Final Government Distribution Atlas IPCC AR6 WGI
1 Table of Contents
2
3 Executive summary ......................................................................................................................................... 8
4 Atlas.1 Introduction .................................................................................................................................... 11
5 Atlas.1.1 Purpose ...................................................................................................................................11
6 Atlas.1.2 Context and framing ...............................................................................................................11
7 Atlas.1.3 Defining temporal and spatial scales and regions ...................................................................12
8 Atlas.1.3.1 Baselines and temporal scales of analysis for projections across scenarios ...........................12
9 Atlas.1.3.2 Global warming levels ............................................................................................................13
10 Atlas.1.3.3 Spatial scales and reference regions .......................................................................................14
11 Atlas.1.3.4 Typological and socio-economic regions ...............................................................................15
12 Atlas.1.4 Combining multiple sources of information for regions .........................................................16
13 Atlas.1.4.1 Observations ...........................................................................................................................16
14 Atlas.1.4.2 Reanalysis ...............................................................................................................................17
15 Atlas.1.4.3 Global model data (CMIP5 and CMIP6) ................................................................................17
16 Atlas.1.4.4 Regional model data (CORDEX) ...........................................................................................18
17 BOX ATLAS.1: .................................................................................................................... CORDEX-CORE
18 .......................................................................................................................................................... 20
19 Atlas.1.4.5 Bias adjustment .......................................................................................................................21
20 Cross-Chapter Box Atlas.1: ...............................................Displaying robustness and uncertainty in maps
21 .......................................................................................................................................................... 21
22 Atlas.2 The online ‘Interactive Atlas’ ....................................................................................................... 26
23 Atlas.2.1 Why an interactive online Atlas in AR6? ...............................................................................27
24 Atlas.2.2 Description of the Interactive Atlas: functionalities and datasets ...........................................27
25 Atlas.2.3 Accessibility, reproducibility and reusability (FAIR principles) ............................................29
26 Atlas.2.4 Guidance for users ..................................................................................................................31
27 Atlas.2.4.1 Purpose of the Interactive Atlas ..............................................................................................31
28 Atlas.2.4.2 Guidelines for the Interactive Atlas ........................................................................................31
29 Atlas.2.4.2.1 Quantitative support for assessments ......................................................................................31
30 Atlas.2.4.2.2 Insights from physical understanding .....................................................................................32
31 Atlas.2.4.2.3 Construction of storylines .......................................................................................................32
32 Atlas.2.4.2.4 Visual information ..................................................................................................................32
33 Atlas.2.4.2.5 Dedicated climate change assessment programs ....................................................................33
34 Atlas.3 Global synthesis .............................................................................................................................. 33
35 Atlas.3.1 Global atmosphere and land surface .......................................................................................34
36 Atlas.3.2 Global ocean ...........................................................................................................................40
37 Atlas.4 Africa ............................................................................................................................................... 41
Do Not Cite, Quote or Distribute Atlas-3 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.4.1 Key features of the regional climate and findings from previous IPCC assessments .............41
2 Atlas.4.1.1 Key features of the regional climate .......................................................................................41
3 Atlas.4.1.2 Findings from previous IPCC assessments .............................................................................41
4 Atlas.4.2 Assessment and synthesis of observations, trends and attribution..........................................42
5 Atlas.4.3 Assessment of model performance .........................................................................................43
6 Atlas.4.4 Assessment and synthesis of projections ................................................................................44
7 Atlas.4.5 Summary .................................................................................................................................45
8 Atlas.5 Asia .................................................................................................................................................. 46
9 Atlas.5.1 East Asia .................................................................................................................................47
10 Atlas.5.1.1 Key features of the regional climate and findings from previous IPCC assessments .............47
11 Atlas.5.1.1.1 Key features of the regional climate .......................................................................................47
12 Atlas.5.1.1.2 Findings from previous IPCC assessments .............................................................................47
13 Atlas.5.1.2 Assessment and synthesis of observations, trends and attribution..........................................47
14 Atlas.5.1.3 Assessment of model performance .........................................................................................48
15 Atlas.5.1.4 Assessment and synthesis of projections ................................................................................49
16 Atlas.5.1.5 Summary .................................................................................................................................50
17 Atlas.5.2 North Asia ...............................................................................................................................51
18 Atlas.5.2.1 Key features of the regional climate and findings from previous IPCC assessments .............51
19 Atlas.5.2.1.1 Key features of the regional climate .......................................................................................51
20 Atlas.5.2.1.2 Findings from previous IPCC assessments .............................................................................51
21 Atlas.5.2.2 Assessment and synthesis of observations, trends and attribution..........................................51
22 Atlas.5.2.3 Assessment of model performance .........................................................................................52
23 Atlas.5.2.4 Assessment and synthesis of projections ................................................................................53
24 Atlas.5.2.5 Summary .................................................................................................................................54
25 Atlas.5.3 South Asia ...............................................................................................................................54
26 Atlas.5.3.1 Key features of the regional climate and findings from IPCC previous assessments .............54
27 Atlas.5.3.1.1 Key features of the regional climate .......................................................................................54
28 Atlas.5.3.1.2 Findings from previous IPCC assessments .............................................................................55
29 Atlas.5.3.2 Assessment and synthesis of observations, trends and attribution..........................................55
30 Atlas.5.3.3 Assessment of model performance .........................................................................................56
31 Atlas.5.3.4 Assessment and synthesis of projections ................................................................................57
32 Atlas.5.3.5 Summary .................................................................................................................................58
33 Atlas.5.4 Southeast Asia.........................................................................................................................58
34 Atlas.5.4.1 Key features of the regional climate and findings from previous IPCC assessments .............58
35 Atlas.5.4.1.1 Key features of the regional climate .......................................................................................58
36 Atlas.5.4.1.2 Findings from previous IPCC assessments .............................................................................59
37 Atlas.5.4.2 Assessment and synthesis of observations, trends and attribution..........................................59
38 Atlas.5.4.3 Assessment of model performance .........................................................................................60
39 Atlas.5.4.4 Assessment and synthesis of projections ................................................................................60
Do Not Cite, Quote or Distribute Atlas-4 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.5.4.5 Summary .................................................................................................................................61
2 Atlas.5.5 Southwest Asia .......................................................................................................................61
3 Atlas.5.5.1 Key features of the regional climate and findings from previous IPCC assessments .............61
4 Atlas.5.5.1.1 Key features of the regional climate .......................................................................................61
5 Atlas.5.5.1.2 Findings from previous IPCC assessments .............................................................................62
6 Atlas.5.5.2 Assessment and synthesis of observations, trends and attribution..........................................62
7 Atlas.5.5.3 Assessment of model performance .........................................................................................63
8 Atlas.5.5.4 Assessment and synthesis of projections ................................................................................64
9 Atlas.5.5.5 Summary .................................................................................................................................65
10 Atlas.6 Australasia ...................................................................................................................................... 66
11 Atlas.6.1 Key features of the regional climate and findings from previous IPCC assessments .............66
12 Atlas.6.1.1 Key features of the regional climate .......................................................................................66
13 Atlas.6.1.2 Findings from previous IPCC assessments .............................................................................66
14 Atlas.6.2 Assessment and synthesis of observations, trends and attribution..........................................67
15 Atlas.6.3 Assessment of climate model performance ............................................................................68
16 Atlas.6.4 Assessment and synthesis of projections ................................................................................68
17 Atlas.6.5 Summary .................................................................................................................................70
18 Atlas.7 Central and South America .......................................................................................................... 70
19 Atlas.7.1 Central America and the Caribbean ........................................................................................71
20 Atlas.7.1.1 Key features of the regional climate and findings from previous IPCC assessments .............71
21 Atlas.7.1.1.1 Key features of the regional climate .......................................................................................71
22 Atlas.7.1.1.2 Findings from previous IPCC assessments .............................................................................72
23 Atlas.7.1.2 Assessment and synthesis of observations, trends and attribution..........................................72
24 Atlas.7.1.3 Assessment of model performance .........................................................................................72
25 Atlas.7.1.4 Assessment and synthesis of projections ................................................................................73
26 Atlas.7.1.5 Summary .................................................................................................................................74
27 Atlas.7.2 South America ........................................................................................................................75
28 Atlas.7.2.1 Key features of the regional climate and findings from previous IPCC assessments .............75
29 Atlas.7.2.1.1 Key features of the regional climate .......................................................................................75
30 Atlas.7.2.1.2 Findings from previous IPCC assessments .............................................................................75
31 Atlas.7.2.2 Assessment and synthesis of observations, trends and attribution..........................................76
32 Atlas.7.2.3 Assessment of model performance .........................................................................................77
33 Atlas.7.2.4 Assessment and synthesis of projections ................................................................................78
34 Atlas.7.2.5 Summary .................................................................................................................................79
35 Atlas.8 Europe ............................................................................................................................................. 79
36 Atlas.8.1 Key features of the regional climate and findings from previous IPCC assessments .............80
37 Atlas.8.1.1 Key features of the regional climate .......................................................................................80
38 Atlas.8.1.2 Findings from previous IPCC assessments .............................................................................80
Do Not Cite, Quote or Distribute Atlas-5 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.8.2 Assessment and synthesis of observations, trends and attribution..........................................81
2 Atlas.8.3 Assessment of model performance .........................................................................................83
3 Atlas.8.4 Assessment and synthesis of projections ................................................................................84
4 Atlas.8.5 Summary .................................................................................................................................86
5 Atlas.9 North America ................................................................................................................................ 86
6 Atlas.9.1 Key features of the regional climate and findings from previous IPCC assessments .............87
7 Atlas.9.1.1 Key features of the regional climate .......................................................................................87
8 Atlas.9.1.2 Findings from previous IPCC assessments .............................................................................87
9 Atlas.9.2 Assessment and synthesis of observations, trends, and attribution.........................................88
10 Atlas.9.3 Assessment of model performance .........................................................................................89
11 Atlas.9.4 Assessment and synthesis of projections ................................................................................90
12 Atlas.9.5 Summary .................................................................................................................................92
13 Atlas.10 Small islands ................................................................................................................................... 93
14 Atlas.10.1 Key features of the regional climate and findings from previous IPCC assessments .............93
15 Atlas.10.1.1 Key features of the regional climate .......................................................................................93
16 Atlas.10.1.2 Findings from previous IPCC assessments .............................................................................93
17 Atlas.10.2 Assessment and synthesis of observations, trends and attribution..........................................93
18 Atlas.10.3 Assessment of model performance .........................................................................................95
19 Atlas.10.4 Assessment and synthesis of projections ................................................................................95
20 Atlas.10.5 Summary .................................................................................................................................97
21 Cross-Chapter Box Atlas.2: ................ Climate information relevant to water resources in Small Islands
22 .......................................................................................................................................................... 97
23 Atlas.11 Polar regions ................................................................................................................................. 101
24 Atlas.11.1 Antarctica ..............................................................................................................................102
25 Atlas.11.1.1 Key features of the regional climate and findings from previous IPCC assessments ...........102
26 Atlas.11.1.1.1 Key features of the regional climate ...............................................................................102
27 Atlas.11.1.1.2 Findings from previous IPCC assessments.....................................................................103
28 Atlas.11.1.2 Assessment and synthesis of observations, trends and attribution........................................103
29 Atlas.11.1.3 Assessment of model performance .......................................................................................105
30 Atlas.11.1.4 Assessment and synthesis of projections ..............................................................................107
31 Atlas.11.1.5 Summary ...............................................................................................................................108
32 Atlas.11.2 Arctic ....................................................................................................................................109
33 Atlas.11.2.1 Key features of the regional climate and findings from previous IPCC assessments ...........109
34 Atlas.11.2.1.1 Key features of the regional climate ...............................................................................109
35 Atlas.11.2.1.2 Findings from previous IPCC assessments.....................................................................109
36 Atlas.11.2.2 Assessment and synthesis of observations, trends and attribution........................................110
37 Atlas.11.2.3 Assessment of model performance .......................................................................................111
38 Atlas.11.2.4 Assessment and synthesis of projections ..............................................................................112
Do Not Cite, Quote or Distribute Atlas-6 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.11.2.5 Summary ...............................................................................................................................114
2 Atlas.12 Final remarks ............................................................................................................................... 114
3 References .................................................................................................................................................... 116
4 Figures ........................................................................................................................................................ 161
5
6
Do Not Cite, Quote or Distribute Atlas-7 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Executive summary
2
3 This Atlas chapter assesses changes in mean climate at regional scales, in particular observed trends and
4 their attribution and projected future changes. The main focus is on changes in temperature and precipitation
5 (including snow and derived variables in polar regions) over land regions, though other variables, including
6 for oceanic regions, are also discussed. Projected changes are presented both as relative to levels of global
7 warming and for future time periods under a range of emissions scenarios. In order to facilitate summarizing
8 assessment findings, a new set of WGI reference regions is used within the chapter which were derived
9 following broad consultation and peer review. These are used in other chapters for summarizing regional
10 information. This includes the assessment of climatic impact-driver changes in Chapter 12, which
11 incorporates the changes in mean climate assessed in the Atlas. Another important new development since
12 AR5 is the AR6 WGI Interactive Atlas, which is described in this chapter and is used to generate results both
13 for the Atlas and other regional chapters. It is also a resource allowing exploration of datasets underpinning
14 assessment findings in other chapters of the report.
15
16 Observed trends and projections in regional climate
17
18 Most land areas have warmed faster than the global average (high confidence) and very likely by at
19 least 0.1℃ per decade since 1960. A surface temperature change signal has likely emerged over all
20 land areas. Many areas very likely warmed faster since the 1980s, including areas of northern, eastern
21 and south-western Africa, Australia, Central America, Amazonia and West Antarctica (0.2°C–0.3°C
22 per decade), the Arabian Peninsula, Central and East Asia and Europe (0.3°C–0.5°C per decade) and
23 Arctic and near-Arctic land regions (up to 1°C per decade, or more in a few areas). {Figure Atlas.11,
24 Interactive Atlas, Sections Atlas.3.1, Atlas.4.2, Atlas.5.1.2, Atlas.5.2.2, Atlas.5.3.2, Atlas.5.4.2, Atlas.5.5.2,
25 Atlas.6.1.2, Atlas.6.2.2, Atlas.7.2, Atlas.8.2, Atlas.9.2, Atlas.10.2, Atlas.11.1.2, Atlas.11.2.2}
26
27 Significant positive trends in precipitation have been observed in most of North Asia, parts of West
28 Central Asia, South-eastern South America, Northern Europe, Eastern North America, Western
29 Antarctica and the Arctic (medium confidence). Significant negative trends have been observed in the
30 Horn of Africa and southwest Western Australia (high confidence), parts of the Russian Far East,
31 some parts of the Mediterranean and of the Caribbean, Southeast and Northeast Brazil and southern
32 Africa (medium confidence), with the last attributed to anthropogenic warming of the Indian Ocean. In
33 the many other land areas there are no significant trends in annual precipitation over the period 1960–2015
34 though increases in average precipitation intensity have been observed in the Sahel and Southeast Asia
35 (medium confidence). {Figure Atlas.11, Interactive Atlas, Sections Atlas.3.1, Atlas.4.2, Atlas.5.1.2,
36 Atlas.5.2.2, Atlas.5.3.2, Atlas.5.4.2, Atlas.5.5.2, Atlas.6.1.2, Atlas.6.2.2, Atlas.7.2, Atlas.8.2, Atlas.9.2,
37 Atlas.10.2, Atlas.11.1.2, Atlas.11.2.2}.
38
39 The observed warming trends are projected to continue over the 21st century (high confidence) and
40 over most land regions at a rate higher than the global average. At a global warming level of 4°C (i.e.
41 relative to an 1850-1900 baseline) it is likely that most land areas will experience a further warming
42 (from a 1995–2014 baseline) of at least 3°C and in some areas significantly more, including increases of
43 4°C–6°C in the Sahara/Sahel, Southwest, Central and Northern Asia; Northern South America and
44 Amazonia, West Central and Eastern Europe; and Western, Central and Eastern North America, and
45 up to 8°C or more in some Arctic regions. Across each of the continents, higher warming is likely to occur
46 in northern Africa, the central interior of southern and Western Africa; in northern Asia; in Central Australia;
47 in Amazonia; in northern Europe and northern North America (high confidence). Ranges of regional
48 warming for global warming levels of 1.5ºC, 2ºC, 3ºC and 4ºC and for other time periods and emissions
49 scenarios are available in the Interactive Atlas from Coupled Model Intercomparison Project Phases 5 and 6
50 (CMIP5, CMIP6) and Coordinated Regional Climate Downscaling Experiment (CORDEX) projections.
51 {Figure Atlas.12, Interactive Atlas, Sections Atlas.4.4, Atlas.5.1.4, Atlas.5.2.4, Atlas.5.3.4, Atlas.5.4.4,
52 Atlas.5.5.4, Atlas.6.4, Atlas.7.4, Atlas.8.4, Atlas.9.4, Atlas.10.4, Atlas.11.4}
53
54 For given global warming levels, model projections from CMIP6 show future regional warming and
55 precipitation changes that are similar to those projected by CMIP5. However, the larger climate
Do Not Cite, Quote or Distribute Atlas-8 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 sensitivity in some CMIP6 models and differences in the model forcings lead to a wider range of and
2 higher projected regional warming in CMIP6 compared to CMIP5 projections for given time periods
3 and emissions scenarios. {Figure Atlas.13, Sections Atlas.4.4, Atlas.5.1.4, Atlas.5.2.4, Atlas.5.3.4,
4 Atlas.5.4.4, Atlas.5.5.4, Atlas.6.1.4, Atlas.6.2.4, Atlas.7.4, Atlas.8.4, Atlas.9.4, Atlas.10.4, Atlas.11.1.4,
5 Atlas.11.2.4}
6
7 Precipitation will change in most regions, either through changes in mean values or the characteristics
8 of rainy seasons or daily precipitation statistics (high confidence). Regions where annual precipitation
9 is likely to increase include the Ethiopian Highlands; East, South and North Asia; Southeast South
10 America; northern Europe; northern and eastern North America and the polar regions. Regions
11 where annual precipitation is likely to decrease include northern and southwest southern Africa and
12 the Sahel, Indonesia, northern Arabian Peninsula, southwest Australia, Central America, southwest
13 South America and southern Europe. Changes in monsoons are likely to result in increased precipitation in
14 northern China and in South Asia in summer (high confidence). Precipitation intensity will increase in many
15 areas, including in some where annual mean reductions are likely (e.g., Southern Africa) (high confidence).
16 Ranges of regional mean precipitation change for global warming levels of 1.5°C, 2°C, 3°C and 4°C and for
17 other time periods and emissions scenarios are available in Interactive Atlas from CMIP5, CORDEX and
18 CMIP6 projections. {Figure Atlas.13, Interactive Atlas, Sections Atlas.4.4, Atlas.5.1.4, Atlas.5.2.4,
19 Atlas.5.3.4, Atlas.5.4.4, Atlas.5.5.4, Atlas.6.1.4, Atlas.6.2.4, Atlas.7.4, Atlas.8.4, Atlas.9.4, Atlas.10.4,
20 Atlas.11.1.4, Atlas.11.2.4}
21
22 Cryosphere, Polar Regions and Small Islands
23
24 Many aspects of the cryosphere either have seen significant changes in the recent past or will see them
25 during the 21st century (high confidence). Snow cover duration has very likely reduced over Siberia
26 and Eastern and Northern Europe. Also, it is virtually certain that snow cover will experience a decline
27 in these regions and over most of North America during the 21st century, in terms of water equivalent,
28 extent and annual duration. Over the Hindu Kush-Himalaya, glacier mass is likely to decrease
29 considerably (nearly 50%) under the RCP4.5 and RCP8.5 scenarios. Snow cover has declined over
30 Australia as has annual maximum snow mass over North America (medium confidence). Some high-latitude
31 regions have experienced increases in winter snow (parts of northern Asia, medium confidence) or will do so
32 in the future (very likely in parts of northern North America) due to the effect of increased snowfall
33 prevailing over warming-induced increased snowmelt. {Sections 2.3.2.2, 3.4.2, Atlas.5.2.2, Atlas.5.3.4,
34 Altas.6.2, Atlas.8.2, Atlas.8.4, Atlas.9.2, Atlas.9.4}
35
36 It is very likely that the Arctic has warmed at more than twice the global rate over the past 50 years
37 and that the Antarctic Peninsula experienced a strong warming trend starting in 1950s. It is likely that
38 Arctic annual precipitation has increased, with the highest increases during the cold season. Antarctic
39 precipitation and surface mass balance showed a significant positive trend over the 20th century, while
40 strong interannual variability masks any existing trend over recent decades1 (medium confidence).
41 Significant warming trends are observed in other West Antarctic regions and at selected stations in East
42 Antarctica since the 1950s (medium confidence). Under all assessed emission scenarios, both Polar regions
43 are very likely to have higher annual mean surface air temperatures and more precipitation, with temperature
44 increases higher than the global mean, most prominently in the Arctic. {Sections Atlas.11.1.2, Atlas.11.1.4,
45 Atlas.11.2.2, Atlas.11.2.4}
46
47 It is very likely that most Small Islands have warmed over the period of instrumental records.
48 Precipitation has likely decreased since the mid-20th century in some parts of the Pacific poleward of
49 20° latitude in both hemispheres and in the Caribbean in June-July-August. It is very likely that sea
50 levels will continue to rise in Small Island regions and that this will result in increased coastal flooding.
1
The term ‘recent decades’ refers to a period of approximately 30 to 40 years which ends within the period 2010–2020. This is used as many studies
in the literature will analyse datasets over a range of climatologically significant periods (i.e., 30 years or more) with precise start and end dates and
periods depending on data availability and the year of the study. An equivalent approximate description using specific years would be ‘since the
1980s’.
Do Not Cite, Quote or Distribute Atlas-9 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Observed temperature trends are generally in the range of 0.15°C–0.2°C per decade. Rainfall trends in most
2 other Pacific and Indian Ocean Small Islands are mixed and largely non-significant. There is limited
3 evidence and low agreement on the cause of the Caribbean drying trend, though it is likely that both this and
4 the Pacific drying trends will continue in coming decades with drying also projected in the part of the West
5 Indian and Atlantic Oceans. Small Islands regions in the Western and equatorial Pacific and North Indian
6 Ocean are likely to be wetter in the future. {Cross-Chapter Box Atlas.2, Sections Atlas.10.2, Atlas.10.4}
7
8 Model Evaluation, Technical Infrastructure and the Interactive Atlas
9
10 The regional performance of CMIP6 global climate models (GCMs) has improved overall compared to
11 CMIP5 in simulating mean temperature and precipitation, though large errors still exist in some
12 regions. In particular, improvements have been seen over Africa which has belatedly become a focus
13 for GCM model development. Other specific improvement include over East Asia for temperature and the
14 winter monsoon, over parts of South Asia for the summer monsoon, over Australia (including influences of
15 modes of variability), in simulation of Antarctic temperatures and Arctic sea ice. Notable errors include large
16 cold biases in mountain ranges in South Asia, a significant wet bias over Central Asia, in the East Asia
17 summer monsoon and in Antarctic precipitation. An in-depth evaluation of CMIP6 models is lacking for
18 several regions (North and Southeast Asia, parts of West Central Asia, Central and South America), though
19 CMIP5 models have been evaluated for many of these. {Sections Atlas.4.3, Atlas.5.1.3, Atlas.5.2.3,
20 Atlas.5.3.3, Atlas.5.4.3, Atlas.5.5.3, Atlas.6.1.3, Atlas.6.2.3, Atlas.7.3, Atlas.8.3, Atlas.9.3, Atlas.10.3,
21 Atlas.11.1.3, Atlas.11.2.3}
22
23 Since AR5, the improvement in regional climate modelling and the growing availability of regional
24 simulations through coordinated dynamical downscaling initiatives such as CORDEX, have advanced
25 the understanding of regional climate variability, adding value to CMIP global models, particularly in
26 complex topography zones, coastal areas and small islands, and in the representation of extremes (high
27 confidence). In particular, regional climate models with polar-optimized physics are important for estimating
28 the regional and local surface mass balance and are improved compared to reanalyses and GCMs when
29 evaluated with observations (high confidence). There is still a lack of high-quality and high-resolution
30 observational data to assess observational uncertainty in climate studies, and this compromises the ability to
31 evaluate models (high confidence). {Sections Atlas.4.3, Atlas.5.1.3, Atlas.5.2.3, Atlas.5.3.3, Atlas.5.4.3,
32 Atlas.5.5.3, Atlas.6.1.3, Atlas.6.2.3, Atlas.7.3, Atlas.8.3, Atlas.9.3, Atlas.10.3, Atlas.11.1.3, Atlas.11.2.3}
33
34 Significant improvements in technical infrastructure, open tools and methodologies for accessing and
35 analysing observed and simulated climate data, and the progressive adoption of FAIR (findability,
36 accessibility, interoperability, and reusability) data principles have very likely broadened the ability to
37 interact with these data for a wide range of activities, including fundamental climate research,
38 providing inputs into assessments of impacts, building resilience and developing adaptations. Tools to
39 analyse and assess climate information have improved to allow development of information that goes beyond
40 averages (e.g., on future climate thresholds and extremes) and that is relevant for regional climate risk
41 assessments. {Sections Atlas.2.2, Atlas.2.3}
42
43 The Interactive Atlas is a new WGI product developed to take advantage of the interactivity offered
44 by web applications by allowing flexible and expanded exploration of some key products underpinning
45 the assessment (including extreme indices and climatic impact-drivers). This provides a transparent
46 interface for access to authoritative IPCC results, facilitating their use in applications and climate services.
47 The Interactive Atlas implements FAIR principles and builds on open tools and therefore is an important step
48 towards making IPCC results more reproducible and reusable. {Section Atlas.2 and Interactive Atlas}
49
50
Do Not Cite, Quote or Distribute Atlas-10 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.1 Introduction
2
3 Atlas.1.1 Purpose
4
5 The Atlas is the final chapter of this Working Group I (WGI) Sixth Assessment Report (AR6) and comprises
6 the Atlas Chapter and an online interactive tool, the Interactive Atlas. The Atlas assesses fundamental
7 aspects of observed, attributed and projected changes in regional climate in coordination with other WGI
8 chapters (Chapters 2, 3, 4, 6, 8, 9, 10, 11, 12). In particular, it provides analyses and assessments of regional
9 changes in mean climate (specifically surface temperature, precipitation and some cryospheric variables,
10 such as snow cover and surface mass balance) and expands on and integrates results from other chapters
11 across different spatial and temporal scales. The Atlas considers multiple lines of evidence including
12 assessment of different global and regional observational datasets, attribution of observed trends and
13 multiple model simulations from the Coupled Model Intercomparison Projects CMIP5 (Taylor et al., 2012a)
14 and CMIP6 (Eyring et al., 2016; O’Neill et al., 2016), and the COordinated Regional Downscaling
15 EXperiment (CORDEX) (Gutowski Jr. et al., 2016). The Atlas chapter also assesses model performance and
16 summarizes cross-referenced findings from other chapters relevant for the different regions.
17
18 The Interactive Atlas allows for a flexible spatial and temporal analysis of the results presented in the Atlas
19 and other Chapters, supporting and expanding on their assessments. In particular, the Interactive Atlas
20 includes information from global observational datasets (and paleoclimate information) assessed in Chapter
21 2, and projections of relevant extreme indices (used in Chapter 11) and climatic impact-drivers (used in
22 Chapter 12) allowing for a regional analysis of the results (Section Atlas.2.2). It provides information on
23 climatic impact-drivers relevant to sectoral and regional chapters of the Working Group II (WGII) report,
24 being informed by and complementing the work of Chapter 12 in creating a bridge to WGII. Similarly, a
25 specific aim of the integration is synthesising information drawn from across multiple chapters that is
26 relevant to the WGII report and the mitigation and sectoral chapters of the Working Group III (WGIII)
27 report.
28
29 An overview of the main components of the Atlas chapter is provided in Figure Atlas.1. The Interactive
30 Atlas is described in Section Atlas.2 and is available online at http://ipcc-atlas.ifca.es.
31
32
33 [START FIGURE ATLAS.1 HERE]
34
35 Figure Atlas.1: The main components of the Atlas chapter with, upper right, a screenshot from the online Interactive
36 Atlas.
37
38 [END FIGURE ATLAS.1 HERE]
39
40
41 Atlas.1.2 Context and framing
42
43 Information on global and regional climate change in the form of maps, tables, graphs and infographics has
44 always been a key output of IPCC reports. With the consensus that climate has changed and will continue to
45 do so, policymakers are focusing more on understanding its implications, which often requires an increase in
46 regional and temporal details of observed and future climate. The WGI contribution of the AR5 included a
47 globally comprehensive coverage of land regions and some oceanic regions in the Atlas of Global and
48 Regional Climate Projections (IPCC, 2013a), focusing on projected changes in temperature and
49 precipitation. In the WGII contribution, Chapter 21, Regional Context (Hewitson et al., 2014) included
50 continental scale maps of observed and future temperature and precipitation changes, subcontinental changes
51 in high percentiles of daily temperature and precipitation, and a table of changes in extremes over
52 subcontinental regions (updating an assessment in the Special Report on Managing the Risks of Extreme
53 Events and Disasters to Advance Climate Change Adaptation; SREX). However, there was only limited
54 coordination between these two contributions despite the largely common data sources and their relevance
55 across the two working groups and to wider communities of climate change-related policy and practice. This
Do Not Cite, Quote or Distribute Atlas-11 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 resulted in inefficiencies and the potential for confusing or inconsistent information. The Atlas, with its links
2 with other WGI/II/III chapters, has been designed to help address this.
3
4 Given the aims of the Atlas, there are several important factors to consider. There is a clear requirement for
5 climate change information over a wide range of ‘regions’, and classes thereof, and temporal scales. There is
6 also often the need for integrated information relevant for policy, practice and awareness raising. However,
7 most other chapters in WGI are disciplinary, focusing on specific processes in the climate system or on its
8 past or future behaviour, and have limited space to be spatially and temporally comprehensive. The Atlas
9 provides an opportunity to facilitate this integration and exploration of information.
10
11 Developing this information often requires a broad range of data sources (various observations, global and
12 regionally downscaled baselines and projections) to be analysed and combined and, where appropriate,
13 reconciled. This is a topic which is assessed from a methodological perspective in Chapter 10 using a limited
14 set of examples (see also Cross-Chapter Box 10.3). The Atlas then builds on this work with a more
15 comprehensive treatment of the available results, largely (but not exclusively) based on CMIP5, CMIP6, and
16 CORDEX, to provide wider coverage and to further demonstrate techniques and issues. These multiple lines
17 of evidence are integrated in the Interactive Atlas, a new AR6 WGI product described in Section Atlas.2
18 allowing for flexible spatial and temporal analysis of this information with a predefined granularity (e.g.,
19 flexible seasons, regions, and baselines and future periods of analysis including time-slices and warming
20 levels).
21
22 Generating information relevant to policy or practice requires understanding the context of the systems that
23 they focus on. In addition to the hazards these systems face, their vulnerability and exposure, and the related
24 socio-economic and other physical drivers, also need to be understood. To ensure this relevance, the Atlas is
25 informed by the assessments in Chapter 12 and the regional and thematic chapters and cross-chapter papers
26 of WGII. Therefore, it focuses on generating information on climatic impact-drivers and hazards applicable
27 to assessing impacts on and risks to human and ecological systems whilst noting the potential relevance of
28 these to related contexts such as the Sustainable Development Goals and the Sendai Framework for Disaster
29 Risk Reduction.
30
31 Transparency and reproducibility are promoted in the Atlas chapter implementing FAIR principles for
32 Findability, Accessibility, Interoperability, and Reusability of data (Wilkinson et al., 2016). More
33 specifically, the Interactive Atlas provides full metadata of the displayed products (describing both the
34 underlying datasets and the applied postprocessing) and most of the figures included in the Atlas chapter can
35 be reproduced using the scripts and data provided in the WGI-Atlas repository (see Iturbide et al., 2021 and
36 https://github.com/IPCC-WG1/Atlas).
37
38
39 Atlas.1.3 Defining temporal and spatial scales and regions
40
41 Over the past decades scientists have engaged in a wide array of investigations aimed at quantifying and
42 understanding the state of the components of the land surface-ocean-atmosphere system, the complex nature
43 of their interactions and impacts over different temporal and spatial scales. As a result, a great deal has been
44 learned about the importance of an appropriate choice of these scales when estimating changes due to
45 internal climate variability, trends, characterization of the spatiotemporal variability and quantifying the
46 range of and establishing confidence in climate projections. It is therefore important to be able to explore a
47 whole range of spatial and temporal scales and this section presents the basic definitions of those, and the
48 domains of analysis, used by the Atlas accounting for potential synergies between WGI and WGII.
49
50
51 Atlas.1.3.1 Baselines and temporal scales of analysis for projections across scenarios
52
53 Chapter 1 has extensively explored this topic in Section 1.4.1 and Cross-Chapter Box 1.2. A summary of the
54 main points relevant to the Atlas chapter and the Interactive Atlas are provided here.
55
Do Not Cite, Quote or Distribute Atlas-12 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 There is no standard baseline in the literature although the World Meteorological Organization (WMO)
2 recommends a 30-year baseline approach such as the climate normal period 1981–2010. However, it retains
3 the 1961–1990 period as the historical baseline for the sake of supporting long-term climate change
4 assessments (WMO, 2017). Using the WMO standards also provides sample sizes relevant to calculating
5 changes in statistics other than the mean. The AR6 WGI has established the 1995–2014 period as recent past
6 baseline period – for similar reasons to the 1986–2005 period used in AR5 WGI (IPCC, 2013b) – since 2014
7 (2005) is the final year of the historical simulations of the models (more details in Cross-Chapter Box 1.2).
8
9 The choice of a baseline can significantly influence the analysis results for future changes in mean climate
10 (Hawkins and Sutton, 2016) (Cross-Chapter Box 1.2) as well as its variability and extremes. Thus, assessing
11 the sensitivity of results to the baseline period is important. The Interactive Atlas (see Section Atlas.2) allows
12 users to explore and investigate a wide range of different baseline periods when analysing changes for future
13 time-slices or global warming levels:
14
15 • 1995–2014 period (AR6 20-year baseline),
16 • 1986–2005 period (AR5 20-year baseline),
17 • 1981–2010 period (WMO 30-year climate normal),
18 • 1961–1990 period (WMO 30-year long-term climate normal).
19 • 1850–1900 period (baseline used in the calculation of global warming levels).
20
21 This promotes cross-Working Group consistency and facilitates comparability with previous reports and
22 across datasets. For instance, the AR5 and long-term WMO baselines facilitate the intercomparison of
23 CMIP5, CORDEX and CMIP6 projections since all have historical simulations in these periods. Using more
24 recent baselines introduces discontinuity for the CMIP5 and CORDEX models, since historical simulations
25 end in 2005. A pragmatic approximation to deal with this issue is to use scenario data to fill the missing
26 segment, for example for 2006 to 2014 use the first years of RCP8.5-driven transient projections in which
27 the emissions are close to those observed. This approach is used in the Atlas chapter and Chapter 12.
28
29 When assessing changes over the recent past, many studies analyse datasets using a range of climatologically
30 significant periods (i.e., 30 years or more) with precise start and end dates depending on data availability and
31 the year of the study. To account for this, when generating assessments from this literature the term ‘recent
32 decades’ is used to refer to a period of approximately 30 to 40 years which ends within the period 2010–
33 2020. An equivalent approximate description using specific years would be ‘since the 1980s’.
34
35 Regarding the future reference periods, the Interactive Atlas presents projected global and regional climate
36 changes at near-, mid- and long-term periods, respectively 2021–2040, 2041–2060 and 2081–2100, for a
37 range of emissions scenarios (Section Atlas.1.4.3, Cross-Chapter Box 1.4).
38
39
40 Atlas.1.3.2 Global warming levels
41
42 Noting the approach taken in the recent IPCC Special Report on Global Warming of 1.5°C (SR1.5) above
43 1850–1900 levels (IPCC, 2018b), the Atlas also presents global and regional climate change information at
44 different Global Warming Levels (GWLs, see Cross-Chapter Box 11.1). In particular, to provide policy-
45 relevant climate information and represent the range of outcomes from the emissions scenario and time
46 periods considered, GWLs of 1.5°C, 2°C, 3°C and 4°C are considered. The information is computed from all
47 available scenarios (e.g. only 1.5°C and 2°C GWL information can be computed from projections under the
48 SSP1-2.6 scenario). The Interactive Atlas allows comparison of timings for global warming across the
49 different scenarios and of spatial patterns of change, for example information at 2°C GWL is calculated from
50 SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5 projections (Section 4.2.4).
51
52 To calculate GWL information for the datasets used in the Atlas (CMIP6 and CMIP5; see Section Atlas.1.4),
53 this chapter adopted the procedure used in Cross-Chapter Box 11.1. A model future climate simulation
54 reaches the defined GWL of 1.5°C, 2°C, 3°C or 4°C when its global near-surface air temperature change
55 averaged over successive 20-year periods first attains that level of warming relative to its simulation of the
Do Not Cite, Quote or Distribute Atlas-13 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 1851-1900 climate. (1851–1900 defines the pre-industrial baseline period for calculating the required global
2 surface temperature baseline, Cross-Chapter Box 1.2). Note that this process is different from the one used in
3 the SR1.5 report which used 30-year future periods. If a projection stabilizes before reaching the required
4 threshold it is unable to simulate climate at that GWL and is thus discarded. For CORDEX simulations, the
5 periods of the driving GCM are used, as in Nikulin et al. (2018). Detailed reproducible information on the
6 GWLs used in the Atlas is provided in the Atlas repository (Iturbide et al., 2021).
7
8 Climate information at many temporal scales and over a wide range of temporal averaging periods is
9 required for the assessment of climate change and its implications. These range from annual to multi-decadal
10 averages required to characterise low-frequency variability and trends in climate to hourly or instantaneous
11 maximum or minimum values of impactful climate variables. In between, information on, for example,
12 seasonal rainfall is important and implies the need to include averaging periods whose relevance are
13 geographically dependent. As a result, the Atlas chapter presents results over a wide range of timescales,
14 from daily to decadal, and averaging periods with the Interactive Atlas allowing a choice of user-defined
15 seasons and a range of predefined daily to multi-day climate indices.
16
17
18 Atlas.1.3.3 Spatial scales and reference regions
19
20 Many factors influence the spatial scales and regions over which climate information is required and can be
21 reliably generated. Despite all efforts in researching, analysing and understanding climate and climate
22 change, a key factor in determining spatial scales at which analysis can be undertaken is the availability and
23 reliability of data, both observational and from model simulations. In addition, information is required over a
24 wide range of spatial domains defined either from a climatological or geographical perspective (e.g., a region
25 affected by monsoon rainfall or a river basin) or from a socio-economic or political perspective (e.g., least-
26 developed countries or nation states). Chapter 1 provides an overview of these topics (Section 1.5.2). This
27 subsection discusses some relevant issues, summarizes recent advances in defining domains and spatial
28 scales used by AR6 analyses and how these can be explored with the Interactive Atlas.
29
30 Recent IPCC reports – AR5 Chapter 14 (Christensen et al., 2013) and SR1.5 Chapter 3 (Hoegh-Guldberg et
31 al., 2018) – have summarized information on projected future climate changes over subcontinental regions
32 defined in the SREX report (Seneviratne et al., 2012) and later extended in the AR5 from the 26 regions in
33 SREX to include the polar, Caribbean, two Indian Ocean, and three Pacific Ocean regions (hereafter known
34 as the AR5 WGI reference regions) (Figure Atlas.2a). In recent literature, new subregions have been used,
35 for example for North and South America, Africa and Central America, together with the new definition of
36 reference oceanic regions. Iturbide et al. (2020) describes an updated version of the reference regions which
37 is used in this report (hereafter known as AR6 WGI reference regions) and is shown in Figure Atlas.2b. The
38 goal of these subsequent revisions was to improve the climatic consistency of the regions so they represented
39 sub-continental areas of greater climatic coherency.
40
41
42 [START FIGURE ATLAS.2 HERE]
43
44 Figure Atlas.2: WGI reference regions used in the (a) AR5 and (b) AR6 reports (Iturbide et al., 2020). The latter
45 includes both land and ocean regions and it is used as the standard for the regional analysis of
46 atmospheric variables in the Atlas chapter and the Interactive Atlas. The definition of the regions and
47 companion notebooks and scripts are available at the Atlas repository (Iturbide et al., 2021).
48
49 [END FIGURE ATLAS.2 HERE]
50
51
52 The rationale followed for the definition of the reference regions was guided by two basic principles: 1)
53 climatic consistency and better representation of regional climate features and 2) representativeness of model
54 results (i.e., sufficient number of model grid boxes). The finer resolution of CMIP6 models (as compared, on
55 average, to CMIP5) yields better model representation of the reference regions allowing them to be revised
Do Not Cite, Quote or Distribute Atlas-14 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 for better climatic consistency (e.g., dividing heterogeneous regions) while preserving the model
2 representation. Figure Atlas.3 illustrates this issue displaying the number of grid boxes (over land for land
3 regions) in the AR6 reference regions for two Interactive Atlas reference grids of horizontal resolutions of 1°
4 and 2°, representative of the typical resolution of CMIP6 and CMIP5 models respectively. This figure shows
5 that the new reference regions are well suited for the assessment of model results, with poorest model
6 coverage for the New Zealand (NZ), Caribbean (CAR) and Madagascar (MAD) regions.
7
8
9 [START FIGURE ATLAS.3: HERE]
10
11 Figure Atlas.3: Number of land grid boxes (blue numbers) for the AR6 WGI reference regions for the reference
12 grids representative of (a) CMIP6 and (b) CMIP5, at horizontal resolutions of 1° and
13 2°respectively. Colour shading indicate regions with fewer than 250 grid boxes indicating the number
14 of grid boxes (darkest shading if fewer than 20 grid boxes). The polygons show the AR6 WGI
15 reference regions of Figure Atlas.2.
16
17 [END FIGURE ATLAS.3 HERE]
18
19
20 AR6 WGI (land and open ocean) reference regions are used in the Interactive Atlas as the default
21 regionalization for atmospheric variables. However, these regions are not optimum for the analysis of
22 oceanic variables since, for instance, the five upwelling regions (Canary, California, Peru, Benguela and
23 Somali) are mostly included in ‘land’ regions. Therefore, the alternative set of oceanic regions defined by
24 their biological activity (Figure Atlas.4) is used in the Interactive Atlas for the regional analysis of oceanic
25 variables (see Fay and McKinley, 2014; Gregor et al., 2019). Due to the many potential definitions of the
26 regions relevant for WGI and WGII, some additional typological and socio-economic regions have also been
27 included in the Interactive Atlas.
28
29
30 Atlas.1.3.4 Typological and socio-economic regions
31
32 In addition to contiguous spatial domains discussed in the previous section, some domains are defined by
33 specific climatological, geographical, ecological or socio-economic properties where climate is an important
34 determinant or influencer. In these cases the domains are subject to particular physical processes that are
35 important for its climatology or that involve systems affected by the climate in a way that observations and
36 climate model simulations can be used to understand. Many of these are the basis of the chapters and cross-
37 chapter papers of the AR6 WGII report, namely river basins, biodiversity hotspots, tropical forests, cities,
38 coastal settlements, deserts and semi-arid areas, the Mediterranean, mountains and polar regions. It is
39 therefore important to generate climate information relevant to these typological domains and some
40 examples of these used in the Interactive Atlas are shown in Figure Atlas.4.
41
42
43 [START FIGURE ATLAS.4 HERE]
44
45 Figure Atlas.4: Typological and socio-economic regions used in the Interactive Atlas. (a) Eleven ocean regions
46 defined by their biological activity used for the regional analysis of oceanic variables; (b) ocean
47 regions for small islands, including the Caribbean (CAR) and the North Indian Ocean (ARS and
48 BOB); (c) land monsoon regions of North America, South America, Africa, Asia and Australasia; (d)
49 major river basins; (e) mountain regions; (f) WGII continental regions. These regions can be used
50 alternatively to the reference regions for the regional analysis of climatic variables in the Interactive
51 Atlas. The definition of the regions and companion notebooks and scripts are available at the Atlas
52 repository (Iturbide et al., 2021).
53
54 [END FIGURE ATLAS.4 HERE]
55
56
Do Not Cite, Quote or Distribute Atlas-15 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.1.4 Combining multiple sources of information for regions
2
3 This section introduces the observational data sources and reanalyses that are used in the assessment of
4 regional climate change and for evaluating and bias-adjusting the results of models (more information on
5 observational reference datasets is available in Annex I). It also introduces the different global and regional
6 climate model outputs that are used for regional climate assessment considering both historical and future
7 climate projections (Annex II). Many of these models are run as part of coordinated Model Intercomparison
8 Projects (MIPs), including CMIP5, CMIP6 and CORDEX, described below. Combining information from
9 these multiple data sources is a significant challenge (see Section 10.5 for an in-depth treatment of the
10 problem) though if they can be used to generate robust information on regional climate change it can guide
11 policy and support decisions responding to these changes. An important and necessary part of this process is
12 to check for consistency amongst the data sources which is discussed in the final section.
13
14
15 Atlas.1.4.1 Observations
16
17 There are various sources of observational information available for global and regional analysis.
18 Observational uncertainty is a key factor when assessing and attributing historical trends, so assessment
19 should build on integrated analyses from different datasets (disparity, inadequacy and contradictions in
20 existing datasets are assessed in Section 10.2). The Atlas chapter can supplement and complement Chapter
21 10 by providing the opportunity to visualise and expand on its assessment. This includes displaying maps of
22 density of stations observations (including those that are used in the different datasets) and assessing
23 observational uncertainty by using multiple datasets.
24
25 Two of the most commonly used variables in climate studies are gridded surface air temperature and
26 precipitation. There are many datasets available (Annex I) and Chapter 2 provides an assessment of key
27 global datasets, including blended land-air and sea-surface temperature datasets to assess Global Mean
28 Surface Temperature (GMST). The Atlas analyses separately atmospheric and oceanic variables and for the
29 former a number of common global datasets supporting the assessment done in other chapters is used,
30 including those selected in Chapter 2, but considering land-only information for the blended products. In
31 particular, for air temperature the Atlas uses CRUTEM5 – the land component of the HadCRUT5 dataset –
32 (Osborn et al., 2021), Berkeley Earth (Rohde and Hausfather, 2020) and the Climatic Research Unit CRU
33 TS4 (version 4.04 used here) (Harris et al., 2020). For precipitation the Atlas includes CRU TS4, the Global
34 Precipitation Climatology Centre (GPCC, v2018 used here) (Schneider et al., 2011), and Global Precipitation
35 Climatology Project (GPCP; monthly version 2.3 used here) (Adler et al., 2018). Although the ultimate
36 source of these datasets is surface station reported values (GPCP also includes satellite information), each
37 has access to different numbers of stations and lengths of records and employs different ways of creating the
38 gridded product and ensuring quality control. For oceanic variables, the most widely used sea surface
39 temperature (SST) datasets are HadSST4 (Kennedy et al., 2019), which is the oceanic component of the
40 HadCRUT5 dataset, ERSST (Huang et al., 2017a), and KaplanSST (Kaplan et al., 1998).
41
42 Figure Atlas.5 shows the spatial coverage of the total number of observation stations for different periods
43 (1901–1910, 1971–1980, and 2001–2010) for two illustrative datasets: the CRU TS4.0 dataset for
44 precipitation and the SST data in HadSST4. The former illustrates spatially the declining trend of station
45 observation data used in the precipitation datasets for certain regions (South America, Africa) after the
46 1990s. This demonstrates the regional inhomogeneity and temporal change in station density, which is in
47 part a consequence of many stations not reporting to the WMO networks and their data being held
48 domestically or regionally. During early years a limited number of observations are available. This
49 information is used in the Interactive Atlas to blank out regions not constrained with observations in those
50 datasets providing station density information.
51
52
53
54
55
Do Not Cite, Quote or Distribute Atlas-16 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 [START FIGURE ATLAS.5 HERE]
2
3 Figure Atlas.5: Number of stations per 0.5° x 0.5° grid cell reported over the periods of 1901–1910, 1971–1980, and
4 2001–2010 (rows 1–3) and global total number of stations reported over the entire globe (bottom row)
5 for precipitation in the CRU TS4.0 dataset (left) and the HadSST4 dataset (right). Further details on
6 data sources and processing are available in the chapter data table (Table Atlas.SM.15).
7
8 [END FIGURE ATLAS.5 HERE]
9
10
11 In addition to surface observations, satellites have been widely used to produce rainfall estimates. The
12 advantage of satellite-based rainfall products is their global coverage including remote areas but there is
13 significant uncertainty in these products over complex terrain (Rahmawati and Lubczynski, 2018; Satgé et
14 al., 2019). Another recent development has been on gridded datasets for climate extremes based on surface
15 stations, such as HadEX3 (Dunn et al., 2020).
16
17 There are some studies assessing observational datasets globally (Beck et al., 2017; Sun et al., 2018b) and
18 regionally (Manzanas et al., 2014; Salio et al., 2015; Prakash, 2019), reporting large differences among them
19 and stressing the importance of considering observational uncertainty in regional climate assessment studies.
20 Uncertainty in observations is also a key limitation for the evaluation of climate models, particularly over
21 regions with low station density (Kalognomou et al., 2013; Kotlarski et al., 2019). More detailed information
22 on these issues is provided in Section 10.2.
23
24 For regional studies, observational datasets with global coverage are complemented by a range of regional
25 observational analyses and gridded products, such as E-OBS (Cornes et al., 2018) over Europe, Daymet
26 (Thornton et al., 2016) over North America, or APHRODITE (Yatagai et al., 2012) over Asia. These are
27 highlighted in various other chapters and the Atlas expands on their treatment, complementing discussions
28 on discrepancies/conflicts in observations presented in Chapter 10 and expanding on and replicating their
29 results for other regions. In particular, the Interactive Atlas includes the global and regional observational
30 products described here to assess observational uncertainty over the different regions analysed.
31
32
33 Atlas.1.4.2 Reanalysis
34
35 There are currently many atmospheric reanalysis datasets with different spatial resolution and assimilation
36 algorithms (see Annex I and Section 1.5.2). There are also substantial differences among these datasets due
37 to the types of observations assimilated into the reanalyses, the assimilation techniques that are used, and the
38 resolution of the outputs, amongst other reasons. For example, 20CR (Slivinski et al., 2019) only assimilates
39 surface pressure and sea surface temperature to achieve the longest record but at relatively low resolution,
40 while ERA-20C (Poli et al., 2016) only assimilates surface pressure and surface marine winds. At the other
41 extreme, very sophisticated assimilation systems using multiple surface, upper air and Earth observation data
42 sources are employed, for example ERA5 (Hersbach et al., 2020) and JRA-55 (Harada et al., 2016), which
43 also have much higher resolutions. Most reanalysis datasets cover the entire globe, but there are also high-
44 resolution regional reanalysis datasets which provide further regional detail (Kaiser-Weiss et al., 2019).
45
46 The Atlas and Interactive Atlas use information from ERA5 and from the bias-adjusted version WFDE5
47 (Cucchi et al., 2020) which is combined with ERA5 information over the ocean and used as the ISIMIP
48 observational reference dataset W5E5 (Lange, 2019b). This reference is also used in the Atlas for model
49 evaluation (Section Atlas.1.4.4) and for bias-adjusting model outputs (Section Atlas.1.4.5).
50
51
52 Atlas.1.4.3 Global model data (CMIP5 and CMIP6)
53
54 The Atlas chapter (and the Interactive Atlas) uses global model simulations from both CMIP5 and CMIP6,
55 mainly historical and future projections performed under ScenarioMIP (O’Neill et al., 2016). This facilitates
Do Not Cite, Quote or Distribute Atlas-17 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 backwards comparability and thus the detection of new salient features and findings from recent science and
2 the latest CMIP6 ensemble. The selection of the models is based on availability of scenario data for the
3 variables assessed in the Atlas chapter and for those included in the Interactive Atlas (Section Atlas.2.2). In
4 particular, in order to harmonize the results obtained from the different scenarios as much as possible, only
5 models providing data for the historical scenario and at least two emission scenarios, RCP2.6, RCP4.5 and/or
6 RCP8.5 (for CMIP5) and SSP1-2.6, SSP2-4.5, SSP3-7.0 and/or SSP5-8.5 (for CMIP6), were chosen,
7 resulting in 29 and 35 models, respectively (see Cross-Chapter Box 1.4 for a description of the scenarios). In
8 the Atlas chapter (similarly to the regional Chapters 11 and 12) a single simulation is taken from each model
9 (see Section Atlas.12 for limitations of this choice). Since the RCP and SSP emission scenarios are not
10 directly comparable due to different regional forcing (Section 4.2.2), the Atlas includes Global Warming
11 Levels (GWLs) as an alternative dimension of analysis (see Cross-Chapter Box 11.1), which allows
12 intercomparison of results from different scenarios as an alternative to the standard analysis based on time-
13 slices for particular scenarios (Section Atlas.1.3.1). This dimension allows for enhanced comparability of
14 CMIP5 and CMIP6, since it constrains the regional patterns to the same global warming level for both
15 datasets.
16
17 Building on this information, the Interactive Atlas displays a number of (mean and extreme) indices and
18 climatic impact-drivers (CIDs) considering both atmospheric and oceanic variables (see Section Atlas.2.2).
19 Some of these indices have been selected in coordination with Chapters 11 and 12, in order to support and
20 extend the assessment performed in these chapters (see Annex VI for details on the indices). In order to
21 harmonize this information, the indices have been computed for each individual model on the original model
22 grids and the results have been interpolated to a common 2° (for CMIP5) and 1° (CMIP6) horizontal
23 resolution grids. In addition, for the sake of comparability with CMIP6 results (in particular when using
24 baselines going beyond 2005), the historical period of the CMIP5 and CORDEX datasets has been extended
25 to 2006–2014 using the first years of RCP8.5-driven transient projections (see Section Atlas.1.3.1). Tables
26 listing the CMIP5 and CMIP6 models used in the Atlas and in the Interactive Atlas for different scenarios
27 and variables are included as Supplementary Material (Tables Atlas.SM.1 and Atlas.SM.2, respectively);
28 moreover, full inventories including details on the specific ESGF versions are given in the Atlas GitHub
29 repository (Iturbide et al., 2021).
30
31 Chapter 3 and Flato et al. (2013) describe the evaluation of CMIP6 and CMIP5 models, respectively,
32 assessing surface variables and large-scale indicators. Section 10.3.3, assesses the general capability of
33 GCMs to produce climate output for regions.
34
35 Information from the existing CMIP5 and CMIP6 datasets is supplemented with downscaled regional
36 climate simulations from CORDEX. This facilitates an assessment of the effects from higher resolution
37 including whether this modifies the projected climate change signals compared to global models and adds
38 any value, especially in terms of high-resolution features and extremes.
39
40
41 Atlas.1.4.4 Regional model data (CORDEX)
42
43 Global model data, as generated by the CMIP ensembles, although available globally, have spatial
44 resolutions that are limited for reproducing certain processes and phenomena relevant for regional analysis
45 (around 2° and 1° for CMIP5 and CMIP6, respectively). The Coordinated Regional Climate Downscaling
46 Experiment (CORDEX) (Gutowski Jr. et al., 2016) facilitates worldwide application of Regional Climate
47 Models (RCMs, see Section 10.3.1.2), focusing on a number of regions (see Figure Atlas.6) with a typical
48 resolution of 0.44° (but also at 0.22° and 0.11° over some domains, such as Europe). However, only a few
49 simulations are available for some domains (Annex II, Tables AII.1 and AII.2), thus limiting the level of
50 analysis and assessment that can be performed using CORDEX data in some regions. Moreover, there are
51 regions where several domains overlap, thus providing additional lines of evidence. The use of multi-domain
52 grand ensembles to work globally with CORDEX data have recently been proposed (Legasa et al., 2020;
53 Spinoni et al., 2020). Ongoing efforts, such as the multi-domain CORDEX-CORE simulations are promoting
54 more homogeneous coverage and thus more systematic treatment of CORDEX domains (see Box Atlas.1).
55
Do Not Cite, Quote or Distribute Atlas-18 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 [START FIGURE ATLAS.6 HERE]
2
3 Figure Atlas.6: CORDEX domains showing the curvilinear domain boundaries resulting from the original
4 rotated domains. The topography corresponding to the standard CORDEX 0.44° resolution is shown
5 to illustrate the orographic gradients over the different regions.
6
7 [END FIGURE ATLAS.6 HERE]
8
9
10 A lot of progress has been made by the regional climate modelling community since AR5 (Table AII.2) to
11 produce and make available evaluation (reanalysis-driven) simulations over the different CORDEX domains
12 along with downscaled CMIP5 historical and future climate projection information under a range of emission
13 scenarios, mainly RCP2.6, RCP4.5 and RCP8.5 (Tables AII.3 and AII.4). However, these ensembles cover
14 only a fraction of the uncertainty range spanned by the full CMIP5 ensemble in the different domains (e.g.,
15 Ito et al., 2020b), Figure Atlas.16, Figure Atlas.17, Figure Atlas.21, Figure Atlas.22, Figure Atlas.24, Figure
16 Atlas.26, Figure Atlas.28 and Figure Atlas.29). Therefore, comparison of CMIP5 and CORDEX results
17 should be performed carefully, providing results not only for the full CMIP5 ensemble but also for the sub-
18 ensemble formed by the driving models since results can diverge (Fernández et al., 2019; Iles et al., 2020).
19
20 The Atlas chapter and the Interactive Atlas use CORDEX information for the following eleven individual
21 CORDEX domains (out of the fourteen domains shown in Figure Atlas.6): Northern, Central and South
22 America, Europe, Africa, South, East and Southeast Asia, Australasia, Arctic and Antarctica; in addition,
23 oceanic information has been used from the Mediterranean domain, which provides simulations from
24 coupled atmosphere-ocean regional climate models. In order to harmonize the information across domains
25 and to maximize the size of the resulting ensembles, all the available simulations for each individual
26 CORDEX domain (including the standard 0.44° CORDEX and the 0.22° CORDEX-CORE) have been
27 interpolated to a common regular 0.5°-resolution grid to provide a grand ensemble covering the historical
28 and future emission RCP2.6, RCP4.5 and RCP8.5 scenarios, and also the reanalysis-driven simulations for
29 evaluation purposes. In the case of the European domain, the dataset considered is the 0.11° simulations
30 (CORDEX EUR-11, the same dataset as used in Chapter 12) which has been interpolated to a regular 0.25°
31 resolution grid (the same used for the regional observations). In the case of the Mediterranean domain,
32 oceanic information (sea surface temperature) is interpolated to a regular 0.11º grid. In all cases, the indices
33 are computed on the original grids and the interpolation process is applied to the resulting indices. Moreover,
34 for the sake of comparability with CMIP6 results (in particular when using baseline periods beyond 2005),
35 the historical period of the CORDEX datasets has been extended to 2006–2014 using the first years of
36 RCP8.5-driven transient projections in which the emissions are close to those observed (see Section
37 Atlas.1.3.1); note that this procedure is also applied to CMIP5 simulations.
38
39 For the different CORDEX domains, the full ensembles of models (GCM-RCM matrix) used in the Atlas for
40 the different scenarios and variables are described in the Supplementary Material (Tables Atlas.SM.3 to
41 Atlas.SM.14) and in the Atlas repository (Iturbide et al., 2021), including full metadata relative to ESGF
42 versions used and the periods with data available for the different simulations. In particular, the historical
43 scenario information is only available from 1970 onwards for some models and therefore the common period
44 1970–2005 is used for historical CORDEX data in the Atlas. As a result, the WMO baseline period 1961–
45 1990 is not available in the Interactive Atlas for CORDEX data.
46
47 Sections Atlas.4 to Atlas.11 assess research on CORDEX simulations over different regions, analysing past
48 and present climate as well as future climate projections. They also focus on regional model evaluation in
49 order to extend and complement the validation of global models done in Chapter 3, considering the specific
50 regional climate and relevant large-scale and regional phenomena, drivers and feedbacks (Section 10.3.3).
51 Besides the literature assessment, some simple evaluation diagnostics have been computed for the
52 simulations used in the Atlas chapter to provide some basic information on model performance across
53 regions. In particular biases for mean temperature and precipitation have been calculated for the eleven
54 CORDEX domains analysed.
55
Do Not Cite, Quote or Distribute Atlas-19 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Figure Atlas.7 shows mean temperature and precipitation biases over the North American domain in RCM
2 simulations driven by reanalysis and historical GCM simulations (see Section 10.3.2.5). Annual and seasonal
3 (DJF and JJA) biases are computed for both the RCMs and driving GCMs. Biases in the reanalysis-driven
4 RCMs result from intrinsic model errors, with the results displayed being spatially aggregated for each
5 reference region. This same analysis is performed for the GCM-driven RCM simulations over the historical
6 period 1986–2005. This allows comparison of the intrinsic bias of the RCMs with the biases resulting when
7 driven by the different GCMs and patterns of behaviour in the RCMs, for example intrinsic warm and dry
8 biases in ENA and WNA respectively or reduced RCM warm biases compared to the CCCma GCM in NEN
9 and ENA. Similar results for the other CORDEX domains are included as Supplementary Material (Figures
10 Atlas.SM.1 to Atlas.SM.10).
11
12
13 [START FIGURE ATLAS.7 HERE]
14
15 Figure Atlas.7: Evaluation of annual and seasonal air temperature and precipitation for the six North America
16 subregions NWN, NEN, WNA, CNA, ENA and NCA (land only), for CORDEX-NAM RCM
17 simulations driven by reanalysis or historical GCMs. Seasons are June-July-August (JJA) and
18 December-January-February (DJF). Rows represent subregions and columns correspond to the
19 models. Magenta text indicates the driving historical CMIP5 GCMs (including ERA-Interim in first
20 set of slightly separated columns) and the black text to the right of the magenta text represents the
21 driven RCMs. The colour matrices show the mean spatial biases; all biases have been computed for
22 the period 1985–2005 relative to the observational reference (E5W5, see Section Atlas.1.4.2). Further
23 details on data sources and processing are available in the chapter data table (Table Atlas.SM.15).
24
25 [END FIGURE ATLAS.7 HERE]
26
27
28 [START BOX ATLAS.1 HERE]
29
30 BOX ATLAS.1: CORDEX-CORE
31
32 The main objective of CORDEX-CORE is to provide a global homogeneous foundation of high-resolution
33 RCM projections to improve understanding of local phenomena and facilitate impact and adaptation research
34 worldwide (Gutowski Jr. et al., 2016). The experimental framework is designed to produce homogeneous
35 regional projections for most inhabited land regions using nine CORDEX domains at 0.22° resolution
36 (Figure Atlas.6): North, Central and South America (NAM, CAM, SAM), Europe (EUR), Africa (AFR),
37 East, South and Southeast Asia (EAS, WAS, SEA) and Australasia (AUS). Due to computational
38 requirements, three GCMs were selected to drive the simulations, HADGEM2-ES, MPI-ESM, and NorESM
39 covering, respectively the spread of high, medium and low equilibrium climate sensitivities from the CMIP5
40 ensemble at a global scale (with MIROC5, EC-Earth, GFDL-ES2M as secondary GCMs) focusing on two
41 scenarios RCP2.6 and RCP8.5 (see Box Atlas.1, Figure 1). Two RCMs have contributed so far to this
42 initiative (REMO and RegCM4) constituting an initial homogeneous downscaled ensemble to analyse mean
43 climate change signals and hazards (Teichmann et al., 2020; Coppola et al., 2021), and there are ongoing
44 efforts to extend the CORDEX-CORE ensemble with additional regional simulations (e.g., the COSMO-
45 CLM community) to increase the ensemble size. CORDEX-CORE simulations are distributed as part of the
46 information available for the different CORDEX domains at the Earth System Grid Federation (ESGF).
47
48 CORDEX-CORE spans the spread of the CMIP5 climate change signals for interquartile ranges of annual
49 mean temperature and precipitation for most of the reference regions covered (Teichmann et al., 2020)(see
50 Box Atlas.1, Figure 1). However, it is still a small ensemble and for other variables like extremes or climatic
51 impact-drivers has only been partially investigated in Coppola et al. (2021) and needs further analysis.
52
53
54
55
56
Do Not Cite, Quote or Distribute Atlas-20 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 [START BOX ATLAS.1, FIGURE 1 HERE]
2
3 Box Atlas.1, Figure 1: Temperature (left) and precipitation (right) climate change signals at the end of the
4 century (2070–2099) for the entire CMIP5 ensemble (box-whisker plots) and the
5 CORDEX-CORE driving GCMs (grey symbols) of the respective CORDEX-CORE
6 results (non-grey symbols) in the South Asia (SAS) reference region. The shape of the grey
7 symbols represents the climate sensitivity of the driving GCMs: triangles pointing upwards
8 (low ECS), circles (medium ECS), triangles pointing downwards (high ECS). The
9 corresponding RCM results are drawn using the same symbols, but in orange for REMO and in
10 blue for RegCM. The bottom panels show the warming signal by 2070–2099 over the
11 CORDEX regions for RCP2.6 (left) and RCP8.5 (right) (Teichmann et al., 2020).
12
13 [END BOX ATLAS.1, FIGURE 1 HERE]
14
15 [END BOX ATLAS.1 HERE]
16
17
18 Atlas.1.4.5 Bias adjustment
19
20 Bias adjustment is often applied to data from climate model simulations to improve their applicability for
21 assessing climate impacts and risk (for instance in the Inter-Sectoral Impact Model Intercomparison Project,
22 ISIMIP (Rosenzweig et al., 2017). Bias-adjustment approaches (Section 10.3.1.3) are particularly beneficial
23 when threshold-based indices are used, but they can introduce other biases, in particular when applied
24 directly to coarse-resolution GCMs (Cross-Chapter Box 10.2). Bias-adjustment techniques should be chosen
25 carefully for a particular application. In the Atlas, bias adjustment is not applied systematically (in particular
26 it is not applied for the variables assessed in the Atlas chapter), and only the threshold-dependent extreme
27 indices and climatic impact-drivers (CIDs) included in the Interactive Atlas are bias-adjusted (in particular
28 frost days, and TX35, TX40 in coordination with Chapter 12). To facilitate integration with WGII, the Atlas
29 uses the same bias-adjustment method as in ISIMIP3 (Lange, 2019a) and the same observational reference
30 (W5E5, see Section Atlas.1.4.2), upscaled to the same resolution as the model to avoid downscaling artefacts
31 (Cross-Chapter Box 10.2). The ISIMIP3 bias-adjustment method is a trend-preserving approach that is
32 recommended for general applications, as it reduces biases while preserving the original climate change
33 signal (Casanueva et al., 2020). Following the recommendations given in Chapter 10, results in the
34 Interactive Atlas are displayed for both the adjusted and the raw model output.
35
36
37 [START CROSS-CHAPTER BOX ATLAS.1 HERE]
38
39 Cross-Chapter Box Atlas.1: Displaying robustness and uncertainty in maps
40
41 Coordinators: José Manuel Gutiérrez (Spain), Erich Fischer (Switzerland)
42 Contributors: Alessandro Dosio (Italy), Melissa I. Gomis (France/Switzerland), Richard G. Jones (UK),
43 Maialen Iturbide (Spain), Megan Kirchmeier-Young (Canada/USA), June-Yi Lee (Republic of Korea),
44 Stéphane Sénési (France), Sonia I. Seneviratne (Switzerland), Peter W. Thorne (Ireland/UK), Xuebin Zhang
45 (Canada)
46
47 Spatial information on observed and projected future climate changes has always been a key output of IPCC
48 reports. This information is typically represented in the form of maps of historical trends (from observational
49 datasets) and of projected changes for future reference periods and scenarios relative to baseline periods
50 (from multi-model ensembles). These maps usually include information on the robustness or uncertainty of
51 the results such as the significance of trends or the consistency of the change across models. Visualization of
52 this information combines two aspects that are intertwined: the core methodology (measures and thresholds)
53 and its visual implementation. For observed trends, robustness can be simply ascertained by using an
54 appropriate statistical significance test. However, for multi-model mean changes, the consistency across
55 models for the sign of change (model agreement) and the magnitude of change relative to unforced climate
56 variability (signal-to-noise ratio) provide two complementary measures allowing for simple or more
Do Not Cite, Quote or Distribute Atlas-21 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 comprehensive approaches to represent robustness and uncertainty. While they can be visually represented in
2 various ways with more or less complexity (Retchless and Brewer, 2016), the most common implementation
3 for maps in the climate science community remains the overlay of symbols and/or masking of the primary
4 variable. This Cross-Chapter Box reviews the approaches followed in previous IPCC reports and describes
5 the methods used across this WGI report, presenting the rationale and discussing its relative merits and
6 limitations.
7
8 The objectives in AR6 for representing robustness and uncertainty in maps are 1) adopting a method that can
9 be as coherent as possible across the different global/regional chapters while accommodating different needs,
10 2) being visually consistent across WGs, and 3) making the different layers of information on the maps as
11 accessible as possible for the reader. As a result, a single approach is selected for observations and two
12 alternative approaches (simple and advanced) are adopted for projected future changes. It is important to
13 highlight that, as in previous reports, these approaches are implemented in maps at a grid-box level and,
14 therefore, are not informative for larger spatial scales (e.g., over AR6 reference regions) where the
15 aggregated signals are less affected by small-scale variability leading to an increase in robustness. This is
16 particularly relevant for the AR6 regional assessments and approaches (e.g., for trend detection and
17 attribution, Cross-Chapter Box 1.4, Section 11.2.4) which are performed for climatological regions and not
18 at grid-box scale (Chapters 11, 12, Atlas). Both small and large scales are relevant (e.g., adaptation occurs at
19 the smaller scales but also at the level of countries, which are typically larger than a few grid boxes). They
20 are both addressed in the Interactive Atlas, which implements the above approaches for representing
21 robustness in maps at the grid-box level, but also allows analysing region-wide signals (e.g., AR6 WG I
22 reference regions, monsoon regions, etc.), helping to isolate background changes happening at larger scales
23 (Section Atlas.2.2).
24
25 Approaches used in previous reports
26
27 Recent IPCC reports adopted different approaches for mapping uncertainty/robustness, including their
28 calculation method and/or their visual implementation. In AR5 WGI ‘+’ symbols were used to represent
29 significant trends in observations at grid-box level. For future projections, different methods for mapping
30 robustness were assessed (AR5 Box 12.1, Collins et al., 2013), while proposing as a reference an approach
31 based on relating the multi-model mean climate change signal to internal variability, calculated as the
32 standard deviation of non-overlapping 20-year means in the pre-industrial control runs. Regions where the
33 multi-model mean change exceeded two standard deviations of the internal variability and where at least
34 90% of the models agreed on the sign of change were stippled (as an indication of a robust signal). Regions
35 where the multi-model change was less than one standard deviation were hatched (small multi-model mean
36 signal). However, this category did not distinguish areas with consistent small changes from areas of
37 significant but opposing/divergent signals. In addition, the unstippled/unhatched areas were left undefined,
38 since the categories were not mutually exclusive.
39
40 AR5 WGII (Hewitson et al., 2014) used hatching to represent non-significant trends in observations. For
41 future projections, an elaborated approach with four mutually exclusive and exhaustive categories was
42 proposed (to avoid some of the limitations of the AR5 WGI approach): very strong agreement (same as in
43 WGI); strong agreement; divergent change; and little or no change. These depended on the percentage of
44 models showing change greater than the baseline variability and/or agreeing on sign of change (using a 66%
45 agreement threshold). Leaving the robust regions uncovered minimized any interference with the perception
46 of underlying colours that encoded the primary information of the figure.
47
48 The two special reports IPCC SR1.5 (Hoegh-Guldberg et al., 2018) and SROCC (IPCC, 2019a, 2019c)
49 adopted a simplified approach, using only model agreement (≥66% of models agree on sign of change) to
50 characterize robustness. However, cross-hatching was used in SR1.5 to highlight robust areas where models
51 agree, whereas the SROCC used hatching/shading to represent regions where models disagree. Similarly,
52 stippling was used in SR1.5 to indicate regions with significant trends, whereas it was used in SROCC to
53 represent regions where the trends were not significant.
54 Recent methodologies
55
Do Not Cite, Quote or Distribute Atlas-22 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Since AR5 there has been a growing interest for disentangling small consistent climate change signals from
2 significant divergent opposite changes resulting in conflicting information (Tebaldi et al., 2011), and
3 different statistical tests have been applied to assess the significance of signals working with the individual
4 models forming the ensemble (Dosio and Fischer, 2018; Yang et al., 2018; Morim et al., 2019). Moreover,
5 new approaches have been proposed to identify large changes of opposite sign that compensate in the mean
6 (Zappa et al., 2021). Recent literature has also highlighted the respective risks of Type I vs Type II errors,
7 which can be associated with the determination of robustness in analysed signals (Lloyd and Oreskes, 2018;
8 Knutson et al., 2019). Type I errors are identifying signals when there are none, while Type II errors are
9 concluding there is no signal when there is one. In the case of grid-box level analysis, the focus on small-
10 scale features with inherently large signal-to-noise ratio may emphasize noise even though signals are
11 present when aggregated at larger scale (Sections 11.2.4 and 11.2.5). Consequently, changes averaged over
12 regions or a number of grid-boxes emerge from internal variability at a lower level of warming than at the
13 grid-box level (e.g., Cross-Chapter Box Atlas.1, Figure 2). Hence, focus on grid-box significance enhances
14 the risk of Type II errors for overlooking signals significant at the level of AR6 regions. The significance of
15 signals is also affected by interdependence of single simulations considered in a given ensemble, for example
16 when several come from the same modelling group and share parameterizations or model components
17 (Knutti et al., 2013; Maher et al., 2021). The risk of Type II errors increases when a model ensemble
18 includes several related simulations showing no signal.
19
20 The AR6 WGI approach
21
22 AR6 WGI adapts the approaches applied in previous IPCC reports into a comprehensive framework based
23 on the two general principles followed by AR5 WGII: 1) not obscuring (with stippling or hatching) the areas
24 where relevant/robust information needs to be highlighted (since stippling and hatching obstruct the
25 visualisation of the colours, which can affect the perception/interpretation of the underlying data); 2) using
26 mutually exclusive and exhaustive categories to avoid leaving areas undefined. The three adopted
27 approaches (one for observations and two for model projections) are described in Cross-Chapter Box Atlas.1,
28 Table 1. This framework integrates as much as possible the specificities of each WGI Chapter, proposing in
29 some cases alternative thresholds.
30
31 Approach A is intended for observations and consists of two categories, one for areas with significant trends
32 (colour, no overlay) and one for non-significant ones (coloured areas overlaid with ‘x’), typically using a
33 two-sided test for a significance level of 0.1; Chapter 2 and Atlas trends have been calculated using OLS
34 regression accounting for serial correlation (Santer et al., 2008).
35
36 Approach B is the simple alternative for model projections. It consists of two categories, one for model
37 agreement (at least 80% of the models agree on the sign of change; colour, no overlay) and the other one for
38 non-agreement (hatching). It is noted that model agreement is computed using ‘model democracy’ (i.e.
39 without discarding/weighting models), since quantifying and accounting for model interdependence (shared
40 building blocks) still remain challenging (Section 4.2.6). Different thresholds have been used in previous
41 reports and in the literature. 80% has been widely used in CORDEX studies (Dosio and Fischer, 2018;
42 Kjellström et al., 2018; Nikulin et al., 2018; Yang et al., 2018; Akperov et al., 2019; Rana et al., 2020)
43 partially due to the small ensemble sizes available in some cases; this also helps to reduce the impact of
44 model interdependence in the final results. Although 90% (used in AR5 WGI) provides high confidence on
45 the forced change, it is deemed too stringent for precipitation-like variables and regional assessments and
46 was therefore not included (see Cross-Chapter Box Atlas.1, Figure 1). The 66% threshold, which has been
47 used in previous reports (e.g., SR1.5 and SROCC) and in the literature, is not used to avoid communicating
48 weak confidence. Cross-Chapter Box Atlas.1, Figure 1 illustrates the application of this approach.
49
50 Approach C is a more advanced alternative for model projections, extending the AR5 WGI and simplifying
51 the AR5 WGII approaches (fewer categories). It consists of three categories: ‘robust change’, ‘conflicting
52 change’, and ‘no change or no robust change’ (see the details in Cross-Chapter Box Atlas.1, Table 1). The
53 first two categories can be interpreted as areas where the climate change signal likely emerges from internal
54 variability (i.e., it exceeds the variability threshold in ≥66% of the models). The variability threshold is
55 defined as 𝛾 = √2 ∙ 1.645 ∙ 𝜎20𝑦𝑟 , where 𝜎20𝑦𝑟 is the standard deviation of 20-year means, computed from
Do Not Cite, Quote or Distribute Atlas-23 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 non-overlapping periods in the pre-industrial control (after detrending with a quadratic fit as in AR5 WG1);
2 in cases where this information is not available (e.g., for CORDEX or HighResMIP), the following
3 approximation is used instead: 𝛾 = √2⁄20 ∙ 1.645 ∙ 𝜎1𝑦𝑟 , where 𝜎1𝑦𝑟 is the interannual standard deviation
4 measured in a linearly detrended modern period (note that for white noise 𝜎20𝑦𝑟 = 𝜎1𝑦𝑟 ⁄√20). The factor √2
5 is used as in the AR5 WGI approach to account for the fact that the variability of a difference in means (the
6 climate change signal) is of interest. This approach is an evolution of AR5 WGI method with three notable
7 differences: (a) AR6 uses a lower threshold for internal variability (1.645 corresponding to a 90% confidence
8 level, instead of 2 as used in AR5 WG1); (b) the threshold on agreement in sign is lowered from ≥90% to
9 ≥80%, leading to more grid boxes classified as robust as opposed to conflicting signal; (c) the AR6 method
10 compares signal to variability in each individual model and consequently introduces a 66% cut-off on
11 significant changes implying that the climate change signal likely emerges from internal variability in the
12 baseline period. This change is motivated by the criticisms of internal variability of the models, while it can
13 differ largely across models.
14
15 Cross-Chapter Box Atlas.1, Figure 1 illustrates the application of this method considering the effect of the
16 baseline period (1850–1900 versus 1995–2014) and shows that it provides similar results to related
17 approaches proposed in the literature (Zappa et al., 2021).
18
19 The two alternative approaches discussed above allow visualisation of different level of detail of information
20 on the projected change and are intended for different communication purposes. Approach B just informs on
21 the consistency of the sign of change independent of its significance relative to internal variability, whereas
22 approach C puts the projected changes into context of internal variability and allows highlighting of areas of
23 conflicting signals. It is important to note that different approaches can be applied to the same variable
24 between different chapters for different communication purposes. For example, in maps showing multi-
25 model mean changes of precipitation, Chapter 4 adopts the approach C but Chapter 8 applies the approach B.
26
27 In terms of visual implementation, the approach follows recommendations resulting from conversations with
28 IPCC national delegations: 1) having a consistent approach across WGs would aid consistency and reduce
29 the risk of confusion; 2) defining ‘hatching’ as ‘diagonal lines’ in the caption would aid accessibility for non-
30 expert audiences; 3) a clear and concise legend that explains what these patterns represent should be included
31 directly in the figure; 4) information about model uncertainty should be overlaid such that it does not detract
32 from the data underneath.
33
34 Since stippling is commonly used to represent statistical significance, hatching was chosen to ‘obscure’ the
35 problematic categories in the above approaches; it also facilitates the visualisation of uncertainty in the
36 Interactive Atlas when zooming in. To avoid confusion, methods or thresholds that were unrelated to the
37 three approaches hereby presented were visualised with a different pattern (i.e., model improvement between
38 low- and high-resolution simulations in Chapter 3; agreement between observation-based products in
39 Chapter 5; correlation between two variables in Chapter 6).
40
41
42 [START CROSS-CHAPTER BOX ATLAS.1, TABLE 1 HERE]
43
44 Cross-Chapter Box Atlas.1, Table 1: Approaches for representing robustness (uncertainty) in maps of observed
45 (approach A) and projected (approaches B and C) climate changes.
46
Approach Category Definition Visual Code
A.1. Significant Significant (0.1 level) trend Colour
A. Observations (no overlay)
(significance)
A.2. Non significant Non-significant trend Cross
B. Model B.1. High model ≥80% of models agree on sign of change. Colour
Do Not Cite, Quote or Distribute Atlas-24 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
projections. agreement For Chapter 6 (<5 model ensembles): more (no overlay)
Simple approach than (n-1)/n models agree on the sign of
(agreement) change
B.2. Low model <80% agree on sign of change. Diagonal
agreement For Chapter 6: fewer than (n-1)/n models
agree on the sign of change
C.1. Robust signal ≥66% of models show change greater than Colour
C. Model (significant change and variability threshold 𝛾 and (no overlay)
projections. high agreement) ≥80% of all models agree on sign of change
Advanced
approach C.2. No change or no <66% of models show change greater than Reverse diagonal
(significant robust signal variability threshold
change and
agreement) C.3. Conflicting ≥66% of models show change greater than Crossed lines
signals (significant variability threshold 𝛾 and
change but low <80% of all models agree on sign of change
agreement)
1
2 [END CROSS-CHAPTER BOX ATLAS.1, TABLE 1 HERE]
3
4
5 [START CROSS-CHAPTER BOX ATLAS.1, FIGURE 1 HERE]
6
7 Cross-Chapter Box Atlas.1, Figure 1: Illustration of the simple (top) and advanced (bottom) approaches (B and
8 C in Cross-Chapter Box Atlas.1, Table 1) for uncertainty representation in
9 maps of future projections. Annual multi-model mean projected relative
10 precipitation change (%) from CMIP6 for the period 2040–2060 (left) and
11 2080–2100 (right) relative to the baseline periods 1995–2014 (a–d) and 1850–
12 1900 (e–g) under a high-emission (SSP3-7.0) future. Diagonal and crossed
13 lines follow the indications in Cross-Chapter Box Atlas.1, Table 1. Further
14 details on data sources and processing are available in the chapter data table
15 (Table Atlas.SM.15).
16
17 [END CROSS-CHAPTER BOX ATLAS.1, FIGURE 1 HERE]
18
19
20 Uncertainty at the grid-box and regional scales: interpreting hatched areas
21
22 There is no one-size-fits-all method for representing robustness or uncertainty in future climate projections
23 from a multi-model ensemble. One of the main challenges is the dependence of the significance on the
24 spatial scale of interest: while a significant trend may not be detected at every location, a fraction of
25 locations showing significant trends can be sufficient to indicate a significant change over a region,
26 particularly for extremes (e.g., it is likely that annual maximum 1-day precipitation has intensified over the
27 land regions globally even though there are only about 10% of weather stations showing significant trends;
28 Figure 11.13). The approach adopted in WGI works at a grid-box level and, therefore, is not informative for
29 assessing climate change signals over larger spatial scales. For instance, an assessment of the amount of
30 warming required for a robust climate change signal to emerge can strongly depend on the considered spatial
31 scale. A robust change in the precipitation extremes averaged over a region or a number of grid-boxes
32 emerge at a lower level of warming than at the grid-box level because of larger variability at the smaller
33 scale (Cross-Chapter Box Atlas.1, Figure 2).
34
35 [START CROSS-CHAPTER BOX ATLAS.1, FIGURE 2 HERE]
36
37 Cross-Chapter Box Atlas.1, Figure 2: Climate change signals are more separable from noise at larger spatial
38 scales. The figure is showing the global warming level associated with the
Do Not Cite, Quote or Distribute Atlas-25 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 emergence of a significant increase in the probability due to anthropogenic
2 forcing in the 1-in-20-year daily precipitation event using a 500-year sample
3 from the CanESM2 large ensemble simulations. The left panel uses data
4 analysed over a single grid box, with no spatial aggregation, while the right
5 box uses data averaged over 25 grid boxes to represent moderate spatial
6 aggregation. Aggregation over 25 grid boxes reduces natural variability,
7 resulting in a smaller warming required for a clear separation between the
8 signal and noise. Adapted from Kirchmeier‐Young et al. (2019).
9
10 [END CROSS-CHAPTER BOX ATLAS.1, FIGURE 2 HERE]
11
12 [END CROSS-CHAPTER BOX ATLAS.1 HERE]
13
14
15 Atlas.2 The online ‘Interactive Atlas’
16
17 The WGI Interactive Atlas is a new AR6 product developed as part of the Atlas in consultation with other
18 chapters to facilitate flexible synthesis information for regions, and to support the Technical Summary (TS)
19 and the Summary for Policymakers (SPM), as well as the handshake with WGII. It includes multiple lines of
20 evidence to support the assessment of observed and projected climate change by offering information for
21 regions using both time-slices across scenarios and Global Warming Levels (GWLs). Coordination has been
22 established with other chapters (particularly the regional chapters) adopting their methodological
23 recommendations (Chapter 10) and using common datasets and agreed extreme indices and climatic impact-
24 drivers (CIDs) to support and expand their assessment (Chapters 11 and 12).
25
26 The Interactive Atlas allows for flexible spatial and temporal analysis (Section Atlas.1.3) with a predefined
27 granularity (predefined climatological and typological regions and user-defined seasons) through a wide
28 range of maps, graphs and tables generated in an interactive manner building on a collection of global and
29 regional observational datasets and climate projections (including CMIP5, CMIP6 and CORDEX; Section
30 Atlas.1.4). In particular, the Interactive Atlas provides trends and changes for observations and projections in
31 the form of interactive maps for predefined historical and future periods of analysis, the former including
32 recent-past and paleoclimate (see Cross-Chapter Box 2.1) and the latter including future time-slices (near,
33 medium and long term) across scenarios (RCPs and SSPs; see Cross-Chapter Box 1.4) and GWLs (1.5°C,
34 2°C, 3°C and 4°C; see Cross-Chapter Box 11.1). It also provides regional information (aggregated spatial
35 values) for a number of predefined (reference and typological) regions in the form of time series, annual
36 cycle plots, scatter plots (e.g., temperature versus precipitation), table summaries, and ensemble and seasonal
37 stripe plots. This allows for a comprehensive analysis (and intercomparison, particularly using GWLs as a
38 dimension of integration) of the different datasets at a global and regional scale.
39
40 The Interactive Atlas can be consulted online at http://ipcc-atlas.ifca.es. Figure Atlas.8 illustrates the main
41 functionalities available: the controls at the top of the window allow the interactive selection of the dataset,
42 variable, period (reference and baseline) and season which define a particular product of interest (e.g., annual
43 temperature change from CMIP6 under SSP3-7.0 for the long-term 2081–2100 period relative to 1995–2104
44 in this illustrative case). Regionally aggregated information can be obtained interactively by clicking on one
45 or several subregions on the map and by selecting one of the several options available for visuals (time
46 series, annual-cycle plots, scatter and stripe plots) and tables.
47
48
49 [START FIGURE ATLAS.8 HERE]
50
51 Figure Atlas.8: Screenshots from the Interactive Atlas. (a) The main interface includes a global map and controls to
52 define a particular choice of dataset, variable, period (reference and baseline) and season (in this
53 example, annual temperature change from CMIP6 for SSP3-7.0 for the long-term 2081–2100 period
54 relative to 1995–2104). (b–e) Various visuals and summary tables for the regionally averaged
55 information for the selected reference regions.
56
Do Not Cite, Quote or Distribute Atlas-26 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 [END FIGURE ATLAS.8 HERE]
2
3
4 A major goal during the development of the Interactive Atlas has been ensuring transparency and
5 reproducibility of results, and promoting open science and Findability, Accessibility, Interoperability, and
6 Reuse (FAIR) principles (Wilkinson et al., 2016) described in Atlas.2.3. As a result, full metadata are
7 provided in the Interactive Atlas for each of the products, and the scripts used to generate the intermediated
8 products (e.g., extreme indices and CIDs) and figures are available online in a public repository (Iturbide et
9 al., 2021), which also includes simple notebooks illustrating key parts of the code suitable for reusability.
10 These scripts are based on the climate4R open-source framework (Iturbide et al., 2019) and full metadata
11 have been generated for all final products using the METACLIP framework (Bedia et al., 2019), which
12 builds on standards and describes provenance of the datasets as well as the post-processing workflow.
13
14
15 Atlas.2.1 Why an interactive online Atlas in AR6?
16
17 The idea of an interactive online Atlas was first discussed in the IPCC Expert Meeting on Assessing Climate
18 Information for the Regions (IPCC, 2018a). The meeting stressed the need for the AR6 regional Atlas to go
19 beyond the AR5 experience in supporting and expanding the assessment of key variables/indices and
20 datasets conducted in all chapters, ensuring traceability, and facilitating the ‘handshake’ between WGI and
21 WGII. One of the main limitations of previous products, including the AR5 WGI Atlas (IPCC, 2013a), is
22 their static nature with inherent limited options and flexibility to provide comprehensive regional climate
23 information for different regions and applications. For instance, the use of standard seasons limits the
24 assessment in many cases, such as regions affected by monsoons or seasonal rainband migrations or other
25 phenomena-driven seasons. The limited number of variables which can be treated on a printed Atlas also
26 prevents the inclusion of relevant extreme indices and CIDs. The development of an online Interactive Atlas
27 for AR6 was proposed as a solution to overcome these obstacles, facilitating the flexible exploration of key
28 variables/indices and datasets assessed in all chapters through a wide range of maps, graphs and tables
29 generated in an interactive manner, and thus also providing support to the TS and SPM. One of the main
30 concerns raised by this new online interactive product was the potential danger of having an unmanageable
31 number of final products impossible to assess following the IPCC review process. This was addressed by
32 designing the Interactive Atlas with limited and predefined functionality and granularity thus facilitating the
33 review process and including use of open-source tools and code for traceability and reproducibility of results.
34
35
36 Atlas.2.2 Description of the Interactive Atlas: functionalities and datasets
37
38 The Interactive Atlas builds on the work done in the context of the Spanish National Adaptation Plan
39 (PNACC – AdapteCCa; http://escenarios.adaptecca.es) to develop an interactive online application
40 centralizing and providing key regional climate change information to assist the Spanish climate change
41 impact and adaptation community. The functionalities included in the AR6 WGI Interactive Atlas are an
42 evolution of those implemented in AdapteCCa and have been adapted and extended to cope with the
43 particular requirements of the datasets and functionalities it includes. In particular, the Interactive Atlas
44 allows analysis of global and regional information on past trends and future climate changes through a wide
45 range of maps, graphs and tables generated in an interactive manner and building on six basic products (see
46 Figure Atlas.8):
47 1. Global maps of variables averaged over time-slices across scenarios and GWLs, with robustness
48 represented using the approaches described in Cross-Chapter Box Atlas.1.
49 2. Temporal series, displaying all individual ensemble members and the multi-model mean, with
50 robustness represented as ranges across the ensemble (25th–75th and 10th–90th percentile ranges).
51 The selected reference period of analysis is also displayed as context information, either a time-slice
52 (near-, mid- or long-term) or a GWL (defined for a given model as the first 20-year period where its
53 average surface temperature change first reaches the GWL relative to its 1850–1900 temperature).
54 3. Annual cycle plots representing individual models, the multi-model mean and ranges across the
55 ensemble.
Do Not Cite, Quote or Distribute Atlas-27 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 4. Stripe and seasonal stripe plots, providing visual information on changes across the ensemble
2 (different models in rows with the multi-model mean on the top) and across seasons (months in
3 rows, using the signal from the multi-model mean), respectively.
4 5. Two-variable scatter plots (e.g., temperature versus precipitation).
5 6. Tables with summary information.
6
7 The first of these products provides spatial information about the ensemble mean, while the latter five
8 convey (spatially) aggregated information of the multi-model ensemble for particular region(s) selected by
9 the user from a number of predefined alternatives (see Sections Atlas.1.3.3 and Atlas.1.3.4 for reference and
10 typological regions, respectively).
11
12 The Interactive Atlas includes both atmospheric (daily mean, minimum and maximum temperatures,
13 precipitation, snowfall and wind) and oceanic (sea surface temperature, pH, sea ice, and sea level rise)
14 essential variables assessed in the Atlas chapter and Chapters 4, 8 and 9, as well as some derived extreme
15 indices used in Chapter 11 and a selection of CIDs used in Chapter 12 (see Annex VI):
16 • Maximum of maximum temperatures (TXx) – see Chapter 11.
17 • Minimum of minimum temperatures (TNn) – see Chapter 11.
18 • Maximum 1-day precipitation (RX1day) – see Chapter 11.
19 • Maximum 5-day precipitation (RX5day) – see Chapter 11.
20 • Consecutive Dry Days (CDD) – see Chapter 11.
21 • Standardized Precipitation Index (SPI-6) – see Chapters 11 and 12.
22 • Frost days (FD), both raw and bias adjusted – see Chapters 11 and 12.
23 • Heating Degree Days (HD) – see Chapter 12.
24 • Cooling Degree Days (CD) – see Chapter 12.
25 • Days with maximum temperature above 35°C (TX35), both raw and bias adjusted – see Chapter 12.
26 • Days with maximum temperature above 40°C (TX40), both raw and bias adjusted – see Chapter 12.
27
28 The essential variables are computed for observations and reanalysis datasets as described in Sections
29 Atlas.1.4.1 and Atlas.1.4.2 (note that the Atlas does not include observational datasets for extremes). Trend
30 analyses are available for two alternative baseline periods (1961–2015 and 1980–2015, selected according to
31 data availability). This expands the information available in Chapter 2 for global observational datasets,
32 including new periods of analysis and new regional observational datasets which provide further insight into
33 observational uncertainty.
34
35 Both essential variables and indices/CIDs are computed for CMIP5, CMIP6 and CORDEX model
36 projections (Sections Atlas.1.4.3 and Atlas.1.4.4). The calculations are performed on the original model grids
37 and results are interpolated to the reference regular grids at horizontal resolutions of 2° (CMIP5), 1°
38 (CMIP6) and 0.5° (CORDEX) (Iturbide et al., 2021). Information is available for the historical, SSP1-2.6,
39 SSP2-4.5, SSP3-7.0 and SSP5-8.5 scenarios for CMIP6, and historical, RCP2.6, RCP4.5 and RCP8.5 for
40 CMIP5 and CORDEX, as documented in Annex II Tables 1-2 (for CMIP5/CMIP6) and Annex II Tables 3-
41 14 (for the different CORDEX domains). All products (maps, graphs and tables) are available for different
42 reference periods of analysis, either time-slices (2021–2040, 2041–2060 and 2081–2100 for near-, mid- and
43 long-term future periods, respectively; see Section Atlas.1.3.1), or GWLs (1.5°C, 2°C, 3°C or 4C; see
44 Section Atlas.1.3.2), with changes relative to a number of alternative baselines (including 1850–1900 pre-
45 industrial, and 1995–2014 recent past; see Section Atlas.1.3.1). Note that instead of blending the information
46 from the different scenarios, the Interactive Atlas allows comparison of the GWL spatial patterns and timings
47 across the different scenarios (see Cross Chapter Box 11.1).
48
49 Some of the above indices (in particular TX35 and TX40) are highly sensitive to model biases and the
50 application of bias-adjustment techniques is recommended to alleviate this problem (see Cross-Chapter Box
51 10.2). Bias adjustment is performed as explained in Section Atlas.1.4.5.
52
53 The Interactive Atlas implements the approaches for representing robustness in maps at the grid-box level
54 described in Cross-Chapter Box Atlas.1. These approaches are not necessarily informative for assessing
Do Not Cite, Quote or Distribute Atlas-28 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 trends and climate change signals over larger spatial scales where signals are less affected by small-scale
2 variability leading to an increase in robustness. For regional analysis, the Interactive Atlas allows the
3 analysis of aggregated region-wide signals and assessing their robustness at a regional scale thus
4 complementing the previous approach for grid-box robustness representation. For example, Figure Atlas.9
5 shows large hatched areas for maximum 5-day precipitation in the South Asia region. When aggregated
6 spatially, the region exhibits a robust wetting signal, with most ensemble members agreeing on the sign. This
7 highlights that signals may not have emerged at the station- or grid-box scale but have clearly at aggregated
8 scales, particularly for variables with high variability (e.g., extreme precipitation or cold extremes; see
9 Cross-Chapter Box Atlas.1).
10
11 The advanced approach for representing robustness includes a new category for identifying conflicting
12 signals, where models are projecting significant changes but of opposite signs. This is demonstrated in
13 Figure Atlas.9 which shows a region of central Africa where models have significant changes in surface
14 winds with some projecting increases and others decreases. This is clearly demonstrated in the time-series
15 below the map which shows these wind-speed changes aggregated over the CAF reference region for each of
16 the CMIP6 models and the opposing signals in many of these.
17
18
19 [START FIGURE ATLAS.9 HERE]
20
21 Figure Atlas.9: Analysing robustness and uncertainty in climate change signals across spatial scales using the
22 Interactive Atlas. The left panel shows projected annual relative changes for maximum 5-day
23 precipitation from CMIP6 at 3°C of global warming level relative to the 1850–1900 baseline, through
24 a map of the ensemble mean changes (panel top) and information on the regional aggregated signal
25 over the South Asia reference region as time series (panel bottom). This shows non-robust changes
26 (diagonal lines) at the grid-box level (due to the large local variability), but a robust aggregated signal
27 over the region. The right panel shows projected surface wind speed changes from CMIP6 models for
28 2041–2060 relative to a 1995–2014 baseline under the SSP5-8.5 scenario, again with the ensemble
29 mean changes in the map (panel top) and regionally aggregated time series over Central Africa for
30 each model (panel bottom). This shows conflicting changes (crossed lines) at the grid-box level due to
31 signals of opposite sign in the individual models displayed in the time series.
32
33 [END FIGURE ATLAS.9 HERE]
34
35
36 Atlas.2.3 Accessibility, reproducibility and reusability (FAIR principles)
37
38 The accessibility and reproducibility of scientific results have become a major concern in all scientific
39 disciplines (Baker, 2016). During the design and development of the Interactive Atlas, special attention was
40 paid to these issues in order to ensure the transparency of the products feeding into the Interactive Atlas
41 (which are all publicly available). Accessibility is implemented in collaboration with the IPCC Data
42 Distribution Centre (DDC), since all products underpinning the Interactive Atlas, including the intermediate
43 products required for the indices and CIDs (monthly aggregated data), are curated and distributed by the
44 IPCC-DDC and include full provenance information as part of their metadata. Atlas products are generated
45 using the open source climate4R framework (Iturbide et al., 2019) for data processing (e.g. regridding,
46 aggregation, index calculation, bias adjustment), evaluation and quality control (when applicable). Full
47 metadata are generated for all final products using the METACLIP framework (Bedia et al., 2019), based on
48 the Resource Description Framework (RDF) standard to describe the datasets and data-processing workflow.
49
50 In summary, a number of actions have been conducted in order to implement open access, reproducibility
51 and reusability of results, including:
52 • Use of standards and open-source tools.
53 • Open access to raw data and derived Atlas products via the IPCC-DDC.
54 • Provision of full provenance metadata describing the product generation workflow.
55 • Access to code through an online repository (Iturbide et al., 2021), including the scripts needed for
56 calculating the intermediate datasets and for reproducing some of the figures of the Atlas chapter.
Do Not Cite, Quote or Distribute Atlas-29 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 • Provision of annotated (Jupyter) notebooks describing key elements of the code to provide guidance
2 and facilitate reusability.
3
4 All final products visualized in the Interactive Atlas can be exported in a variety of formats, including PNG
5 and PDF for bitmap and vector information, respectively. Moreover, in the case of the global maps, the final
6 data underlying these products can be downloaded in GIS format (GeoTIFF), thus facilitating reusability of
7 the information. Note that the images are final IPCC products (covered by the IPCC terms of use), whereas
8 the underlying data are distributed by the IPCC-DDC under a more flexible license which facilitates
9 reusability. Moreover, a comprehensive provenance metadata description has been generated, including all
10 details needed for reproducibility, from the data sources to the different post-processes applied to obtain the
11 final product. In these cases, there is also the possibility to download a PNG file augmented with attached
12 metadata information (in JSON format). This metadata information (including the source code generating the
13 product) can be accessed and interpreted automatically using specific JSON software/libraries. However, for
14 the sake of simplicity, a human-readable version of the metadata is accessible directly from the Interactive
15 Atlas describing the key information along the workflow.
16
17 Provenance is defined as a ‘record that describes the people, institutions, entities, and activities involved in
18 producing, influencing, or delivering a piece of data or a thing’. This information can be used to form
19 assessments about their quality, reliability or trustworthiness. In the context of the outcomes of the
20 Interactive Atlas, having an effective way of dealing with data provenance is a necessary condition to ensure
21 not only the reproducibility of results, but also to build trust on the information provided. However, the
22 relative complexity of the data and the post-processing workflows involved may prevent a proper
23 communication of data provenance with full details for reproducibility. Therefore, a special effort was made
24 in order to build a comprehensive provenance metadata model for the Interactive Atlas products.
25
26 Provenance frameworks are typically based on RDF (Resource Description Framework), a family of World
27 Wide Web Consortium (W3C) specifications originally designed as a metadata model (Candan et al., 2001).
28 It is an abstract model that has become a general method for conceptual description of information for the
29 Web, using a variety of syntax notations and serialization formats. METACLIP (Bedia et al., 2019) exploits
30 RDF through specific vocabularies, written in the OWL ontology language, describing different aspects
31 involved in climate product generation, from the data source to the post-processing workflow, extending
32 international standard vocabularies such as PROV-O (Moreau et al., 2015). The METACLIP vocabularies
33 are publicly available in the METACLIP repository (http://github.com/metaclip/vocabularies).
34
35 METACLIP emphasises the delivery of ‘final products’ (understood as any piece of information that is
36 stored in a file, such as a plot or a map) with a full semantic description of its origin and meaning attached.
37 METACLIP ensures ‘machine readability’ through reuse of well-defined, standard metadata vocabularies,
38 providing semantic interoperability and the possibility of developing database engines supporting advanced
39 provenance analytics. Therefore, this framework has been adopted to generate provenance information and
40 attach it as metadata to the products generated by the Interactive Atlas. A specific vocabulary (‘ipcc_terms’)
41 is created alongside the inclusion of new products in the Interactive Atlas and uses the controlled
42 vocabularies existing from CMIP and CORDEX experiments. As an example, Figure Atlas.10 shows the
43 semantic vocabularies needed to encode the information of the typical workflow for computing (from bias-
44 adjusted data) any of the climate indices (extreme or CIDs) included in the Interactive Atlas.
45
46
47 [START FIGURE ATLAS.10 HERE]
48
49 Figure Atlas.10: Schematic representation of the Interactive Atlas workflow, from database description,
50 subsetting and data transformation to final graphical product generation (maps and plots).
51 Product-dependent workflow steps are depicted with dashed borders. METACLIP specifically
52 considers the different intermediate steps consisting of various data transformations, bias adjustment,
53 climate index calculation and graphical product generation, providing a semantic description of each
54 stage and the different elements involved. The different controlled vocabularies describing each stage
55 are indicated by the colours, with gradients indicating several vocabularies involved, usually meaning
56 that specific individual instances are defined in ‘ipcc_terms’ extending generic classes of
Do Not Cite, Quote or Distribute Atlas-30 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 ‘datasource’. These two vocabularies, dealing with the primary data sources have specific annotation
2 properties linking their own features with the CMIP5, CMIP6 and CORDEX Data Reference Syntax,
3 taking as reference their respective controlled vocabularies. All products generated by the Interactive
4 Atlas provide a METACLIP provenance description, including a persistent link to a reproducible
5 source code under version control.
6
7 [END FIGURE ATLAS.10 HERE]
8
9
10 Atlas.2.4 Guidance for users
11
12 Atlas.2.4.1 Purpose of the Interactive Atlas
13
14 The primary purpose of the IPCC is to provide a policy relevant, non-prescriptive assessment of the state of
15 knowledge on climate change and its impacts. This purpose is different from the provision of information
16 targeted to implement climate policies, which is the focus of climate services and national climate change
17 assessment communities. IPCC assessments are based on quantitative observational and model-generated
18 data that are also used in many activities supporting the development of climate policies. However, the
19 functionality of the Interactive Atlas is primarily aimed at supporting the knowledge assessment.
20
21 Much of the assessment in this report is based on multiple lines of evidence (Cross Chapter Box 10.3). The
22 Interactive Atlas facilitates combining multiple observational and model-generated datasets and spatial and
23 temporal analyses that combine to support statements on the characteristics of the climate system. The use of
24 predefined spatial and temporal aggregations imposes constraints on the ability to make specific or tailored
25 assessments but does provide essential background and uncertainty information to generate broad findings
26 and provide confidence statements on these. Also, the inclusion of a selection of extremes and CIDs is a new
27 element in the Interactive Atlas and facilitates broader application including the handshake with WGII.
28 Below, some guidelines on the use, interpretation and limitations of the Interactive Atlas are given.
29
30
31 Atlas.2.4.2 Guidelines for the Interactive Atlas
32
33 Atlas.2.4.2.1 Quantitative support for assessments
34 Many assessment statements make use of evidence derived from observed changes, model projections, and
35 process-oriented attribution of changes to human interventions. The Interactive Atlas shows a small subset of
36 available observations that document climate change, namely surface air temperature and total precipitation
37 (and thus not including observations of other atmospheric and Earth system components used as part of the
38 evidence base for the report). Only datasets that have (near) global or large regional gridded spatial coverage
39 and go back multiple decades are used. For each variable multiple datasets are included, but some of these
40 have overlapping native ground-station observations and so are not independent (see Section Atlas.1.4.1).
41 The datasets show patterns of substantial spatial and temporal variability, and the empirical evidence of a
42 non-stationary climatology needs to be filtered from this information. Issues with quality, representativity,
43 and mutual consistency lead to constraints on their use for attribution of causes of trends (see Section
44 10.4.1.2 for examples). The practice of attributing trends and extreme events to human causes gives
45 confidence that these trends are expected to continue in the (near) future, provided the human drivers of
46 climate change remain unchanged. However, large internal variability at decadal time scales can be mis-
47 interpreted as an anthropogenic influence on the likelihood of extreme events, and in that case extrapolation
48 of trends cannot be expected to be a reliable predictor for the future (Schiermeier, 2018).
49
50 The Interactive Atlas gives access to a specific set of climate variables from a large number of climate model
51 simulations, particularly the (global) CMIP5, CMIP6 and (regional) CORDEX archives. The global model
52 outputs generally give a relatively coarse picture of climate change, which is an important line of evidence
53 for the detection and attribution of climate change, but is rarely directly applicable for local climate change
54 assessment or support of policy design (van den Hurk et al., 2018). To provide additional detail, downscaling
55 global projections with regional climate models (RCMs) or statistical downscaling can be undertaken but
Do Not Cite, Quote or Distribute Atlas-31 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 also adds a source of uncertainty as it involves additional modelling (see Section 10.3).
2
3 The information displayed in the Interactive Atlas allows a number of sources of uncertainty to be
4 quantified. ‘Observational uncertainty’ is represented by the use of multiple (albeit often not completely
5 independent) observational datasets. ‘Uncertainty due to internal variability’ cannot be quantified directly
6 since multiple realizations from historic and future projections are not accessible (the Interactive Atlas uses a
7 single realization of each model). The use of a large collection of model systems allows for an elaborate
8 quantification of ‘model uncertainty’. In addition, a comparison of CMIP5 and CMIP6 supports evidence of
9 progress in model quality since the AR5, while the evaluation of the added value of RCMs reveals model
10 uncertainty related to spatial resolution (see Section 10.3). Finally, the assessment of ‘scenario uncertainty’
11 is supported by the inclusion of multiple emission scenarios for both CMIP5, CORDEX and CMIP6.
12
13 The communication of uncertainty has a profound influence on the perception of information that is
14 exchanged during the communication process. An assessment of uncertainty communication and the barriers
15 to climate information construction is given in Section 10.5.4.
16
17
18 Atlas.2.4.2.2 Insights from physical understanding
19 The detailed technical findings in IPCC reports also serve as an important benchmark resource for the
20 research community. The Interactive Atlas complements the IPCC assessment report as a repository of
21 scientific information on global and regional climate and its representation in coordinated model ensemble
22 experiments. Regional climate is governed by a mixture of drivers, such as circulation patterns, seasonal
23 monsoons, annual cycles of snow and regional land-atmosphere feedbacks. Global warming may affect
24 regional climate characteristics by altering the dynamics of their drivers. The Interactive Atlas allows the
25 comparison of different levels of global warming on specific regional climate features but is not designed for
26 advanced analysis of the relationship between drivers and regional climate characteristics. For this, tailored
27 analysis protocols need to be applied, such as the aggregation of climate change information from ensembles
28 of regional climate projections, and stratification according to drivers of regional climate such as patterns of
29 atmospheric circulation (Lenderink et al., 2014). The analysis of complex regional climate characteristics
30 resulting from compound drivers also require additional expert knowledge and data processing (Thompson et
31 al., 2016). Section 12.6.2 assesses various categories of climate services, including tailored analysis of
32 regional climate processes.
33
34
35 Atlas.2.4.2.3 Construction of storylines
36 Communicating the full extent of available information on future climate for a region, including a
37 quantification of uncertainties, can act as a barrier to the uptake and use of such information (Lemos et al.,
38 2012; Daron et al., 2018). To address the need to simplify and increase the relevance of information for
39 specific contexts, recent studies have adopted narrative and storyline approaches (Hazeleger et al., 2015;
40 Shepherd et al., 2018) (see Sections 1.4.4 and 10.5.3 for definitions and further discussion on these
41 concepts). The use of region-specific climate storylines, including a role for local mechanisms, drivers and
42 societal impacts generally requires detailed information that is typically not provided by the Interactive
43 Atlas. However, background information and basic (scenario) assumptions can be derived from the
44 Interactive Atlas which can be considered to provide an expert knowledge base from which to build targeted
45 storylines and climate information.
46
47
48 Atlas.2.4.2.4 Visual information
49 The visual communication of climate information can take many forms. Besides the standard visual products
50 typically used for communicating global and regional climate information to practitioners (e.g., maps, time
51 series or scatter plots), the Interactive Atlas incorporates new visuals, for example ‘stripes’ (RMetS, 2019),
52 facilitating the communication of key messages (e.g., warming and consistency across models) to a less
53 technical audience. The various tabular and graphical representation alternatives included as options in the
54 Interactive Atlas (Figure Atlas.8) facilitate exploring the information interactively from different
55 perspectives and for different levels of detail, thus favouring the communication with the larger and diverse
Do Not Cite, Quote or Distribute Atlas-32 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 audience of IPCC products.
2
3 To support the use of visuals provided in the Interactive Atlas for application to different audiences, new
4 insights since the AR5 have emerged from a range of scientific disciplines, including the cognitive and
5 psychological sciences (Harold et al., 2016). Studies have used interviews and online surveys to assess
6 interpretations of visuals used to communicate climate information and uncertainties (Daron et al., 2015;
7 Lorenz et al., 2015; McMahon et al., 2015; Retchless and Brewer, 2016). They commonly find wide-ranging
8 interpretations and varied understandings of climate information amongst respondents due to the choice of
9 visuals. In addition, Taylor et al. (2015) found that preferences for a particular visualization approach do not
10 always align with approaches that achieve greatest accuracy in interpretation. Choosing appropriate visuals
11 for a particular purpose and audience can be informed by testing and evaluation with target groups.
12
13
14 Atlas.2.4.2.5 Dedicated climate change assessment programs
15 Communication aimed at informing the general public about assessed scientific findings on climate change
16 have a different purpose and format than if intended to inform a specific target audience to support
17 adaptation or mitigation policies (Whetton et al., 2016). The growing societal engagement with climate
18 change means IPCC reports are increasingly used directly by businesses, the financial sector, health
19 practitioners, civil society, the media, and educators at all levels. The IPCC reports could effectively be
20 considered a tiered set of products with information relevant to a range of audiences.
21
22 The Interactive Atlas does provide access to a collection of observational and modelling datasets, presented
23 in a form that supports the distillation of information on observed and projected climate trends at the regional
24 scale. Access to the repository of underlying datasets enables further processing for particular purposes. As
25 noted above, it is not the intention nor the ambition of this IPCC assessment and the Interactive Atlas
26 component to provide a climate service for supporting targeted policies. For this an increasing number of
27 dedicated climate change assessment programs have been carried out, aiming at mapping climate change
28 information relevant for adaptation and mitigation decision support.
29
30 For instance, the European Environment Agency (2018) provides an overview of European national climate
31 change scenario programs. Most of these use CMIP5 (or earlier) global climate change ensembles driven by
32 an agreed set of greenhouse-gas emission scenarios, followed by downscaling using RCMs and/or statistical
33 methods, in order to generate regionally representative hydrometeorological indicators of climate change. In
34 some cases, output of selected downscaled global and regional models is provided to users (Whetton et al.,
35 2012; Daron et al., 2018). Uptake by users is strongly dependent on providing justification of the selection or
36 for the downscaling procedure and if further steps are needed to tailor the information to local scales (Lemos
37 et al., 2012). More comprehensive programs provide probabilistic climate information by careful analysis
38 and interpretation of ensembles of model outputs (Lowe et al., 2018). The information is generally tailored to
39 professional practitioners with expertise to interpret and process this probabilistic data. This top-down
40 probabilistic information chain is not always able to highlight the essential climate change information for
41 users, and alternative bottom-up approaches are encouraged (Frigg et al., 2013). Section 12.6.2 assesses
42 climate services including the national climate assessments and user uptake.
43
44
45 Atlas.3 Global synthesis
46
47 Most other chapters in WGI assess past or future behaviour of specific aspects of the global climate system
48 and this section introduces some of the key results, specifically from Chapters 2, 4 and 9. This provides a
49 global overview on observations and information from the CMIP5 and CMIP6 ensembles to underpin the
50 regional assessments in the rest of the Atlas Chapter and the results displayed in the Interactive Atlas. Thus,
51 its aim is not to generate an assessment of regional climate change directly but to provide the global context
52 for this information derived later in the Atlas. Section Atlas.3.1 considers global atmospheric and land-
53 surface information with global ocean information in Section Atlas.3.2.
54
55
Do Not Cite, Quote or Distribute Atlas-33 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.3.1 Global atmosphere and land surface
2
3 The principal atmospheric quantities of interest for understanding how climate change may impact human
4 and ecological systems, as well as being key global indicators of change, are surface air temperature and
5 precipitation. They are therefore a significant focus of the regional climate assessments in the following
6 regional sections of the chapter (Atlas.4 to Atlas.11) and of the Interactive Atlas. Changes in these variables
7 over land during the recent past (1960–2015) are shown in Figure Atlas.11 using results from two global
8 datasets (assessed in Chapter 2) to illustrate both where there is robust information on observed trends and
9 observational uncertainty.
10
11 For temperature, a clear signal of warming is seen over most land areas with an amplification at high
12 latitudes, though all continents apart from Europe also have regions where trends are not significant.
13 Significant changes in annual mean precipitation are seen over much more limited areas though with
14 consistent increasing trends over some northern high-latitude regions and decreasing trends over smaller
15 regions in tropical Africa, the Americas and Southwest Asia. The information conveyed in Figure Atlas.11
16 on both consensus in the signal of change and on observational uncertainty is used in this chapter as a line of
17 evidence to assess historical observed trends.
18
19 As an alternative way of viewing and summarising information in the observational data, the panels (c) and
20 (d) in Figure Atlas.11 show the time at which any significant temperature trends from the Berkeley Earth and
21 CRUTEM5 datasets, averaged over the reference regions, emerged from interannual variability – with a
22 signal to noise ratio greater than two (Hawkins et al., 2020). In the former a regionally averaged warming
23 signal has emerged over all of the land reference regions. In the latter, emergence times are only calculated
24 for those regions which have data available in more than 50% of the land area (unlike Berkeley Earth,
25 CRUTEM does not include spatial interpolation, see Section 2.3.1.1.3) and these are similar for all but one of
26 the regions indicating that observational uncertainty does not change the main conclusion of widespread
27 emergence of surface temperature signals over land regions.
28
29
30 [START FIGURE ATLAS.11 HERE]
31
32 Figure Atlas.11: Observed linear trends of signals in annual mean surface air temperature (a, b) and
33 precipitation (e, f) in the Berkeley Earth, CRU TS and GPCC datasets (see Section Atlas.1 for
34 dataset details). Trends are calculated for the common 1960–2015 period and are expressed as ºC per
35 decade for temperature and relative change (with respect to the climatological mean) per decade for
36 precipitation. Crosses indicate regions where trends are not significant (at a 0.1 significance level) and
37 the black lines mark out the reference regions defined in Section Atlas.1. Panels c and d display the
38 period in which the signals of temperature change in data aggregated over the reference regions
39 emerged from the noise of annual variability in the respective aggregated data. Emergence time is
40 calculated for (c) Berkeley Earth (as used in (a)) and CRUTEM5. Regions in the CRUTEM5 map are
41 shaded grey when data are available over less than 50% of the land area of the region. Further details
42 on data sources and processing are available in the chapter data table (Table Atlas.SM.15).
43
44 [END FIGURE ATLAS.11 HERE]
45
46 As described earlier, information on projected future changes is required both at different time periods in the
47 future under a range of emissions scenarios but also for different global warming levels. Figure Atlas.12
48 shows the global surface air temperature change projection calculated from the CMIP6 ensemble mean of for
49 the middle of the century under the SSP1-2.6 and SSP3-7.0 emissions scenarios compared to the end of the
50 century warming under SSP3-7.0 and for a global warming level of 2°C. The patterns of changes are similar
51 to the observed warming and there is a high level of consistency with CMIP5 in terms of both patterns and
52 magnitude of change (Interactive Atlas). However, for the long-term future, warming in the CMIP6
53 ensemble is generally higher, reflecting the increase in the top end of the range of climate sensitivities
54 amongst the CMIP6 GCMs (Figure Atlas.13).
55
56
Do Not Cite, Quote or Distribute Atlas-34 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 [START FIGURE ATLAS.12 HERE]
2
3 Figure Atlas.12: Global temperature changes projected for the mid-century under SSP1-2.6 (left, top) and SSP3-
4 7.0 (left, bottom) compared with a 2°C global warming level (right, top) and the end of the
5 century under SSP3-7.0 (right, bottom) from the CMIP6 ensemble. Note that the future period
6 warmings are calculated against a baseline period of 1995–2014 whereas the global mean warming
7 level is defined with respect to the baseline period of 1851–1900 used to define global warming
8 levels. The other three SSP-based maps would show greater warmings with respect to this earlier
9 baseline. Further details on data sources and processing are available in the chapter data table (Table
10 Atlas.SM.15).
11
12 [END FIGURE ATLAS.12 HERE]
13
14
15 Figure Atlas.12 demonstrates how temperature is projected to increase for all regions, and at a greater rate
16 than the global average over many land regions and with significant amplification in the Arctic. It also shows
17 the higher mid-century warming and significantly higher end of century warming under the high-emission
18 SSP3-7.0 scenario compared to the low-emission SSP1-2.6 scenario. Conversely, comparing the projected
19 2°C Global Warming Level change with that projected additional warming compared to the recent past under
20 the SSP1-2.6 scenario demonstrates the much smaller additional warming projected under this low-emission
21 scenario. Finally, the maps display the CMIP6 ensemble mean projection but it is important to explore the
22 full range of outcomes from the ensemble, for example when undertaking a comprehensive risk assessment
23 in which temperature is an important hazard. This can be explored regionally in the Interactive Atlas
24 (Section Atlas.2) by viewing the timeseries of changes for all of the models within the ensemble over the
25 AR6 WGI reference regions (Figure Atlas.2:).
26
27 Changes in annual mean precipitation present a more complex picture with regions of decrease as well as
28 increase and areas where there is model disagreement on the sign of the change, even when the signal is
29 strong in the long-term future period as shown in Cross-Chapter Box Atlas.1, Figure 1. However, as with the
30 temperature changes, there is a high level of consistency in the patterns and magnitude of the precipitation
31 changes, with changes in some areas being larger in the long-term future period. Considering changes over
32 land, Cross-Chapter Box Atlas.1, Figure 1 also shows that at lower warming levels there are many regions,
33 especially in the southern hemisphere, where there is no robust signal of change from the models.
34
35 In addition to displaying results from global model ensembles as maps of projected changes and their
36 robustness or as timeseries of the projected temporal evolution of the median and range of a climate statistic,
37 it is often useful to generate area-averaged summaries of these statistics under different future emission
38 scenarios or at specific global warming levels. This is demonstrated in Figure Atlas.13 and forms the basis of
39 a common set of analyses which are presented for the reference regions in the regional assessments in the
40 following Sections Atlas.4 to Atlas.11. It shows the range of projected changes compared to the 1850–1900
41 and the recent past 1995–2014 baseline periods for the CMIP5 and CMIP6 ensembles. The first four panels
42 show: annual mean changes in temperature globally and over land only for various global warming levels
43 and emission scenarios and time periods (left pair) and then again globally and for global land, changes in
44 precipitation and temperature at the same global warming levels (right pair). The second four panels provide
45 the same temperature and precipitation information globally and for global land only in the December–
46 February and July–August seasons. These results demonstrate the consensus between the two ensembles for
47 increased warming over land areas and increases in global precipitation at all warming levels and that global
48 land precipitation increases more. They also show the increased precipitation response in DJF reflecting the
49 large precipitation increases in the northern hemisphere higher latitudes in winter. Finally, they demonstrate
50 the greater warming projected by the CMIP6 ensemble, as an average over the ensemble and the upper end
51 of the range. See Chapter 4 for an in-depth assessment of these results.
52
53
54
55
56
Do Not Cite, Quote or Distribute Atlas-35 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 [START FIGURE ATLAS.13 HERE]
2
3 Figure Atlas.13: Changes in annual mean surface air temperature and precipitation averaged over the global
4 (left) and global land areas (right) in each horizontal pair of panels. The top left two panels show
5 the median (dots) and 10th–90th percentile range across each model ensemble for temperature change,
6 for two datasets (CMIP5 and CMIP6) and two scenarios (SSP1-2.6/RCP2.6 and SSP5-8.5/RCP8.5).
7 The first twelve bars represent the projected changes over three time periods (near-term 2021–2040,
8 mid-term 2041–2060 and long-term 2081–2100) compared to the baseline period of 1995–2014 and
9 the remaining four bars represent the additional warming projected relative to the same baseline to
10 reach four global warming levels (GWL; 1.5°C, 2°C, 3°C and 4°C). The top right two panels show
11 scatter diagrams of temperature against precipitation changes, displaying the median (dots) and 10 th–
12 90th percentile ranges for the same four GWLs, again representing the additional changes for the
13 global temperature to reach the respective GWL from the baseline period of 1995–2014. In all panels
14 the dark (light) grey lines or dots represent the CMIP6 (CMIP5) simulated changes in temperature and
15 precipitation between the 1850–1900 baseline used for calculating GWLs and the recent past baseline
16 of 1995–2014 used to calculate the changes in the bar diagrams and scatter-plots. Changes are
17 absolute for temperature and relative for precipitation. The script used to generate this figure is
18 available online (Iturbide et al., 2021) and similar results can be generated in the Interactive Atlas for
19 flexibly defined seasonal periods. Further details on data sources and processing are available in the
20 chapter data table (Table Atlas.SM.15).
21
22 [END FIGURE ATLAS.13 HERE]
23
24
25 Global warming leads to systematic changes in regional climate variability via various mechanisms such as
26 thermodynamic responses via altered lapse rates (Kröner et al., 2017; Brogli et al., 2019) and land-
27 atmosphere feedbacks (Boé and Terray, 2014). These can modify temporal and spatial variability of
28 temperature and precipitation, including an altered seasonal and diurnal cycle and return frequency of
29 extremes. Regional influences from and feedbacks with sea surface, clouds, radiation and other processes
30 also modulate the regional response to enhanced warming, both locally and, via teleconnections, remotely.
31
32 Given their potential to influence extremes in temperature, precipitation and other climatic impact-drivers
33 and hazards, and thus risks to human and ecological systems, it is important to understand these links for
34 developing adaptations in response to clear anthropogenic influences on individual hazards. This will also
35 support the related fields of disaster risk reduction and global sustainable development efforts (Steptoe et al.,
36 2018). They demonstrated that 15 regional hazards shared connections via the El Niño–Southern Oscillation,
37 with the Indian Ocean Dipole, North Atlantic Oscillation and the Southern Annular Mode (see Annex IV)
38 being secondary sources of significant regional interconnectivity (Figure Atlas.14). Understanding these
39 connections and quantifying the concurrence of resulting hazards can support adaptation planning as well as
40 multi-hazard resilience and disaster risk reduction goals.
41
42
43 [START FIGURE ATLAS.14 HERE]
44
45 Figure Atlas.14: Influence of major modes of variability (see Annex IV) on regional extreme events relevant to
46 assessing multi-hazard resilience. Ribbon colours define the driver from which they originate and
47 their width is proportional to the correlation. Crossed lines represent where there is conflicting
48 evidence for a correlation or where the driver is not directly related to the hazard and dots represent
49 drivers that have both a positive and negative correlation with the hazard. Figure is copied from
50 Steptoe et al. (2018) / CCBY4.0.
51
52 [END FIGURE ATLAS.14 HERE]
53
54
55 The main modes of variability influencing global and regional climate are comprehensively described in
56 Annex IV. In the context of the assessment in the Atlas chapter, they are important because of their influence
57 on the variability of temperature (Part A) and precipitation (Part B) in regions around the world. This is
Do Not Cite, Quote or Distribute Atlas-36 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 quantified in Table Atlas.1: which lists the fraction of interannual variance in seasonal mean temperature and
2 precipitation explained by variability in these modes. The table provides information on the influence of the
3 teleconnections for selected seasons for the interannual to decadal modes and at an annual scale for the
4 multi-decadal. The columns related to the interannual to decadal modes focus on the seasons where these
5 connections are strongest but each mode of variability will often have influences in other seasons (for more
6 details see Annex IV). The table shows that for many regions, seasonal temperature and precipitation is
7 substantially modulated by these modes of variability, all regions feel some influence and variability in
8 ocean basins often has influence in multiple remote regions.
9
10
11 [START TABLE ATLAS.1 HERE]
12
13 Table Atlas.1: Regional mapping of the teleconnections associated with the main modes of variability (Annex IV).
14 Fraction of surface air temperature and precipitation variance explained at interannual timescale by
15 each mode of variability (columns) for each AR6 region (rows) based on the coefficient of
16 determination R2. Units are in percent and non-significant values based on t-statistics at the 95% level
17 confidence are indicated by a cross. HadCRUT (HAD), GISTEMP (GIS), BerkeleyEarth (BE), and
18 CRU-TS (CRU) observed datasets are used to assess the strength of the teleconnection for surface air
19 temperature and GPCC and CRU-TS are used for precipitation. The colour scale given on label bars
20 shown at the bottom quantifies the values of the explained variance and also stands for the sign of the
21 teleconnection for the positive phase of the mode. All data are linearly detrended prior the
22 computation of the regression. Note that results are sensitive to the choice of the detrending function
23 (linear, loess filter, 3-order polynomial function) but by few percent at most, which is well below the
24 range of the observational uncertainty assessed here through the use of several observational products.
25
2-meter temperature
NAM SAM ENSO IOB IOD AZM AMM PDV AMV
DJF DJF DJF MAM SON JJA JJA Annual Annual
1959-2019 1979-2019 1959-2019 1958-2019 1958-2019 1958-2019 1958-2019 1900-2014 1900-2014
HAD
HAD
HAD
HAD
HAD
HAD
HAD
HAD
HAD
CRU
CRU
CRU
CRU
CRU
CRU
CRU
CRU
CRU
GIS
GIS
GIS
GIS
GIS
GIS
GIS
GIS
GIS
BE
BE
BE
BE
BE
BE
BE
BE
BE
Africa
Sahara 60 56 60 57 9 15 14 16 8 10 12 9 7 13 5 15 14
Western Africa 24 22 26 28 11 43 50 45 40 18 29 20 15 8 13 8 8 5 4 9
Central Africa 16 22 17 19 14 13 50 41 56 58 13 16 9 12 11 15 13 13 13
North-eastern Africa 19 20 16 21 40 34 41 28 8 5 5 9
South-eastern Africa 9 17 11 12 17 37 36 42 29 7 8 17 6 8 12 5
West-southern Africa 43 45 56 53 17 27 32 32 7 8 4 4 5 5
East-southern Africa 12 13 72 71 75 73 32 40 36 32 4 3
Madagascar 12 35 22 25 16 33 22 26 9 14 7 13 10 10 8 4 5
Asia
West Siberia 45 47 45 44
East Siberia 52 54 53 50 7 3 3 4
Russian Far East 7 8 8 14 12 9 6 6 6 7 4 5
West Central Asia
East Central Asia 7
Tibetan Plateau 19 11 15 14 8 5 4 5 6 8 5 17
East Asia 8 8 6 13 12 14 14
South Asia 9 9 8 10 7 11 14 12 11 9 7 8 4 6 5
Southeast Asia 34 46 36 41 71 76 73 73 7 4 6 5
Arabian Peninsula 30 33 29 35 12 11 7 21 10 11 11
Do Not Cite, Quote or Distribute Atlas-37 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
Australasia
Northern Australia 12 31 19 20 34 46 37 33 21 4 5 5 12
Central Australia 14 12 14 14 19 18 22 24 19 18 18 18 19 21 29 19 4 4 6 9
Eastern Australia 21 22 24 21 20 19 20 21 18 20 18 17 11 10 7 6 5 6 11
Southern Australia 21 20 24 26
New Zealand 15 17
Central and South America
South Central
22 24 16 31 33 36 11 9 17 14 15 21 8 4 7 4
America
North-western South
11 14 13 17 79 86 82 80 59 48 52 56 12 24 15 22 7 15 10 14 11 5 6 9 9
America
Northern South
6 8 50 61 65 46 50 65 64 65 13 23 21 11 8 9 9 5
America
Northeastern South
21 29 28 22 60 54 52 64 11 7 8 9 8 9
America
South American
47 56 59 52 22 27 39 35 15 26 24 23 7 9 6 6 6
Monsoon
South-western South
14 19 19 10 13 22 20 11 8 11 12 7 8 11 7 4
America
South-eastern South
19 22 23 20 5
America
Southern South
8 18 12 15
America
Europe
Mediterranean 25 25 32 28 7 7 23 16 19 18
Western & Central
28 30 27 27 12 13 13 13
Europe
Eastern Europe 33 36 34 35 7 7 7 8
Northern Europe 49 55 53 54 6 8 5 5
North America
North Central
10 10 18 10 15 9 22 15 19 17 7 7 7 7 15 17 17 11 16 11 23 24
America
Western North
4 5 4 5 6 5 6
America
Central North
17 18 17 17 7 7 8 8 9 9 7 11
America
Eastern North
13 11 11 11 12 10 11 10 4 4 4 8 10 9 10
America
North-eastern North
12 27 20 12 6 8 7 5 7 9 14
America
North-western North
10 10 9 11 16 18 17 18 6 8 7 9
America
Small Islands
Caribbean 15 22 15 8 37 19 23 15 20 17 4 6 12
Pacific
Polar Terrestrial Regions
Greenland/Iceland 47 42 38 35 7 42 43 38 51
Russian Arctic 26 17 27 31 9 14 10 10
West Antarctica 12 14 8 10 8 7 24 17 6
East Antarctica 52 25 37
NAM SAM ENSO IOB IOD AZM AMM PDV AMV
Colder Warmer
50 40 30 20 20 30 40 50
Temperature anomalies and percentage of explained variance
Do Not Cite, Quote or Distribute Atlas-38 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
Precipitation
NAM SAM ENSO IOB IOD AZM AMM PDV AMV
DJF DJF DJF MAM SON JJA JJA Annual Annual
1959-2019 1979-2019 1959-2019 1958-2019 1958-2019 1958-2019 1958-2019 1900-2014 1900-2014
Africa GPCC CRU GPCC CRU GPCC CRU GPCC CRU GPCC CRU GPCC CRU GPCC CRU GPCC CRU GPCC CRU
Sahara ` 20 17 14 10 7 10 25 24
Western Africa 17 13 4 7 19 27
Central Africa 8 10 9 11 13 9
North-eastern Africa 7 16 11 32 31
South-eastern Africa 24 20 59 55 4
West-southern Africa 30 22 17 14 11 13
East-southern Africa ``` 36 31 7 7 6 5
Madagascar 7 12 8
Asia
West Siberia ` ` 7 ` ` ` 8 9 ` ` 11 `
East Siberia 11
Russian Far East 9 10 5
West Central Asia 13 17 27 14 4
East Central Asia 39 36
Tibetan Plateau 16 13 7 9 13 4 6
East Asia ``` ``` 19 21 26 20 ``` 8 9 ``` 9 8 ```
South Asia 8
Southeast Asia 31 31 6 51 45 9 14 8 6
Arabian Peninsula 24 20 5 7
Australasia
Northern Australia 14 12 19 18 7 7
Central Australia 13 11 19 21 7 7 5 4
Eastern Australia 14 10 8 7 8 7
Southern Australia 10 11 41 38 8 3
New Zealand
Central and South America
South Central America 16 15 7 7
North-western South
7 16 11 23 ` `` 16 8
America
Northern South America 64 51 22 22 31 16 11 12
North-eastern South America 20 17 12 11 7 8
South American Monsoon 7 ` `` 6
South-western South
10 16 12 19 12
America
South-eastern South America 22 19 13 13 10 13 10 6 4 6 5
Southern South America 13 33 7 9
Europe
Mediterranean 58 58
Western & Central Europe 15 20 10 9 4 8
Eastern Europe
Northern Europe 35 29
North America
Do Not Cite, Quote or Distribute Atlas-39 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
North Central America 25 26 25 29 12 12 11 12 5 6
Western North America 4 5
Central North America 14 10 17 16 3 6 6
Eastern North America 8 10 4
North-eastern North America 24 27 4 16 4
North-western North
15 13 8 4
America
Small Islands
Caribbean 10 18 8 8 12 10 13 5 5
Pacific
Polar Terrestrial Regions
Greenland/Iceland 7 9 7
Russian Arctic 10 6 8
West Antarctica
East Antarctica
NAM SAM ENSO IOB IOD AZM AMM PDV AMV
Drier Wetter
50 40 30 20 20 30 40 50
Precipitation anomalies and percentage of explained variance
1
2 [END TABLE ATLAS.1 HERE]
3
4
5 Atlas.3.2 Global ocean
6
7 As with the atmosphere, there are several key ocean-related quantities which are relevant for understanding
8 how climate change may impact human and ecological systems and/or key global indicators of change.
9 These include ocean surface temperature and heat content, sea-surface height, sea ice cover and thickness,
10 and certain chemical properties such as ocean acidity and oxygen concentration. For example, there is
11 growing awareness of the threat presented by ocean acidification to ecosystem services and its socio-
12 economic consequences are becoming increasingly apparent and quantifiable (Hurd et al., 2018) and SR1.5
13 (IPCC, 2018c) noted a significant impact of low levels of global warming on the state of the global oceanic
14 ecosystems and food security. For instance, 70–90% of coral reefs are projected to decline at a warming
15 level of 1.5°C, with larger losses at 2°C.
16
17 Thus, because of their importance to coastal populations and infrastructure and ocean ecosystems, the
18 Interactive Atlas focuses on change in sea-surface temperature, sea level and pH. Figure Atlas.15:shows
19 projected changes sea-surface temperature and sea level at the end of the century under SSP1-2.6 and SSP5-
20 8.5 emissions, demonstrating the much larger changes seen with the higher-emission scenario. The projected
21 changes in sea level show the significantly greater increases, of up to 1 m locally, under a high-emissions
22 future. Regional details of these projected changes under a range of emission scenarios and time periods can
23 be explored in the Interactive Atlas. An in-depth assessment of these changes is presented in Section 5.3 and
24 Chapter 9.
25
26
27 [START FIGURE ATLAS.15 HERE]
28
29 Figure Atlas.15: Projected changes in sea surface temperature (top), sea level rise (bottom) for 2081–2100 under
30 the SSP1-2.6 (left) and SSP5-8.5 (right) emission scenarios compared to a 1995–2014 baseline
31 period from the CMIP6 ensemble. For sea surface temperature, diagonal lines indicate regions
32 where 80% of the models do not agree on the sign of the projected changes. Further details on data
Do Not Cite, Quote or Distribute Atlas-40 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 sources and processing are available in the chapter data table (Table Atlas.SM.15).
2
3 [END FIGURE ATLAS.15 HERE]
4
5
6 Atlas.4 Africa
7
8 The assessment in this section focuses on changes in average temperature and precipitation (rainfall and
9 snow), including the most recent years of observations, updates to observed datasets, the consideration of
10 recent studies using CMIP5 and those using CMIP6 and CORDEX simulations. Assessment of changes in
11 extremes are in Chapter 11 (Table 11.4–11.6) and climatic impact-drivers in Chapter 12 (Table 12.1-12.12).
12
13
14 Atlas.4.1 Key features of the regional climate and findings from previous IPCC assessments
15
16 Atlas.4.1.1 Key features of the regional climate
17
18 Africa has many varied climates which can be categorized as dry regime in the Saharan region, tropical
19 humid regime in West and East Africa except for parts of the Greater Horn of Africa (alpine) and the Sahel
20 (semi‐arid), and a dry/wet season regime in the northern and southern African region including the Namibian
21 desert; each climate region has its local variations resulting in very high spatial and temporal variations (Peel
22 et al., 2007). Based on the varied climates, nine subregions are defined for Africa (see Figure Atlas.16:) the
23 Mediterranean region (MED) including North Africa, Sahara including parts of the Sahel (SAH), West
24 Africa (WAF), Central Africa (CAF), North Eastern Africa (NEAF), South Eastern Africa (SEAF), West
25 Southern Africa (WSAF), East Southern Africa (ESAF) and Madagascar (MDG).
26
27 The climatic features that characterize the intra-seasonal and inter-annual variability of Africa are mainly the
28 Madden-Julian Oscillation (MJO) which is confined to the deep tropics during boreal winter, Pacific Decadal
29 Variability (PDV), and the shift of the Atlantic Inter Tropical Convergence Zone in response to changes in
30 the meridional SST gradient. A positive phase of the PDV weakens African monsoons (Meehl and Hu,
31 2006)(Figure AIV.8d) and MJO phase 4 suppresses convection over the equatorial Africa (Figure AIV.10a)
32 (see Annex IV). Other features influence specific subregions. For instance, El Niño events increase
33 precipitation in eastern Africa and decreases precipitation in southern Africa. Over southern Africa there is a
34 strong link between ENSO and droughts (Meque and Abiodun, 2015). The positive phase of the IOD
35 increases rainfall in eastern tropical Africa in boreal autumn to early winter (Figure AIV.5d), while the
36 negative phase induces the reduction in rainfall. West African monsoon is influenced by Atlantic Zonal
37 Mode (AZM) with decreased rainfall over the Sahel and increased rainfall over Guinea (Losada et al., 2010).
38 Positive Atlantic Multidecadal Variability (AMV) influences positive anomalies all year round over a broad
39 Mediterranean region, including North Africa.
40
41
42 Atlas.4.1.2 Findings from previous IPCC assessments
43
44 The most recent IPCC reports, AR5 and SR1.5 (Christensen et al., 2013; Hoegh-Guldberg et al., 2018), state
45 that over most parts of Africa, minimum temperatures have warmed more rapidly than maximum
46 temperatures during the last 50 to 100 years (medium confidence). In the same period minimum and
47 maximum temperatures have increased by more than 0.5°C relative to 1850-1900 (high confidence). While
48 the quality of ground observational temperature measurements tends to be high compared to that of
49 measurements for other climate variables, Africa remains an under-represented region as reported in SR1.5
50 (Hoegh-Guldberg et al., 2018; IPCC, 2018c). Based on the Coupled Model Intercomparison Project Phase 5
51 (CMIP5) ensemble and reported in IPCC AR5 and SR1.5, surface air temperatures in Africa are projected to
52 rise faster than the global average increase and likely to increase by more than 2°C and up to 6°C by the end
53 of the century relative to the late 20th century if global warming reaches 2°C (Bindoff et al., 2013; Niang et
54 al., 2014; Hoegh-Guldberg et al., 2018). The higher temperature magnitudes are projected during boreal
55 summer. Southern Africa is likely to exceed the global mean land surface temperature increase in all seasons
Do Not Cite, Quote or Distribute Atlas-41 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 by the end of the century. Temperature projections for East Africa indicate considerable warming under
2 RCP8.5 where average warming across all models is approximately 4°C by the end of the century.
3 According to SROCC, eastern Africa like other regions with smaller glaciers is projected to lose more than
4 80% of its glaciers by 2100 under RCP8.5 (medium confidence) (Hock et al., 2019b).
5
6 West Africa has also experienced an overall reduction of rainfall over the 20th century, with a recovery
7 towards the last 20 years of the century (Christensen et al., 2013). Over the last three decades rainfall has
8 decreased over East Africa especially between March and May/June. Projected rainfall changes over Africa
9 in the mid and late 21st century is uncertain. In regions of high or complex topography such as the Ethiopian
10 Highlands, downscaled projections indicate likely increases in rainfall and extreme rainfall by the end of the
11 21st century. However, Northern Africa and the south-western parts of South Africa are likely to have a
12 reduction in precipitation.
13
14 The consequence of increased temperature and evapotranspiration, and decreased precipitation amount, in
15 interaction with climate variability and human activities, have contributed to desertification in dryland areas
16 in Sub-Saharan Africa (medium confidence) as reported in SRCCL (Mirzabaev et al., 2019).
17
18
19 Atlas.4.2 Assessment and synthesis of observations, trends and attribution
20
21 Figure Atlas.11: shows observed trends in annual mean surface temperature and indicates it has been rising
22 rapidly over Africa over 1960–2015 and with significant increases in all regions of 0.1°C–0.2°C per decade
23 and higher over some northern, eastern and south-western regions (high confidence) (see also Interactive
24 Atlas). This is confirmed by an independent analysis performed for a longer period (1961–2018) over areas
25 where long-term homogeneous temperature time-series are available (Engelbrecht et al., 2015). More
26 specifically over the Horn of East Africa, the long-term mean annual temperature change between 1930 and
27 2014 showed two distinct but contrary trends: significant decreases between 1930 and 1969 and increases
28 from 1970 to 2014 (Ghebrezgabher et al., 2016). North Africa has an overall warming in observed seasonal
29 temperature (Barkhordarian et al., 2012; Lelieveld et al., 2016) with positive trends in annual minimum and
30 maximum temperatures (Vizy and Cook, 2012). Temperatures over West Africa have increased over the last
31 50 years (Mouhamed et al., 2013; Niang et al., 2014) with a spatially variable warming reaching 0.5°C per
32 decade from 1983 to 2010 (Sylla et al., 2016). West Africa has also experienced a decrease in the number of
33 cool nights as well as more frequent warm days and warm spells (Mouhamed et al., 2013; Ringard et al.,
34 2016). Similarly, East Africa has experienced a significant increase in temperature since the beginning of the
35 early 1980s (Anyah and Qiu, 2012) with an increase in seasonal mean temperature. Over South Africa,
36 positive trends were found in the annual mean, maximum and minimum temperatures for 1960–2003 in all
37 seasons, except for the central interior (Kruger and Shongwe, 2004; Zhou et al., 2010; Collins, 2011; Kruger
38 and Sekele, 2013; MacKellar et al., 2014), where minimum temperatures have decreased significantly
39 (MacKellar et al., 2014). Inland southern Africa, minimum temperatures have increased more rapidly than
40 maximum temperatures (New et al., 2006).
41
42 Most areas lack enough observational data to draw conclusions about trends in annual precipitation over the
43 past century. In addition, many regions of Africa have discrepancies between different observed precipitation
44 datasets (Sylla et al., 2013; Panitz et al., 2014). A statistically significant (95% confidence level) decrease in
45 rainfall and the number of rainy days is reported in autumn over the east, central and north-eastern parts of
46 South Africa in spring and summer during 1960–2010 (MacKellar et al., 2014; Kruger and Nxumalo, 2017).
47 Central Africa has experienced a significant decrease in total precipitation which is likely associated with a
48 significant decrease of the length of the maximum number of consecutive wet days (Aguilar et al., 2009).
49 Furthermore, rainfall decreased significantly in the Horn of Africa (Tierney et al., 2015) with the largest
50 reductions during the long rains season March to May (Lyon and Dewitt, 2012; Viste et al., 2013; Rowell et
51 al., 2015). Over mountainous areas significant increases are found in the number of rain days around the
52 southern Drakensberg in spring and summer during the period 1960–2010 (MacKellar et al., 2014).
53 Similarly, southern West Africa is observed to have had more intense rainfall from 1950 to 2014 during the
54 second rainy season of September to November (Nkrumah et al., 2019). The Sahel region also had more
55 intense rainfall throughout the rainy season (Panthou et al., 2014, 2018b, 2018a; Sanogo et al., 2015; Gaetani
Do Not Cite, Quote or Distribute Atlas-42 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 et al., 2017; Taylor et al., 2017; Biasutti, 2019) during the period 1980–2010. Southern African rainfall
2 shows a significant downtrend of –0.013 mm day–1 year–1 in recent decades and –0.003 mm day–1 year–1 for
3 longer periods during 1900–2010 (Jury, 2013) (low confidence).
4
5 Temperature increases over Africa in the 20th century can be attributed to the strong evidence of a continent-
6 wide anthropogenic signal in the warming (Hoerling et al., 2006; Min and Hense, 2007; Stott et al., 2010,
7 2011; Niang et al., 2014)(Figure 3.8). More specifically over West Africa, the clear emergence of
8 temperature change (Figure Atlas.11:) is due to the relatively small natural climate variability in the region
9 which generates narrow climate bounds that can be easily surpassed by relatively small climate changes
10 (Niang et al., 2014). Warming over North Africa is largely due to anthropogenic climate forcing (Knippertz
11 et al., 2003; Barkhordarian et al., 2012; Diffenbaugh et al., 2017).
12
13 The drying observed over the Sahel in the 1960s to 1970s has been attributed to warming of the South
14 Atlantic SST and southern African drying as a response to Indian Ocean warming (Hoerling et al., 2006; Dai,
15 2011). Enhanced rainfall intensity since the mid-1980s over the Sahel (Maidment et al., 2015; Sanogo et al.,
16 2015) is associated with increased greenhouse gases suggesting an anthropogenic influence (Biasutti, 2019)
17 (medium confidence). In the last decade, the changes in the timing of onset and cessation of rainfall over
18 Africa have been linked to changes in the progression of the tropical rainband and the Saharan Heat Low
19 (Dunning et al., 2018; Wainwright et al., 2019). Moreover, later onset and earlier cessation of Eastern Africa
20 rainfall is associated with a delayed and then faster movement of the tropical rainband northwards during the
21 boreal spring and northward shift of the Saharan Heat Low (Wainwright et al., 2019) driven by
22 anthropogenic carbon emissions and changing aerosol forcings (medium confidence). Over East Africa, the
23 drying trend is associated with an anthropogenic-forced relatively rapid warming of Indian Ocean SSTs
24 (Williams and Funk, 2011; Hoell et al., 2017); a shift to warmer SSTs over the western tropical Pacific and
25 cooler SSTs over the central and eastern tropical Pacific (Lyon and Dewitt, 2012); multidecadal variability of
26 SSTs in the tropical Pacific, with cooling in the east and warming in the west (Lyon, 2014); and the
27 strengthening of the 200-mb easterlies (Liebmann et al., 2017). However, decadal natural variability from
28 SST variations over the Pacific Ocean has also been associated with the drying trend of the East Africa
29 (Wang et al., 2014; Hoell et al., 2017) with an anthropogenic-forced rapid warming of Indian Ocean SSTs
30 (medium confidence).
31
32
33 Atlas.4.3 Assessment of model performance
34
35 Model development has advanced in the world, but Africa still lags as a focus and in its contribution (James
36 et al., 2018). None of the current generation of general circulation models (GCMs) was developed in Africa
37 (Watterson et al., 2014), and the relevant processes in the continent have not been the priority for model
38 development but treated in a one-size-fit-all approach (James et al., 2018) except for a few studies that
39 focused on convective-permitting climate projections (Stratton et al., 2018; Kendon et al., 2019). However,
40 there are growing efforts to boost African climate science by running and evaluating climate models over
41 Africa (Endris et al., 2013; Kalognomou et al., 2013; Gbobaniyi et al., 2014; Engelbrecht et al., 2015; Klutse
42 et al., 2016; Gibba et al., 2019).
43
44 The CMIP project previously did not result in improved performance for Africa (Flato et al., 2013; Rowell,
45 2013; Whittleston et al., 2017) and culling ensembles based on existing metrics for Africa fails to reduce the
46 range of uncertainty in precipitation projections (Roehrig et al., 2013; Yang et al., 2015; Rowell et al., 2016)
47 but biases over Africa are lower in CMIP6 compared to CMIP5 (Almazroui et al., 2020b). Nonetheless, the
48 CMIP5 ensemble has been evaluated over Africa to advance its application for climate research (Biasutti,
49 2013; Rowell, 2013; Dike et al., 2015; McSweeney and Jones, 2016; Onyutha et al., 2016; Wainwright et al.,
50 2019) as has, more recently, the CMIP6 ensemble (Almazroui et al., 2020b).
51
52 Coordinated Regional Downscaling Experiment (CORDEX) regional climate models have been widely
53 evaluated over Africa. They capture the occurrence of the West African Monsoon jump and the timing and
54 amplitude of mean annual cycle of precipitation and temperature over the homogeneous subregions of West
55 Africa (Gbobaniyi et al., 2014), simulate eastern Africa rainfall adequately (Endris et al., 2013) and over
Do Not Cite, Quote or Distribute Atlas-43 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Southern Africa capture the observed climatological spatial patterns of extreme precipitation (Pinto et al.,
2 2016). They simulate also the phasing and amplitude of monthly rainfall evolution and the spatial
3 progression of the wet season onset well over Southern Africa (Shongwe et al., 2015). However,
4 discrepancies and biases in present-day rainfall are reported over Uganda from the RCM-simulated rainfall
5 compared to three gridded observational datasets (Kisembe et al., 2018). Specifically, they reported that the
6 CORDEX models underestimate annual rainfall of Uganda and struggle to reproduce the variability of the
7 long and short rainy seasons.
8
9
10 Atlas.4.4 Assessment and synthesis of projections
11
12 Research over Africa has improved since AR5 and though SR1.5 (de Coninck et al., 2018) has synthesised
13 new information for the continent, there is still not enough literature on specific areas for assessment. CMIP5
14 and CMIP6 projections (Figure Atlas.16:) are for continued warming, with median projected regional
15 warming for 2080–2100 compared to 1995–2014 of between 1°C and 2°C under SSP1-2.6/RCP2.6
16 emissions and exceeding 4°C and in some regions 5°C under SSP5-8.5/RCP8.5 emissions. The central
17 interior of Southern and Northern Africa are likely to warm faster than equatorial and tropical regions
18 (Interactive Atlas). Projections from CMIP5 show that East Africa is likely to warm by 1.7°C–2.8 °C and
19 2.2°C–5.4°C under the RCP4.5 and RCP8.5 scenarios respectively in the period 2071–2100 relative to 1961–
20 1990 (Ongoma et al., 2018). Over southern Africa, areas in the south-western region of the subcontinent,
21 covering South Africa and parts of Namibia and Botswana, are projected to experience the largest increase in
22 temperature, which are expected to be greater than the global mean warming (Maúre et al., 2018). A large
23 ensemble of CORDEX Africa simulations have been used to project the impact of 1.5°C and 2°C global
24 warming levels (GWLs) (Klutse et al., 2018; Lennard et al., 2018; Maúre et al., 2018; Mba et al., 2018;
25 Nikulin et al., 2018; Osima et al., 2018). While a few studies addressed the whole African continent
26 (Lennard et al., 2018; Nikulin et al., 2018), some focused on specific regions of Africa (Diedhiou et al.,
27 2018; Klutse et al., 2018; Kumi and Abiodun, 2018; Maúre et al., 2018; Mba et al., 2018). CORDEX
28 simulations project robust warming over Africa in excess of the global mean (Lennard et al., 2018; Nikulin
29 et al., 2018) and over West Africa the magnitude of regional warming reaches the 2080–2100 global
30 warming level one to two decades earlier (Mora et al., 2013; Niang et al., 2014; Sylla et al., 2016; Klutse et
31 al., 2018). Temperature increases projected under RCP8.5 over Sudan and northern Ethiopia implying that
32 the Greater Horn of Africa would warm faster than the global mean relative to 1971–2000 (Osima et al.,
33 2018). Over North Africa, summer mean temperatures from CORDEX, CMIP5 (RCP8.5) and CMIP6
34 (SSP5-8.5) are projected to increase beyond 6.0°C by the end of the century with respect to the period 1970–
35 2000 (Schilling et al., 2012; Ozturk et al., 2018; Almazroui et al., 2020b), see also the Interactive Atlas. Note
36 that results for the CORDEX-AFR over the Mediterranean (MED) are consistent with those reported from
37 the CORDEX-EUR dataset (Figure Atlas.24:; Section Atlas.1.3), in agreement with Legasa et al. (2020).
38
39
40 [START FIGURE ATLAS.16 HERE]
41
42 Figure Atlas.16: Regional mean changes in annual mean surface air temperature and precipitation relative to the
43 1995–2014 baseline for the reference regions in Africa (warming since the 1850–1900 pre-
44 industrial baseline is also provided as an offset). Bar plots in the left panel of each region triplet
45 show the median (dots) and 10th–90th percentile range (bars) across each model ensemble for annual
46 mean temperature changes for four datasets (CMIP5 in intermediate colours; subset of CMIP5 used to
47 drive CORDEX in light colours; CORDEX overlying the CMIP5 subset with dashed bars; and CMIP6
48 in solid colours); the first six groups of bars represent the regional warming over two time periods
49 (near-term 2021–2040 and long-term 2081–2100) for three scenarios (SSP1-2.6/RCP2.6, SSP2-
50 4.5/RCP4.5, and SSP5-8.5/RCP8.5), and the remaining bars correspond to four global warming levels
51 (GWL: 1.5°C, 2°C, 3°C, and 4°C). The scatter diagrams of temperature against precipitation changes
52 display the median (dots) and 10th–90th percentile ranges for the above four warming levels for
53 December-January-February-March (DJFM; middle panel) and June-July-August-September (JJAS;
54 right panel), respectively; for the CMIP5 subset only the percentile range of temperature is shown,
55 and only for 3°C and 4°C GWLs. Changes are absolute for temperature (in °C) and relative (as %) for
56 precipitation. See Section Atlas.1.3 for more details on reference regions (Iturbide et al., 2020) and
Do Not Cite, Quote or Distribute Atlas-44 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Section Atlas.1.4 for details on model data selection and processing. The script used to generate this
2 figure is available online (Iturbide et al., 2021) and similar results can be generated in the Interactive
3 Atlas for flexibly defined seasonal period. Further details on data sources and processing are available
4 in the chapter data table (Table Atlas.SM.15).
5
6 [END FIGURE ATLAS.16 HERE]
7
8
9 Projected rainfall changes over Africa in the mid and late 21st century are uncertain in many regions, highly
10 variable spatially and with differing levels of model agreements (Figure Atlas.16:) though with robust
11 projections of decreases in MED and WSAF and increases in NEAF and SEAF by 2080–2100 under high
12 emissions (Interactive Atlas). Some uncertainties are reported over parts of Africa from CORDEX
13 projections (Dosio and Panitz, 2016; Endris et al., 2016; Klutse et al., 2018). For example, large uncertainties
14 are associated with projections at 1.5°C and 2°C of global warming over Central Africa (Mba et al., 2018)
15 and over the Sahel (Gbobaniyi et al., 2014; Sylla et al., 2016). Over southern Africa, enhanced warming is
16 projected to result in a reduction in mean rainfall across the region (Maúre et al., 2018) and in particular over
17 the Limpopo Basin and smaller areas of the Zambezi Basin in Zambia, and also parts of Western Cape,
18 South Africa, under a global warming of 2°C. The projections of reduced precipitation in summer rainfall
19 regions of southern Africa are associated with delayed wet season onset in spring (Dunning et al., 2018) due
20 to a northward shift and delayed breakdown of the Congo Air Boundary (Howard and Washington, 2020).
21 However, projected rainfall intensity over southern Africa is likely to increase and be magnified under
22 RCP8.5 compared with RCP4.5 for the period 2069–2098 relative to the reference period (1976–2005)
23 (Pinto et al., 2018). For West Africa, rainfall projection is uncertain because of the contrasting signals from
24 models (Dosio et al., 2019). Nonetheless, West Africa river basin-scale irrigation potential would decline
25 under 2°C of global warming even for areas where water availability increases (Sylla et al., 2018). The
26 western and eastern Sahel are projected as hotspots for delayed rainfall onset dates of about four days and six
27 days causing reduced length of rainy season in the 1.5°C to 2°C warmer climates under RCP4.5 and RCP8.5
28 scenarios (Kumi and Abiodun, 2018). Projected delay in rainfall cessation dates and a longer length of rainy
29 season over the western part of the Guinea coast is likely under the same scenarios (Sellami et al., 2016;
30 Kumi and Abiodun, 2018)(Figure Atlas.16:). There is a tendency towards an increase in annual mean
31 precipitation over central Sahel and eastern Africa (Figure Atlas.16, (Nikulin et al., 2018) especially over the
32 Ethiopian highlands with up to 0.5 mm day–1 (Osima et al., 2018).
33
34
35 Atlas.4.5 Summary
36
37 The rate of surface temperature increase has generally been more rapid in Africa than the global average and
38 by at least 0.1°C–0.2°C during 1960–2015 (high confidence). Minimum temperatures have increased more
39 rapidly than maximum temperatures over inland southern Africa (medium confidence). Since 1970, mean
40 temperature over East Africa has shown an increasing trend but showed a decreasing trend in the previous 40
41 years (medium confidence).
42
43 The Horn of Africa has experienced significantly decreased rainfall during the long rain season from March
44 to May (high confidence) and drying trends in this and other parts of Africa are attributable to oceanic
45 influences (high confidence), resulting from both internal variability and anthropogenic causes. Drying over
46 the Sahel in the last century was attributed to an increase in the South Atlantic SST and more recently over
47 southern African as a response to anthropogenic-forced Indian Ocean warming. Drying over East Africa is
48 associated with decadal natural variability in SSTs over the Pacific Ocean. The enhanced rainfall intensity
49 over the Sahel in the last two decades is associated with increased greenhouse gases indicating an
50 anthropogenic influence (medium confidence).
51
52 Relative to the late 20th century, annual mean temperature over Africa is projected to rise faster than the
53 global average (very high confidence) with the increase likely to exceed 4°C by the end of the century under
54 RCP8.5 emissions. The central interior of Southern and Northern Africa are likely to warm faster than
55 equatorial and tropical regions (high confidence).
Do Not Cite, Quote or Distribute Atlas-45 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1
2 There are contrasting signals in the projections of rainfall over some parts of Africa until the end of the 21st
3 century (high confidence) though changes in any given region are generally projected with medium
4 confidence. In regions of high or complex topography such as the Ethiopian Highlands, downscaled
5 projections indicate increases in rainfall by the end of the 21st century. However, Northern Africa and the
6 south-western parts of South Africa are likely to have a reduction in precipitation under higher warming
7 levels (high confidence). Over West Africa, rainfall is projected to decrease in the Western Sahel subregion
8 (medium confidence) and increase in the central Sahel subregion (low confidence) and along the Guinea
9 Coast subregion (medium confidence). Rainfall amounts are projected to reduce over the western part of East
10 Africa, but to increase in the eastern part of the region (medium confidence). Southern Africa is projected to
11 have a reduction in annual mean rainfall but increases in rainfall intensity by 2100 (medium confidence).
12
13
14 Atlas.5 Asia
15
16 The assessment in this section focuses on changes in average temperature and precipitation (rainfall and
17 snow), including the most recent years of observations, updates to observed datasets, the consideration of
18 recent studies using CMIP5 and those using CMIP6 and CORDEX simulations. Assessment of changes in
19 extremes are in Chapter 11 (Table 11.7–11.9) and climatic impact-drivers in Chapter 12 (Table 12.4). It
20 covers most Asian territories of the region (Figure Atlas.17:) with the exception of the Russian Arctic, RAR,
21 which is assessed as part of the Arctic in section 11.2. These include West and East Siberia (WSB, ESB) and
22 the Russian Far East (RFE) in the north; West and East Central Asia (WCA, ECA), the Tibetan Plateau
23 (TIB) and East Asia (EAS); and the Arabian Peninsula (ARP), South and Southeast Asia (SAS, SEA) in the
24 south. Figure Atlas.17: supports the assessment of regional mean changes in annual mean surface air
25 temperature and precipitation over Asia. Due to the high climatological and geographical heterogeneity of
26 Asia, the assessment is performed over five sub-continental areas: East Asia (EAS+ECA), North Asia
27 (WSB+ESB+RFE), South Asia (SAS+TIB), Southeast Asia (SEA) and Southwest Asia (ARP+WCA).
28
29
30 [START FIGURE ATLAS.17 HERE]
31
32 Figure Atlas.17: Regional mean changes in annual mean surface air temperature and precipitation relative to the
33 1995–2014 baseline for the reference regions in Asia (warming since the 1850–1900 pre-
34 industrial baseline is also provided as an offset). Bar plots in the left panel of each region triplet
35 show the median (dots) and 10th–90th percentile range (bars) across each model ensemble for annual
36 mean temperature changes for four datasets (CMIP5 in intermediate colours; subset of CMIP5 used to
37 drive CORDEX in light colours; CORDEX overlying the CMIP5 subset with dashed bars; and CMIP6
38 in solid colours); the first six groups of bars represent the regional warming over two time periods
39 (near-term 2021–2040 and long-term 2081–2100) for three scenarios (SSP1-2.6/RCP2.6, SSP2-
40 4.5/RCP4.5, and SSP5-8.5/RCP8.5), and the remaining bars correspond to four global warming levels
41 (GWL: 1.5°C, 2°C, 3°C, and 4°C). The scatter diagrams of temperature against precipitation changes
42 display the median (dots) and 10th–90th percentile ranges for the above four warming levels for
43 December-January-February (DJF; middle panel) and June-July-August (JJA; right panel),
44 respectively; for the CMIP5 subset only the percentile range of temperature is shown, and only for
45 3°C and 4°C GWLs. Changes are absolute for temperature (in °C) and relative (as %) for
46 precipitation. See Section Atlas.1.3 for more details on reference regions (Iturbide et al., 2020) and
47 Section Atlas.1.4 for details on model data selection and processing. Further details on data sources
48 and processing are available in the chapter data table (Table Atlas.SM.15).
49
50 [END FIGURE ATLAS.17 HERE]
51
52
53
54
55
56
57
Do Not Cite, Quote or Distribute Atlas-46 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.5.1 East Asia
2
3 Atlas.5.1.1 Key features of the regional climate and findings from previous IPCC assessments
4
5 Atlas.5.1.1.1 Key features of the regional climate
6 The climatic regions defined for East Asia include central and eastern China, Japan and Korea Peninsula
7 (regions ECA and EAS in Figure Atlas.17:). East Asia is significantly influenced by monsoon systems (see
8 Section 8.3.2.4.2). The seasonal advance or retreat of the East Asian summer monsoon (EASM) rain band is
9 crucial to local climate. The East Asian winter monsoon (EAWM) has significant influence on the weather
10 and climate over East Asia and plays an important role in regulating winter temperatures including strong
11 cold events and snowstorms (Wang and Chen, 2014; Wang and Lu, 2016). The East Asian monsoons exhibit
12 considerable variability on a wide range of time scales, including notable interannual variabilities that
13 includes an effect of the El Niño Southern Oscillation (ENSO) (Wang et al., 2000) and the Indian Ocean
14 Dipole (IOD) (Takaya et al., 2020), and significant inter-decadal variabilities in the 20th century resulted
15 from the effect of Pacific Decadal Variability (PDV) (Zhou et al., 2009), see also Annex IV and Table
16 Atlas.1:. The thermal conditions of both the Tibetan Plateau and related ocean regions play key roles in
17 modulating the intensity of the monsoon circulation. The East Asian monsoons are mainly driven by land-sea
18 thermal contrast and, thus, are deeply affected by global climate change (Ding et al., 2014; Gong et al.,
19 2018).
20
21
22 Atlas.5.1.1.2 Findings from previous IPCC assessments
23 The findings of the IPCC AR5 (Christensen et al., 2013) stated that the EASM and EAWM circulations have
24 experienced an inter-decadal scale weakening since the 1970s, leading to a warmer climate in winter and
25 enhanced mean precipitation along the Yangtze River Valley (30°N) but deficient mean precipitation in
26 North China in summer. Since the middle of the 20th century, it is likely that there has been an increasing
27 trend in winter temperatures across much of Asia (Christensen et al., 2013). The numbers of cold days and
28 nights have decreased and the numbers of warm days and nights have increased over Asia (Hartmann et al.,
29 2013). It is likely that there are decreasing numbers of snowfall events where increased winter temperatures
30 have been observed (Hartmann et al., 2013). The SRCCL reports a land-use change induced cooling as large
31 as –1.5°C in eastern China between 1871 and 2007 (Hartmann et al., 2013). The summer rainfall amount
32 over East Asia shows no clear trend during the 20th century.
33
34 The IPCC AR5 (Christensen et al., 2013) reports a significant increase in mean temperatures in southeastern
35 China, associated with a decrease in the number of frost days under the SRES A2 scenario. The CMIP5
36 model projections indicate an increase of temperature in both boreal winter and summer over East Asia for
37 RCP4.5. Based on CMIP5 model projections, there is medium confidence in an intensified EASM and
38 increased summer precipitation over East Asia. More than 85% of CMIP5 models show an increase in mean
39 precipitation of the EASM, while more than 95% of models project an increase in heavy precipitation events
40 (Christensen et al., 2013). SROCC states that future projections of annual precipitation indicate increases of
41 the order of 5% to 20% over the 21st century in many mountain regions, including the Himalaya and East
42 Asia (Hock et al., 2019b). SR1.5 reports that statistically significant changes in heavy precipitation between
43 1.5°C and 2°C of global warming are found in East Asia (Hoegh-Guldberg et al., 2018).
44
45
46 Atlas.5.1.2 Assessment and synthesis of observations, trends and attribution
47
48 Summer (June–August) mean temperature in Eastern China has increased by 0.82°C since reliable
49 observations were established in the 1950s (Sun et al., 2014). Based on historical meteorological
50 observations, the best estimate of the linear trend of annual mean surface air temperature (SAT) for China
51 with 95% uncertainty ranges is 0.38 ± 0.05°C per decade for 1979–2015 (Li et al., 2017). From 1960 to
52 2010, the increasing trend of temperature was about 0.34°C per decade in the arid region of northwest China,
53 higher than the average over China (Li et al., 2012a; Xu et al., 2015). Over South Korea, warming is 1.4–2.6
54 times larger than global trends. The increase is 1.90°C during 1912–2014 and 0.99°C during 1973–2014
55 (Park et al., 2017) with a 25–45% urbanization contribution. The annual temperature increased in large cities
Do Not Cite, Quote or Distribute Atlas-47 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 at a rate of 0.29 ± 0.08°C per decade compared with 0.11 ± 0.08°C per decade in other stations in South
2 Korea from 1960 to 2010 (Kim et al., 2016a). A relatively high increase in annual mean temperature at the
3 rate of 3.0°C per century was detected in the Tokyo metropolitan area for the period 1901–2015 (Matsumoto
4 et al., 2017). Trends of annual temperature for the period of 1960–2015 are shown in Figure Atlas.11:. Most
5 areas of East Asia have significant warming trends exceeding 0.1°C per decade, and the strongest warming
6 (0.3°C–0.4°C per decade) occurs in northern China.
7
8 Observational studies indicated significant decadal variations in the EAWM (Wang and Lu, 2016; He et al.,
9 2017). It weakened significantly around the late 1980s, being relatively strong during 1976–1987 and weaker
10 during 1988–2001. The EAWM has recovered in intensity after 2004 and caused frequent and prevalent
11 severe cold spells, as well as a number of unusually harsh cold winters in many parts of East Asia during the
12 period 2004–2012 (Wang and Chen, 2014; Kug et al., 2015; Ge et al., 2016; Gong et al., 2018). Negative
13 zonal mean winter SAT anomalies were observed over the whole of East Asia from 1980 to 1988, with
14 positive anomalies observed over high and low latitudes from 1988 to 2010 (Miao and Wang, 2020).
15
16 Precipitation trends over East Asia show considerable regional differences (medium confidence). Mean
17 precipitation has shown negligible sensitivity to the warming trend with consequently limited overall trends
18 in China though summer rainfall daily frequency and intensity show respectively decreasing and increasing
19 trends from 1961 to 2014 (Zhou and Wang, 2017). The summer precipitation trends over eastern China
20 display a dipole pattern, characterized by positive anomalies in the central-eastern China along the Yangtze
21 River Valley and negative anomalies in the north China since the 1950s (Section 8.3.2.4.2). This pattern has
22 changed with the enhanced rainfall in the Huaihe River Valley and decreased in the regions south of the
23 middle and lower reaches of the Yangtze River Valley since the 2000s (Liu et al., 2012; Zhao et al., 2015).
24 The climate in northwest China changed from ‘warm-dry’ to ‘warm-wet’ condition in the mid-1980s (Peng
25 and Zhou, 2017; Wang et al., 2020), with an increases rate of annual precipitation of about 3.7% per decade
26 from 1961 to 2015 (Wu et al., 2019a) and 11.2 mm per decade between 1960 and 2011 in northern Xinjiang
27 (Xu et al., 2015). Mean rainfall and the number of rainy days during the Meiyu-Baiu-Changma period from
28 June to September have increased during 1973–2015 in Korea (Lee et al., 2017). The precipitation trend has
29 caused a large increase in summer precipitation at a rate of 40.6 ± 4.3 mm per decade, resulting in an
30 increase of annual precipitation of 27.7 ± 5.5 mm per decade in South Korea from 1960 to 2010 (Kim et al.,
31 2016a). Precipitation amounts exhibited a slight decrease at both the annual and seasonal scales in Japan for
32 the period 1901–2012 (Duan et al., 2015).
33
34 Agriculture intensification through oasis expansion in Xinjiang region has increased summer precipitation in
35 the Tianshan mountains (Zhang et al., 2009, 2019b; Deng et al., 2015; Guo and Li, 2015; Yao et al., 2016;
36 Xu et al., 2018; Cai et al., 2019) (high confidence from medium evidence with high agreement). However,
37 there is very low confidence of the effect of oasis expansion on the temperature warming trend (Han and
38 Yang, 2013; Li et al., 2013; Yuan et al., 2017).
39
40 In the context of climate warming, intense snowfalls have hit China frequently in recent winters and have
41 caused severe damages to the sustainability of the society (Sun et al., 2019). Observations generally show a
42 decrease in the frequency and an increase in the mean intensity of snowfalls in north-western, north-eastern
43 and south-eastern China and the eastern Tibetan Plateau since the 1960s (Zhou et al., 2018), but the results
44 may depend on the objective criteria for identifying wintertime snowfall (Luo et al., 2020a).
45
46
47 Atlas.5.1.3 Assessment of model performance
48
49 Current climate models perform poorly in simulating the mean precipitation in East Asia, including the phase
50 of the northward progression of the seasonal rain band (Zhang et al., 2018b). Although there has been an
51 improvement in the simulation of mean states, interannual variability, and past climate changes in the
52 progression from CMIP3 to CMIP5, some previously documented biases (such as the ridge position of the
53 western North Pacific Subtropical High and the associated rainfall bias) are still evident in CMIP5 models
54 (Sperber et al., 2013; Zhou et al., 2017). Most models capture the main characteristics of the winter mean
55 circulation over East Asia reasonably well, but they still suffer from difficulty in predicting the interannual
Do Not Cite, Quote or Distribute Atlas-48 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 variability of the EAWM (Shin and Moon, 2018). Models have improved from CMIP5 to CMIP6 for
2 climatological temperature and EAWM (Jiang et al., 2020a). Some CMIP6 models also show improvements
3 in simulating the annual mean and interannual variation of precipitation (Sellar et al., 2019; Tatebe et al.,
4 2019; Wu et al., 2019b). The performance of models is sensitive to cumulus convection schemes and
5 horizontal resolution (Haarsma et al., 2016; Wu et al., 2017; Kusunoki, 2018b). High-resolution atmospheric
6 general circulation models (AGCM) successfully reproduce the intensity and the spatial pattern of the EASM
7 rainfall (Li et al., 2015; Yao et al., 2017; Ito et al., 2020a) and improve the simulation of the diurnal cycle of
8 precipitation rates and the probability density distributions of daily precipitation over Korea, Japan and
9 northern China (Lin et al., 2019), but increasing horizontal resolution (at the typical scales used in GCMs) is
10 not always a panacea for solving model biases (Roberts et al., 2018).
11
12 Recent studies using CORDEX-EA models with resolution about 12–25 km showed that the RCMs produce
13 relatively more detailed regional features of the temperature distribution compared with the driving GCMs
14 (Tang et al., 2016). Over China, RCMs provide more spatial details and in general reduce the biases of their
15 driving GCMs, in particular in DJF (December–January–February) and over areas with complex topography
16 (Wu and Gao, 2020). However, RCMs also show biases in simulating East Asian precipitation and its
17 variability (Park et al., 2016; Zhou et al., 2016; Zou and Zhou, 2016), and do not always show added value
18 compared to the driving GCMs (Li et al., 2018a). For example, by comparing inter-GCM and inter-RCM
19 differences around the Japan archipelago, it was found that RCM generate relatively large differences in
20 precipitation (Suzuki-Parker et al., 2018). The RCM multi-model ensemble produces superior simulation
21 compared to that of a single model (Jin et al., 2016; Guo et al., 2018a). A comparative study of RCMs at
22 different spatial resolutions showed with coarse-resolution they present some limitations and high-resolution
23 RCMs offer added value for several evaluation metrics (Park et al., 2020).
24
25
26 Atlas.5.1.4 Assessment and synthesis of projections
27
28 The development of climate models provides a solid basis for projection of future monsoon changes under
29 different global warming scenarios. Coupled model simulations indicate that East Asia will likely experience
30 higher warming than the global mean conditions across all global warming levels (Figure Atlas.17:) and with
31 the projected warming greater in ECA than EAS. Also, in the CMIP6 ensemble, the multi-model mean and
32 90th percentile warming for a given period and emissions scenario are consistently greater than in the
33 CMIP5 ensemble. Larger warming magnitudes are projected to occur in the southern, north-western, and
34 north-eastern regions of China, parts of Mongolia, the Korean Peninsula, and Japan than in other regions (Li
35 et al., 2018b). Projections indicated that the winter SAT increases over the East Asian continent, and the
36 precipitation increases over the northern East Asian continent in winter reaching 1.5°C and 2.0°C global
37 warming under the RCP4.5 and RCP8.5 scenarios (Miao et al., 2020).
38
39 Projected annual precipitation changes in the CMIP5 and CMIP6 ensembles are positive for all warming
40 levels in ECA and for the higher warming levels in EAS. Changes in precipitation per degree Celsius global
41 warming are larger in DJF than in JJA in ECA but show smaller seasonal difference in EAS (Figure
42 Atlas.17). The EASM precipitation is projected to increase but with a complex spatial structure (Kitoh, 2017;
43 Moon and Ha, 2017). Simulations from CMIP5 models show that compared with the current summer
44 climate, both SAT and precipitation increase significantly over the East Asian continent during the 1.5°C
45 warming period (Chen et al., 2019a), and the main mode of EASM precipitation changing from tripolar to
46 dipolar (Wang et al., 2018). The increase in precipitable water in the wet EASM region is only slightly
47 greater than global average but the increase in precipitation is much greater (Li et al., 2019b). The monsoon
48 circulation in the lower troposphere is projected to strengthen due to the enhanced thermal forcing by the
49 Tibetan Plateau (He et al., 2019; He and Zhou, 2020), which causes the increased summer precipitation over
50 the East Asian continent. Precipitation over eastern China increases for almost all months under global
51 warming in projections from GCMs with different horizontal resolutions (Kusunoki, 2018a). Also, under
52 RCP scenarios, in the 21st century, mean precipitation is projected to increase (Kim et al., 2020) especially
53 in the late afternoons (Oh and Suh, 2018) over the Korean Peninsula due to global warming and associated
54 changes in EASM. Increase in JJA mean precipitation is projected in northern East Asia consistently among
55 the CMIP models, while northward migration of early summer East Asian rain band such as the Meiyu-Baiu-
Do Not Cite, Quote or Distribute Atlas-49 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Changma is delayed along with that of the mid-latitude westerly jet in the future (Horinouchi et al., 2019).
2 However, the geographical distribution of precipitation change tends to depend more on the cumulus
3 convection scheme (Ose, 2017) and horizontal resolution of models rather than on SST distributions. Under
4 the RCP4.5 and the RCP8.5 scenarios, the interannual variability in EASM rainfall is projected by the multi-
5 model ensemble mean to increase in the 21st century (Ren et al., 2017). Further studies showed a projected
6 increase in heavy rainfall together with increases in rainfall intensity (Endo et al., 2017). Multi-model
7 intercomparison indicates significant uncertainties in future projections of climate change in East Asia,
8 although precipitation increases consistently across models (Zhou et al., 2017). Simulations under RCP4.5
9 scenario project that the number of snow days will be reduced by the end of the 21st century relative to
10 1986–2005, primarily owing to the decline of light snowfall events. The total amount is projected to increase
11 in north-western China but decrease in the other subregions (Zhou et al., 2018).
12
13 The increasing temperature trends under RCP scenarios were consistently reproduced in projections using
14 CORDEX-EA models (Kim et al., 2016b) as reported in AR5 using GCMs. However, changes in annual and
15 seasonal mean precipitation exhibit significant inter-RCM differences with larger magnitudes and variability
16 than in the GCMs (Ham et al., 2016; Ozturk et al., 2017; Sun et al., 2018a; Zhang et al., 2018a). RCM
17 simulations project that the Meiyu-Baiu-Changma heavy rainfall will significantly increase in northern Japan
18 at the end of the 21st century under the RCP8.5 scenario (Osakada and Nakakita, 2018), but projected
19 precipitation amount and number of precipitation days in summer around and over Japan differ as a result of
20 RCM uncertainty (Suzuki-Parker et al., 2018). Annual total snowfall is projected to decrease in most parts of
21 Japan except for Japan's northern island under RCP2.6 (Kawase et al., 2021).
22
23 Statistically downscaling of 37 CMIP5 GCMs for Xinjiang, China projected pronounced temperature
24 increase of 0.27℃ to 0.51℃ per decade from 2021 to 2060 while precipitation change was projected to be
25 between –1.7% to 6.8% per decade and varying seasonally and spatially (Luo et al., 2018). A decrease of
26 precipitation was projected in the western region of Xinjiang during summer. More extreme rainfall events
27 were projected to occur during summer and autumn.
28
29
30 Atlas.5.1.5 Summary
31
32 In East Asia annual mean temperature has been increasing since the 1950s (high confidence). The linear
33 trend of annual mean surface air temperature likely exceeded 0.1°C per decade over most of East Asia from
34 1960 to 2015. Trends of annual precipitation show considerable regional differences with areas of both
35 increases and decreases (medium confidence) and with increases over northwest China and South Korea
36 (high confidence). Agricultural intensification through oasis expansion in Xinjiang region has increased
37 summer precipitation in the Tianshan mountains (high confidence).
38
39 GCMs still show poor performance in simulating the mean rainfall and its variability over East Asia,
40 especially over regions characterized by complex topography. The CMIP6 models have improved from
41 CMIP5 for climatological temperature and winter monsoon but show little improvements for the summer
42 monsoon. The RCMs produce relatively more detailed regional features, but do not always produce superior
43 simulations compared with the driving GCMs.
44
45 The annual mean surface temperature over East Asia will very likely increase under all emissions scenarios
46 and global warming levels. Larger warming magnitudes will likely occur in the northern part of EAS and in
47 ECA. Precipitation is likely to increase over land in most of EAS at the end of the 21st century under higher
48 emissions scenario (SSP3-7.0, RCP8.5, SSP5-8.5) and global warming levels and in ECA under all
49 emissions scenarios and global warming levels. Summer precipitation increase is likely to occur in East Asia,
50 corresponding to the strengthened summer monsoon circulation.
51
52
53
54
55
Do Not Cite, Quote or Distribute Atlas-50 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.5.2 North Asia
2
3 Atlas.5.2.1 Key features of the regional climate and findings from previous IPCC assessments
4
5 Atlas.5.2.1.1 Key features of the regional climate
6 North Asia extends from the Ural Mountains in the west to the Pacific Ocean in the east and from the
7 Russian Arctic in the north to Southwest Asia and East Asia in the south. Its most recognizable features are
8 boreal forests and permafrost. In AR6 North Asia is divided into three reference regions (Figure Atlas.17:):
9 West Siberia (WSB) with a continental climate, warm summers and cold winters, many waterlogged areas
10 and several natural zones due to a large extent from south to north and heterogeneity in regional climates;
11 East Siberia (ESB) which is mainly highland with extensive permafrost and a more severe continental
12 climate characterised by harsh, long winters and short, hot summers, and by less precipitation and snow
13 cover than in neighbouring regions; and the Russian Far East (RFE) with a monsoon-influenced climate, cold
14 winters and wet summers in the south and cold winters and cool summers almost without precipitation in the
15 north. WSB and ESB are mainly influenced by NAO and NAM (Annex IV.2.1) and AO with associated
16 atmospheric blocking by the Siberian High that exhibits a pronounced decadal-to-multidecadal variability
17 (see also Table Atlas.1). RFE is under the influence of the ENSO (Annex IV.2.3) and the PDV (Annex
18 IV.2.6) that mostly affect rainfall variability.
19
20
21 Atlas.5.2.1.2 Findings from previous IPCC assessments
22 In the previous IPCC assessment cycles, the three subregions comprising North Asia in this section along
23 with Eastern Europe and the Asian Arctic were considered as either Northern Eurasia or Russia in AR4 and
24 AR5. WGI AR5 stated that for North and Central Asia CMIP5 models had difficulty in representing
25 climatological means of both temperature and precipitation, which is partly related to the scarceness of
26 observational data in northern parts of the region and to issues related to the estimation of biases with coarse
27 resolution models (Christensen et al., 2013). In CMIP3 projections under different RCP scenarios, North
28 Asia temperatures increase more in winter (DJF) than summer (JJA) (Seneviratne et al., 2012). With most
29 models projecting increased precipitation significantly above the 20-year natural variability, it was
30 concluded that precipitation in northern Asia will very likely increase (Christensen et al., 2013).
31
32 SRCCL identified aridisation of the climate in southern Eastern Siberia between 1976 and 2016 as causing
33 an extension of the steppes polewards whilst climate change also extended the vegetation season, increasing
34 forest productivity in most of boreal Siberia, but increasing risk of wildfire and tree mortality (Mirzabaev et
35 al., 2019). SROCC noted the warming climate has caused permafrost thaw and loss of ground ice, and thus
36 land subsidence and collapse, disturbing ecosystems and human infrastructure. Permafrost stability,
37 hydrology and vegetation were also impacted by recent extensive fires burning into the organic soil layer
38 (Meredith et al., 2019). SR1.5 noted that future, higher levels of warming lead to greater impacts in key
39 systems such as the Siberian ecosystems, identified as one of the threatened systems (‘Reason for Concern 1
40 – RFC1’) (Hoegh-Guldberg et al., 2018) with impacts at 2°C expected to be greater than those at 1.5°C
41 (medium confidence).
42
43
44 Atlas.5.2.2 Assessment and synthesis of observations, trends and attribution
45
46 Increases in surface air temperature (SAT) have been observed since the mid-1970s over the whole of North
47 Asia (Frolov et al., 2014), and particularly over the north-eastern part (Gruza et al., 2015) (see also Figure
48 Atlas.18). Trends of annual SAT in the northern part of the region during the last decades were very likely
49 twice as strong as the global average (Figure Atlas.11) (Frolov et al., 2014; Mokhov, 2015; Sherstyukov,
50 2016) with trends in RFE of 0.8°C–1.2°C per decade for the 1976–2014 period and more intense warming
51 strengthening from south to north observed in spring in ESB (Frolov et al., 2014; Ippolitov et al., 2014;
52 Kokorev and Sherstiukov, 2015).
53
54 Recent strong warming in polar regions (Section Atlas.11.2) was accompanied by cooling in winter in mid-
55 latitude regions particularly in the southern part of WSB and ESB (Cohen et al., 2014; Ippolitov et al., 2014;
Do Not Cite, Quote or Distribute Atlas-51 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Gruza et al., 2015; Kharyutkina et al., 2016; Overland et al., 2016; Perevedentsev et al., 2017; Wegmann et
2 al., 2018). These temperature decreases were strongly correlated with significant warming over the Barents-
3 Kara Sea (greater than 2.5°C per decade during 2003–2017) and sea-ice loss suggesting a causal link (Outten
4 and Esau, 2012; Semenov et al., 2012; Overland et al., 2016; Semenov, 2016; Wegmann et al., 2018;
5 Meleshko et al., 2019; Susskind et al., 2019) though recent studies (Blackport et al., 2019; Clark and Lee,
6 2019) have shown that both phenomena result from mid-latitude circulation variability (see also Cross-
7 Chapter Box 10.1). In addition, significant warming in the last decade has halved the cooling trend in
8 southern WSB from –0.6°C per decade during 1976–2012 to –0.3°C per decade during 1976–2018 (Frolov et
9 al., 2014; Roshydromet, 2019) (high confidence).
10
11 Annual precipitation totals very likely increased over North Asia in the last half century along with more
12 heavy and less light precipitation, more freezing rain and less freezing drizzle (Wen et al., 2014; Groisman et
13 al., 2016; Ye et al., 2017; Chernokulsky et al., 2019)(see Figure Atlas.11 and Interactive Atlas). The highest
14 increase was observed over regions of Siberia and RFE with estimated trends of 10–25 mm per decade for
15 the 1976–2014 period (Kokorev and Sherstiukov, 2015) or 5% per decade for the 1976–2018 period
16 (Roshydromet, 2019). Increases over southern RFE are the largest (over 50 mm per decade) and mostly due
17 to positive changes in convective precipitation intensity in the region in summer season (JJA) during 1966–
18 2016 (medium confidence) (Chernokulsky et al., 2019). A decreasing trend was observed in central WSB,
19 northern ESB, the Baikal and Transbaikal regions, the Amur River region, and Primorie territories of RFE
20 (the Kamchatka and Chukchi Peninsulas) with up to –20 mm per decade for the 1976–2014 period (Kokorev
21 and Sherstiukov, 2015) or 15–20% per decade for the 1976–2018 period (Roshydromet, 2019). Overall, solid
22 precipitation predominantly decreased in North Asia and very likely caused both less Snow Cover Extent
23 (SCE) and Snow Water Equivalent (SWE), attributable to the anthropogenic influence with high confidence
24 (Sections 2.3.2.2, 3.4.2).
25
26 Snow characteristics depend on both temperature and precipitation, and observed trends over North Asia
27 show large spatial heterogeneity and interannual variability (Figure Atlas.18:) leading to medium confidence
28 that maximum snow depth has increased over Siberia, the Okhotsk sea coast and in southern RFE since
29 1960s (Callaghan et al., 2011; Loginov et al., 2014), with trends during 1976–2016 of 1.8 cm (in WBS), 1.1
30 cm (in ESB), and 4.6 cm (in RFE) per decade (Bulygina et al., 2017). Snow cover duration increased in
31 Yakutia, Sakhalin Island and some other coastal areas of the Pacific Ocean in RFE during 1980–2009
32 (Callaghan et al., 2011) and decreased in WSB and ESB (Bulygina et al., 2017; Roshydromet, 2019).
33 However, Gorbatenko et al. (2019) reported that in southeastern WSB maximal snow depth has increased by
34 5–20 cm and duration of steady snow cover by between 4 and 10 days during 1989–2016 (Figure Atlas.18).
35
36
37 [START FIGURE ATLAS.18 HERE]
38
39 Figure Atlas.18: Linear trends for the 1980–2015 period based on station data from the World Data Centre of
40 the Russian Institute for Hydrometeorological Information (RIHMI-WDC) (Bulygina et al.,
41 2014). (a) Snow season duration from 1 July to 31 December (days per decade); (b) snow season
42 duration from 1 January to 30 June (days per decade); (c) maximum annual height of snow cover (mm
43 per decade). Trends have been calculated using ordinary least squares regression and the crosses
44 indicate nonsignificant trend values (at the 0.1 level) following the method of Santer et al. (2008) to
45 account for serial correlation. Further details on data sources and processing are available in the
46 chapter data table (Table Atlas.SM.15).
47
48 [END FIGURE ATLAS.18 HERE]
49
50
51 Atlas.5.2.3 Assessment of model performance
52
53 Temperature trends and means derived from reanalysis datasets (JRA-25, MERRA) correctly represented
54 temperature variability shown in observational data over the Asian territory of Russia for the 1976–2010
55 period (Loginov et al., 2014). Assessment of CRU TS 3.22, CRUTEMP4, ERA-Interim and NCEP2 datasets
56 against station data over North Asia for annual and seasonal air temperature has shown that the ERA-Interim
Do Not Cite, Quote or Distribute Atlas-52 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 reanalysis outperforms others for the 1981–2005 period (Kokorev and Sherstiukov, 2015). The latter
2 reanalysis also underestimates summer precipitation and shows large wet biases over Northeast Asia during
3 spring and underestimates mean seasonal temperature over northeast Asia in spring (MAM), autumn (SON),
4 and winter (DJF) but overestimate it in summer (JJA) compared with the CRU dataset (medium confidence)
5 (Ozturk et al., 2017; Top et al., 2021).
6
7 GCMs capture the main synoptic processes affecting North Asia and the CMIP5 ensemble simulates the
8 temporal evolution of the magnitude and position of the Siberian High (SH) over the period 1872–2005 (Fei
9 and Yong-Qi, 2015). CMIP5 models simulate a weakened intensity of the winter SH and a strengthened
10 interannual variability compared to observations (Fei and Yong-Qi, 2015). The characteristics of blocking
11 events over the region (number, duration, intensity and frequency) were reasonably well reproduced by
12 GCMs (Mokhov et al., 2014) and most overestimate the annual mean temperature over northern Eurasia
13 (Interactive Atlas ). Biases in simulated annual surface air temperature simulation primarily come from
14 winter (DJF) season and are relatively smaller in other seasons (Miao et al., 2014; Peng et al., 2019). Most
15 GCMs capture the main decadal SAT trend (Miao et al., 2014) though CMIP5 GCMs fail to capture the
16 decreasing temperature trend over East Siberia (Fei and Yong-Qi, 2015). Possible causes of GCMs inability
17 to represent the recent slowdown of warming is discussed more in Cross-Chapter Box 3.1. For CMIP5,
18 models with higher resolution do not always perform better than those with lower resolutions (medium
19 confidence) (Miao et al., 2014).
20
21 Sixteen CMIP5 model simulations of SAT variability over Eurasia were evaluated against CRU observations
22 for permafrost subregions (Peng et al., 2019), showing a warm bias in northwest Eurasia capturing the
23 climate warming over the 20th century and its acceleration during the late 20th century. CMIP5 GCMs
24 generally underestimate daily temperature range compared with observations over north-eastern Russia
25 (Sillmann et al., 2013; Lindvall and Svensson, 2015). Currently there is no literature on the CMIP6 ensemble
26 over the region though a few single-model studies are available (Voldoire et al., 2019; Wu et al., 2019b).
27
28 There is very limited use of RCMs for North Asia. CORDEX-CAS covers North Asia except parts of RFE
29 and ARCTIC-CORDEX the northern regions (Figure Atlas.6:). For CORDEX-CAS three RCMs (REMO,
30 ALARO-0 and CLMcom) have been used and have warm biases for maximum temperatures, cold biases for
31 minimum temperatures and a wet bias in the north during the winter (Top et al., 2021). Rain gauges,
32 however, are known to have problems in terms of measuring properly solid precipitation (e.g., drifting snow)
33 which can greatly affect the accuracy of precipitation observations over North Asia (Harris et al., 2014).
34
35
36 Atlas.5.2.4 Assessment and synthesis of projections
37
38 CMIP5 and CMIP6 projections are consistent in the direction and ranges of surface temperature change
39 which are higher than the global average and with ensemble mean warming of around 6°C for the 4°C GWL.
40 Projected precipitation changes are also consistent with significant increases in winter, of up to 40% in the
41 ensemble mean for the highest warming levels, and lower increases in summer except for WSB where
42 changes are small and suggest drying at the 4°C GWL (Figure Atlas.17:, Interactive Atlas).
43
44 The CMIP5 ensemble projects a warming of the annual mean SAT over northern Eurasia in the 21st century,
45 likely in the range of 0.8°C–1.0°C (RCP2.6), 2.3°C–3.1°C (RCP4.5) and up to 7.2°C (RCP8.5) (Miao et al.,
46 2014; Peng et al., 2019). Mid-latitude permafrost subregions in Eurasia are projected to warm more than the
47 global mean and non-permafrost territories, with ensemble area-averaged changes of 1.7°C (RCP2.6), 3.2°C
48 (RCP4.5) or 6.4°C (RCP8.5) in 2081–2100 relative to 1986–2005 (Peng et al., 2019).
49
50 Over the Central Asia CORDEX domain, RegCM4.3.5 simulations driven by two different CMIP5 GCMs
51 (HadGEM2-ES and MPI-ESM-MR) project SAT warming for 2071–2100 relative to 1971–2000 of about
52 3°C–4°C during the summer for RCP4.5 to over 7°C for all seasons for RCP8.5. Projected warming is most
53 evident on the large continental Siberian Plateau with boreal and sub-boreal climates and biomes (i.e., taiga
54 forests and tundra) during the winter season (Ozturk et al., 2017). The Voeikov Main Geophysical
55 Observatory (MGO) RCM, driven by five CMIP5 GCMs for the RCP8.5 scenario, projects a faster increase
Do Not Cite, Quote or Distribute Atlas-53 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 in annual minimum temperature as compared with maximum temperature over the whole territory of Russia
2 (Kattsov et al., 2017), and the smallest change in growing season lengths (i.e., periods with daily
3 temperatures over 5°C, 10°C and 15°C) in the area of northern taiga in WSB and ESB comparable with other
4 territories of Russia during the 21st century (Torzhkov et al., 2019).
5
6 For precipitation, MGO RCM projections for the Arctic-CORDEX domain under the RCP8.5 scenario are
7 for increases in annual totals for northern North Asia, a decrease in summer over ESB for 2006–2100
8 relative to 1951–2005 and significant increases in the upper limit of intense precipitation over most of the
9 region in winter (Kattsov et al., 2017; Khlebnikova et al., 2018). Other RCM projections show that in most
10 seasons and for all future periods, precipitation in Siberia is not projected to change with respect to the
11 1971–2000 period, except under the RCP8.5 scenario for the winter and autumn (Ozturk et al., 2017). This
12 very limited and controversial evidence leads to low confidence in RCM precipitation projections for North
13 Asia and since the projections of GCMs and ESMs are more physically consistent, assessment of future
14 precipitation changes is based on CMIP5/CMIP6 presented in Figure Atlas.17: and the Interactive Atlas.
15
16
17 Atlas.5.2.5 Summary
18
19 Annual surface air temperature and precipitation have very likely increased and maximum snow depth has
20 likely increased over most of North Asia since the mid-1970s. The highest warming has been found in spring
21 in ESB and RFE, strengthening from south to north with linear trends of 0.8°C–1.2°C per decade over the
22 1976–2014 period (high confidence). A temperature decrease was identified just in winter in the southern
23 part of WSB and ESB as a result of natural variability, but halved from –0.6°C per decade in 1976–2012 to –
24 0.3°C per decade for the longer 1976–2018 period due to recent warmer winters (high confidence). Over
25 North Asia annual precipitation increases with estimated trends of 5–15 mm per decade in the 1976–2014
26 period have been recorded with an exception over the Kamchatka and the Chukchi Peninsulas where
27 decreases up to –20 mm per decade in the same period have been found (medium confidence). Snow cover
28 duration has very likely decreased over Siberia and increases in maximum snow depths of 1.8 cm, 1.1 cm,
29 and 4.6 cm per decade have been observed for WSB, ESB, and RFE respectively from 1976 to 2016 (limited
30 evidence).
31
32 Most of the CMIP5 and some CMIP6 GCMs overestimate the annual mean air temperature and precipitation
33 over North Asia region (medium confidence). GCMs generally represent the observed decadal temperature
34 trend (medium confidence) and biases primarily come from the winter (DJF) season (high confidence).
35 Results of a very limited number of RCMs applied over the whole region show that they have warmer biases
36 for maximum and colder biases for minimum temperatures (medium agreement, limited evidence). Sparsity
37 of observational data particularly in the northern part of ESB and the whole of the RFE results in low
38 confidence in the assessments of model performance in North Asia.
39
40 Surface air temperature and precipitation in North Asia are projected to increase further (high confidence)
41 with warming higher than the global average and around 6°C at the 4°C GWL. Temperature change in 2080–
42 2099 relative to 1981–2000 is likely in the range of 3°C in summer to 4.9°C in winter under the RCP4.5
43 scenario, and 5.6°C in summer to 9.7°C in winter under the RCP8.5 scenario. Precipitation is projected to
44 increase with ensemble mean changes of 9% in summer under both RCP4.5 and RCP8.5 and of 22% and
45 56% in winter respectively.
46
47
48 Atlas.5.3 South Asia
49
50 Atlas.5.3.1 Key features of the regional climate and findings from IPCC previous assessments
51
52 Atlas.5.3.1.1 Key features of the regional climate
53 The countries in this region are mostly semi-arid to arid and therefore depend heavily on the summer
54 monsoon (June–September, JJAS) which is when most of the precipitation falls over the South Asia region
55 (SAS) (Figure Atlas.17:). The topographic mechanical effect of the Tibetan Plateau (TIB) promotes moisture
Do Not Cite, Quote or Distribute Atlas-54 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 convergence downstream which triggers the early summer monsoon onset particularly over the Bay of
2 Bengal and South China. In winter, Westerly Disturbances (WD) bring moisture from the Atlantic Ocean.
3 The WD interaction with Himalayas causes precipitation over northern and western parts of South Asia that
4 is crucial to maintain the glacier mass balance. The observed teleconnection patterns over SAS for
5 temperature shows cooling effects during NAM and warming effect when in positive phase with ENSO,
6 IOB, AMM and AMV (Annex IV). IOD also influences South Asian precipitation (Annex IV).
7
8
9 Atlas.5.3.1.2 Findings from previous IPCC assessments
10 Recent IPCC reports assessed that it is very likely that the mean annual temperature over South Asia has
11 increased during the past century (Figure 2.21 in Hartmann et al., 2013, Figure 24-2 in Hijioka et al., 2014),
12 and the frequency of cold (warm) days and nights have decreased (increased) across most of Asia since
13 about 1950 (Figure 2.32 in Hartmann et al., 2013). AR5 assessed that there is high confidence that the large-
14 scale patterns of surface temperature are generally well simulated by the CMIP5 models though with
15 problems in some regions, particularly at higher elevations over the Himalayas (Flato et al., 2013). CMIP5
16 models projected for the 21st century a significant increase in temperature over South Asia (high confidence
17 from robust evidence) and in projections of increased summer monsoon precipitation (medium confidence)
18 (Collins et al., 2013). AR5 assessed there is high confidence that high-resolution regional downscaling,
19 which generate results complementary to those from global climate models, adds value to the simulation of
20 spatial variations in climate in regions with highly variable topography (e.g., distinct orography, coastlines),
21 and for mesoscale phenomena and extremes (Flato et al., 2013).
22
23 Inconsistent evidence was found on the declining trends in mean precipitation and increasing droughts from
24 1950 onwards considering 1960–1990 as the baseline period. Similarly, SREX (Table 3-3 in Seneviratne et
25 al., 2012) reported low confidence (due to lack of literature) in trends in climate indices related to extreme
26 precipitation events. The Indian summer monsoon circulation was found to have weakened, but this was
27 compensated by increased local atmospheric moisture content leading to more rainfall (medium confidence).
28 It is likely that the occurrence of snowfall events is decreasing in South Asia along with other regions due to
29 an increase in winter temperatures (Hock et al., 2019b). Based on satellite- and surface-based remote sensing
30 it is very likely that aerosol optical depth has increased over southern Asia since 2000.
31
32
33 Atlas.5.3.2 Assessment and synthesis of observations, trends and attribution
34
35 Recent studies show that annual mean land temperatures over India warmed at a rate of around 0.6°C per
36 century during 1901–2018, which was primarily contributed by a significant increase in annual maximum
37 temperature of 1.0°C per century, while the annual minimum temperature showed a lesser increasing trend of
38 0.18°C per century during this period, with significant rise only in the recent few decades (1981–2010) at a
39 rate of 0.17°C per decade (Srivastava et al., 2017, 2019). The annual average of daily maximum and
40 minimum temperatures has increased over almost all Pakistan with a faster increasing trend in the south
41 (high confidence). Minimum temperatures have increased faster (0.17°C–0.37°C per decade) than maximum
42 temperatures (0.17°C–0.29°C per decade) with the diurnal temperature range reduced (–0.15°C to –0.08°C
43 per decade) in some regions (Khan et al., 2019).
44
45 There has been a noticeable declining trend in rainfall with monsoon deficits occurring with higher
46 frequency in different regions in South Asia (see also Section 8.3.2.4 on the South Asian monsoon).
47 Concurrently, the frequency of heavy precipitation events has increased over India, while the frequency of
48 moderate rain events has decreased since 1950 (high confidence) (Goswami et al., 2006; Dash et al., 2009;
49 Christensen et al., 2013; Krishnan et al., 2016; Kulkarni et al., 2017; Roxy et al., 2017). There is a
50 considerable spread in the seasonal and annual mean precipitation climatology and interannual variability
51 among the different observed precipitation datasets over India (Collins et al., 2013; Prakash et al., 2014; Kim
52 et al., 2018; Ramarao et al., 2018). Yet, the regions of agreement among datasets lend high confidence that
53 there has been a decrease in mean rainfall over most parts of the eastern and central north regions of India
54 (Singh et al., 2014; Roxy et al., 2015; Juneng et al., 2016; Krishnan et al., 2016; Guhathakurta and
55 Revadekar, 2017; Jin and Wang, 2017; Latif et al., 2017). A global modelling study with high resolution
Do Not Cite, Quote or Distribute Atlas-55 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 over South Asia (Sabin et al., 2013) indicated that a juxtaposition of regional land-use changes,
2 anthropogenic-aerosol forcing and the rapid warming signal of the equatorial Indian Ocean was crucial to
3 simulate the observed Indian summer monsoon weakening in recent decades (medium confidence).
4
5 A dipole-like structure in summer monsoon rainfall trends is observed over the northern Indo-Pakistan area
6 with significant increases over Pakistan and decreases over central north India resulting from strengthening
7 (weakening) of vertically integrated meridional moisture transport over the Arabian Sea (Bay of Bengal)
8 (low confidence) (Latif et al., 2017). Positive annual precipitation trends are observed in global and regional
9 datasets (Figure Atlas.11: and Interactive Atlas) during 1960–2015 and over arid provinces of Pakistan (for
10 Rabi and Kharif cropping seasons) during 1951–2015 of 2.8–34.8 mm per decade (Khan et al., 2020) imply
11 high confidence for increased precipitation in Pakistan. Observations located in the monsoon-dominated strip
12 in Pakistan indicate that the mean monsoon onset became earlier during 1971–2010 (Ali et al., 2020).
13
14 Snow and glaciers are the main water resources of all countries in South Asia. Glacier melting is mainly
15 controlled by natural phenomena but anthropogenic emissions of black carbon (BC) are now making a
16 significant contributing to total glacial melting in the Hindu Kush Himalaya (HKH) region (Menon, 2002;
17 Ramanathan et al., 2007; Ramanathan and Carmichael, 2008). BC concentration is seven to ten times higher
18 in mid-altitudes (1000–4000 metres above sea level) than at high altitudes (>4000 metres above sea level).
19 The concentration of BC sampled from the surface of snow/ice samples as well as ice core records shows
20 decreasing ice albedo and an acceleration in glacier melting (Wester et al., 2019) (see also Cross-Chapter
21 Box 10.4). Karakoram and western HKH snow cover is increasing, a phenomena known as the ‘Karakoram
22 anomaly’, and partially attributed to an increase in strength of westerly disturbances (Wester et al., 2019).
23
24 Significant glacier retreat has been observed since 1960 in TIB with lower rates in the interior of the region
25 (Yao et al., 2007). A large interdecadal variation in snow cover is also observed from 1960 to 2010.
26 Observations and model simulations showed that the increasing temperature of frozen grounds is leading to
27 thawing and reduced depth of permafrost, with further significant reductions projected under future global
28 warming scenarios (Yang et al., 2019) (medium confidence).
29
30
31 Atlas.5.3.3 Assessment of model performance
32
33 Whilst simulations of Indian summer monsoon rainfall (ISMR) have improved in CMIP5 compared to
34 CMIP3 in terms of northward propagation, time for peak monsoon and withdrawal (Sperber et al., 2013)
35 they fail to simulate the trends in monsoon rainfall and the post-1950 weakening of monsoon circulation
36 (Saha et al., 2014). This is partially attributed to the failure of coarse resolution CMIP5 models to simulate
37 fine-resolution processes such as orographic effects or land surface feedback, and problems in cloud
38 parameterization result in an overestimation of convective precipitation fraction (Singh et al., 2017a). In
39 CMIP6, a significant improvement is found in capturing the monsoon spatiotemporal patterns over India,
40 particularly in the Western Ghats and north-eastern Himalayan foothills (Gusain et al., 2020). Over Pakistan
41 the CMIP6 models simulate surface temperature better in JJA than DJF (Karim et al., 2020). The CMIP6
42 ensemble underestimates annual mean temperature over all of South Asian with mixed results for
43 precipitation (Almazroui et al., 2020c). The CMIP6 GCMs have a large cold bias in both mean annual
44 maximum and minimum temperatures in the complex Karakorum and Himalayan mountain ranges but
45 exhibit warm biases in mean annual minimum temperature in most of the rest of South Asia.
46
47 Regional climate model (RCM) downscaling of CMIP5 models as part of CORDEX South Asia uses higher
48 resolution (50 km) and improved surface fields such as topography and coastlines to resolve better the
49 complexities of the monsoon and other hydrological processes (Giorgi et al., 2009). The added value of their
50 simulations, relative to the driving GCMs, presents a complex picture. CORDEX RCMs better represent
51 spatial patterns of temperature (Sanjay et al., 2017), the spatial features of precipitation distribution
52 associated with the Indian summer monsoon (Choudhary and Dimri, 2018) and the simulation of monsoon
53 active- and break-phase composite precipitation (Karmacharya et al., 2016). The RCMs follow the driving
54 GCMs in underestimating seasonal mean surface air temperature and overestimating spatial variability in
55 precipitation. They amplify CMIP5 cold biases over almost the entire region, including over the HKH
Do Not Cite, Quote or Distribute Atlas-56 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 region, Afghanistan and southwest Pakistan during winter (Iqbal et al., 2017) and substantial cold biases of
2 6°C–10°C are found over the Himalayan watersheds of the Indus Basin (Nengker et al., 2018; Hasson et al.,
3 2019). Neither RCMs nor their driving CMIP5 GCMs reproduce well the region’s precipitation climatology
4 (Mishra, 2015). In addition, important characteristics of ISMR such as northward and eastward propagation,
5 onset, seasonal rainfall patterns, intra-seasonal oscillations and patterns of extremes did not show consistent
6 improvement (Singh et al., 2017b). Also, these RCM simulations have not demonstrated added value in
7 capturing the observed changes in ISMR characteristics over recent decades though RegCM4 simulations at
8 25 km showed high accuracy in capturing monsoon precipitation characteristics and atmospheric dynamics
9 in historical simulations (Ashfaq et al., 2020).
10
11 Evaluation of four global reanalysis products (ERA5 and ERA-Interim, JRA-55 and MERRA-2, Section
12 Atlas.1.4.2) for snow depth and snow cover over the TIB was performed against 33 in situ station
13 observations, Interactive Multisensor Snow and Ice Mapping System (IMS) snow cover and a satellite
14 microwave snow depth dataset (Orsolini et al., 2019). Most of the reanalyses showed a systematic
15 overestimation. Only ERA-Interim assimilated IMS snow cover at high altitudes, whereas ERA5 did not and
16 the excessive snowfall, snow depth and snow cover in ERA5 was attributed to this difference. The analysis
17 of annual maximum consecutive snow-covered days for the period 1980–2018 over TIB using JRA-55 and
18 Passive Microwave satellite observations showed decreasing trend in all time periods and in recent snow
19 seasons for MERRA-2 (Bian et al., 2020). The uncertainty assessment of model physics in snow modelling
20 over TIB using ground-based observations and high-resolution snow-cover satellite products from the
21 Moderate Resolution Imaging Spectroradiometer (MODIS) and FengYun-3B (FY3B) suggests that errors
22 can be overcome by optimizing parameterizations of the snow cover fraction rather than optimizing physics-
23 scheme options (Jiang et al., 2020b).
24
25
26 Atlas.5.3.4 Assessment and synthesis of projections
27
28 CMIP5 and CMIP6 surface temperature projections are consistent across the range of global warming levels
29 with increases greater than the global average, more so over TIB (Figure Atlas.17:). CMIP6 models show
30 higher sensitivity to greenhouse gas emissions, projecting higher warming for a given emission scenario. The
31 north-western parts of South Asia, mainly covering Karakorum and Himalayan mountain ranges, are
32 projected to warm more (over 6°C under SSP5-8.5, with higher warming in winters than in summer, see
33 Interactive Atlas) and this will accelerate glacier melting in the region. The warming pattern of maximum
34 and minimum temperatures are projected to intensify in higher latitudes compared with mid-latitudes of
35 South Asia in CMIP5 simulations for all RCP scenarios (Ullah et al., 2020).
36
37 Seasonal precipitation projections show increased winter precipitation over the western Himalayas and
38 decreased precipitation over the eastern Himalayas. On the other hand, summer precipitation projections
39 show a robust increase over most of South Asia, with the largest over the arid region of southern Pakistan
40 and adjacent areas of India, under SSP5-8.5 (Almazroui et al., 2020c). Daily bias-adjusted projections from
41 13 CMIP6 GCMs using all emission scenarios project a warmer (3°C–5°C) and wetter (13–30%) climate in
42 South Asia in the 21st century (Mishra et al., 2020).
43
44 With continued global warming and anticipated reductions in anthropogenic aerosol emissions in the future,
45 CMIP5 models project an increase in the mean and variability of summer monsoon precipitation over India
46 by the end of the 21st century, together with substantial increases in daily precipitation extremes (medium
47 confidence) (Gnanaseelan et al., 2020), see also Section 8.4.2.4 on changes in the South Asian monsoon. The
48 CMIP5 GCMs consistently project an increase in the moisture transport over the Arabian Sea and Bay of
49 Bengal towards the end of 21st century, an increase in moisture convergence and consequent increases in
50 monsoon rainfall over the Indo-Pakistan region which are higher under RCP8.5 thanRCP4.5 (Srivastava and
51 Delsole, 2014; Mei et al., 2015; Latif, 2017). Out of 20 CMIP5 GCMs, four showed an increase in
52 magnitude and lengthening of all-India summer monsoon under RCP8.5. The intensity of both strong and
53 weak monsoons is projected to increase during the period 2051–2099 (Srivastava and Delsole, 2014).
54
55 Summer precipitation changes in South Asia are consistent between CMIP3 and CMIP5 projections, but the
Do Not Cite, Quote or Distribute Atlas-57 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 model spread is large for winter precipitation changes. Changes in summer monsoon rainfall will dominate
2 annual changes over South Asia (Woo et al., 2019). CMIP3 GCMs project a gradual increase in annual
3 precipitation over monsoon-dominated areas of Pakistan throughout the 21st Century and increases in humid
4 and semi-arid climate areas (Saeed and Athar, 2018).
5
6 Warming of 2.5°C to 5°C is projected over northern Pakistan and India (Syed et al., 2014). CORDEX-South
7 Asia projections over northeast India under RCP4.5 for the period 2011–2060, show increasing trends for
8 both seasonal maximum and minimum temperature over northeast India (Interactive Atlas). The future
9 projections of South Asian monsoon from the CORDEX-CORE exhibit a spatially robust delay in the
10 monsoon onset, an increase in seasonality, and a reduction in the rainy season length over parts of South
11 Asia at higher levels of radiative forcing (Ashfaq et al., 2020).
12
13 With the TIB continuing to warm, snow cover and snow water equivalent are projected to decrease but with
14 regional differences due to synoptic influences (Wester et al., 2019) and Cross-Chapter Box 10.4. There is
15 limited evidence on whether the ‘Karakoram Anomaly’ will persist in coming decades, but its long-term
16 persistence is unlikely with continued projected warming (high confidence) (Section 9.5.1.1). It is projected
17 that peak river flow at higher altitudes will commence earlier, due to warming influences on snow cover area
18 and snow/glacier melt rates and with more precipitation falling as rain rather than snow, and the magnitude
19 and seasonality of flow will change over South Asia (Charles et al., 2016).
20
21
22 Atlas.5.3.5 Summary
23
24 Mean, minimum and maximum daily temperatures in South Asia are increasing and winters are getting
25 warmer faster than summers (high confidence). The South Asian monsoon has shown contrasting behaviour
26 over India and Pakistan. There is high confidence that there has been a decrease in mean rainfall over most
27 parts of the eastern and central north regions of India and an increase in precipitation in Pakistan.
28
29 Global model performance over the region has improved from CMIP3 to CMIP5 to CMIP6 in the multi-
30 model ensemble mean simulation of the amplitude and phase of the seasonal cycles of temperature and
31 precipitation. However, there was no appreciable improvement in regions with steep orography, and there
32 has remained substantial inter-model spread in seasonal and annual mean temperatures over South Asia with
33 generally cold biases which are largest in the complex Karakorum and Himalayan mountain ranges. CMIP6
34 GCMs also show a dry bias (15–20%) in mean annual precipitation in the majority of South Asia region with
35 a wet bias in Nepal, Pakistan and northern India.
36
37 It is likely that surface temperatures over South Asia (SAS and TIB) will increase greater than the global
38 average, more so over TIB, and with projected increases of 4.6°C (3.4°C– 6.0°C) during 2081–2100
39 compared with 1995–2014 under SSP5-8.5 and 1.3°C (0.7°C– 2.0°C) under SSP1-2.6 (Interactive Atlas).
40 Summer monsoon precipitation in South Asia is likely to increase by the end of the 21st century while winter
41 monsoons are projected to be drier. Over the same time periods CMIP6 models project an increase in annual
42 precipitation in the range 14–36% under SSP5-8.5 and 0.4–16% under SSP1-2.6 (medium confidence).
43
44 With continued warming, TIB snow cover and snow water equivalent are likely to decrease and with more
45 precipitation falling as rain rather than snow in SAS. It is projected that the peak river flow at higher
46 altitudes will commence earlier due to the effect of warming on snow cover and snow/glacier melt rates,
47 causing changes in magnitude and seasonality of flow.
48
49
50 Atlas.5.4 Southeast Asia
51
52 Atlas.5.4.1 Key features of the regional climate and findings from previous IPCC assessments
53
54 Atlas.5.4.1.1 Key features of the regional climate
55 The Southeast Asia region is composed of countries that are part of Indochina (or mainland Southeast Asia)
Do Not Cite, Quote or Distribute Atlas-58 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 and countries that are very archipelagic in nature and have strong land-ocean-atmosphere interactions,
2 including those that are part of the Maritime Continent and the Philippines. Its climate is mainly tropical
3 (i.e., hot and humid with abundant rainfall). Rainfall seasonal variability in the region is mainly affected by
4 the synoptic-scale monsoon systems, the north-south migration of the ITCZ and tropical cyclones (mainly
5 for the Philippines and Indochina) while intraseasonal variability can be influenced by the MJO (Annex IV).
6 Temperature and especially rainfall are also interannually affected by ENSO and Indian Ocean Basin and
7 Dipole (IOB/IOD) modes (Annex IV, Table Atlas.1).
8
9
10 Atlas.5.4.1.2 Findings from previous IPCC assessments
11 WGI AR5 showed that the mean annual temperature of Southeast Asia has been increasing at a rate of
12 0.14°C to 0.20°C per decade since the 1960s, along with an increasing number of warm days and nights, and
13 a decreasing number of cold days and nights (Christensen et al., 2013). AR5 also reported the lack of
14 sufficient observational records to allow for a full understanding of past precipitation trends in most of the
15 Asian region, including Southeast Asia, and that precipitation trends that were available differed
16 considerably across the region and between seasons (Christensen et al., 2013).
17
18 On projected changes, findings from AR5 showed that warming is very likely to continue with substantial
19 subregional variations over Southeast Asia (Christensen et al., 2013). The median increase in temperature
20 over land projected by the CMIP5 ensemble mean ranges from 0.8°C in RCP2.6 to 3.2°C in RCP8.5 by the
21 end of the 21st century. Moderate future increases in precipitation are very likely, with projected ensemble
22 mean increases of 1% in RCP2.6 to 8% in RCP8.5 by 2100. In the SR1.5, there is a projected increase in
23 flooding and runoff over Southeast Asia for a 1.5°C to 2°C global warming, and these will increase even
24 more for a greater than 2°C level of warming (Hoegh-Guldberg et al., 2018).
25
26
27 Atlas.5.4.2 Assessment and synthesis of observations, trends and attribution
28
29 Within the last decade, there has been an increasing number of studies on climatic trends over Southeast
30 Asia, carried out on a regional basis (Thirumalai et al., 2017; Cheong et al., 2018) or focused on specific
31 countries (Cinco et al., 2014; Villafuerte et al., 2014; Mayowa et al., 2015; Villafuerte and Matsumoto, 2015;
32 Guo et al., 2017a; Sa’adi et al., 2017; Supari et al., 2017; Tan et al., 2021). They document virtually certain
33 significant increases in mean as well as extreme temperature. The minimum temperature extremes very likely
34 warmed faster compared to the maximum temperature. Temperatures, including extremes, are strongly
35 influenced by ENSO in the region (Cinco et al., 2014; Thirumalai et al., 2017; Cheong et al., 2018). Over
36 much of the regions, extreme high temperature occurred mostly in April and almost all April extreme
37 temperatures occur in El Niño years (Thirumalai et al., 2017). In most of Southeast Asia (except for the
38 north-eastern areas), there was likely an increase in the number of warm nights with El Niño episodes within
39 the period 1972–2010 (Cheong et al., 2018).
40
41 Changes in mean precipitation are less spatially coherent over Southeast Asia. Over Thailand, the average
42 number of rain days has decreased by 1.3 to 5.9 days per decade while average daily rainfall intensity has
43 increased by 0.24–0.73 mm day–1 per decade (Limsakul and Singhruck, 2016). Precipitation is also affected
44 by ENSO events (Tangang et al., 2017; Supari et al., 2018). Over Southeast Asia, there has been a significant
45 increase in the amount of precipitation and its extremes with La Niña episodes in the past decades, especially
46 during the winter monsoon period (high confidence) (Villafuerte and Matsumoto, 2015; Limsakul and
47 Singhruck, 2016; Cheong et al., 2018).
48
49 Figure Atlas.11 shows trends in mean temperature and precipitation during 1960–2015 for two global
50 datasets, indicating a significant overall warming over Southeast Asia (high confidence), with higher rates of
51 warming in Malaysia, Indonesia, and the southern areas of mainland Southeast Asia (low confidence).
52 Annual mean precipitation trends (see also Interactive Atlas which includes the regional dataset Aphrodite,
53 see section Atlas.1.4.1) over the region are mostly not significant except for increases over parts of Malaysia,
54 Vietnam and southern Philippines (medium confidence).
55
Do Not Cite, Quote or Distribute Atlas-59 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 It is important to note that the availability, quality, and temporal and spatial density of observation data may
2 lead to uncertainties and varying results in Southeast Asia (Juneng et al., 2016). Some efforts have been
3 made to produce better observationally-based gridded datasets for the region (e.g., Nguyen-Xuan et al.,
4 2016; van den Besselaar et al., 2017; Yatagai et al., 2020).
5
6
7 Atlas.5.4.3 Assessment of model performance
8
9 Performance in simulating rainfall over Southeast Asia varies among CMIP5 GCMs (high confidence). Only
10 some are capable of reasonably simulating the rainfall seasonal cycle and spatial pattern (Siew et al., 2013;
11 Raghavan et al., 2018). Over mainland Southeast Asia, the performance of CMIP5 GCMs in simulating
12 rainfall during wet season was superior to that for annual and dry season precipitation (Li et al., 2019a).
13
14 RCMs have been intensively used over the region in recent years in a series of single or multi-model
15 experiments and there is medium confidence that they reproduce reasonably well seasonal climate patterns of
16 temperature, precipitation and large-scale circulation over the different subregions of Southeast Asia with
17 added values compared to their host GCMs (Van Khiem et al., 2014; Kwan et al., 2014; Ngo-Duc et al.,
18 2014, 2017; Juneng et al., 2016; Katzfey et al., 2016; Loh et al., 2016; Raghavan et al., 2016; Cruz et al.,
19 2017; Ratna et al., 2017; Trinh-Tuan et al., 2018; Nguyen‐Thuy et al., 2021). RCM ensemble means tend to
20 outperform the individual models in representing the climatological mean state (Ngo-Duc et al., 2014; Trinh-
21 Tuan et al., 2018; Nguyen‐Thi et al., 2020). There is relatively high consistency among the simulations of
22 historical climate over mainland Southeast Asia compared to those over the Maritime Continent for both
23 seasonal and inter-annual variability (Ngo-Duc et al., 2017). The consistency in rainfall simulations was
24 lower than for temperature simulations.
25
26 Some RCMs showed a systematic cold bias (Manomaiphiboon et al., 2013; Kwan et al., 2014; Ngo-Duc et
27 al., 2014; Loh et al., 2016; Cruz and Sasaki, 2017; Cruz et al., 2017) that was mainly due to model physics
28 (Manomaiphiboon et al., 2013; Kwan et al., 2014) and/or the biases in the SST forcing (Ngo-Duc et al.,
29 2014). A few simulations revealed a warm bias over some areas such as in the Maritime Continent (Cruz et
30 al., 2017) or Vietnam (Van Khiem et al., 2014). The biases for rainfall in GCMs and RCMs over Southeast
31 Asia were found to be less systematic with wet or dry biases depending on the subregions (Manomaiphiboon
32 et al., 2013; Kwan et al., 2014; Van Khiem et al., 2014; Juneng et al., 2016; Nguyen‐Thi et al., 2020; Supari
33 et al., 2020; Tangang et al., 2020) although wet biases were more pronounced in RCMs (Kwan et al., 2014;
34 Van Khiem et al., 2014; Kirono et al., 2015; Juneng et al., 2016; Supari et al., 2020; Tangang et al., 2020).
35 Some RCMs overestimated rainfall interannual variability (Juneng et al., 2016) while some others
36 underestimated it (Kirono et al., 2015). Simulated rainfall amount is sensitive to the choice of convective
37 scheme (Juneng et al., 2016; Ngo-Duc et al., 2017) and the choice of land-surface scheme (Chung et al.,
38 2018). Rainfall biases in current climate simulations can be greatly reduced if a bias correction method such
39 as quantile mapping is applied (Trinh-Tuan et al., 2018). The pattern of tropical cyclone numbers in the
40 region were reasonable represented by RCM outputs (Van Khiem et al., 2014; Kieu-Thi et al., 2016;
41 Herrmann et al., 2020).
42
43
44 Atlas.5.4.4 Assessment and synthesis of projections
45
46 Mean temperature in Southeast Asia is projected to continue to rise through the 21st century (virtually
47 certain, very high confidence). Projections by multi-model regional climate simulations of CORDEX-SEA
48 showed a temperature increment over land under RCP8.5 to range from 3°C to 5°C by the end of 21st
49 century relative to pre-1986–2005 period (Tangang et al., 2018). For the same periods, the average mean
50 temperature increase over land projected by CMIP5 (CMIP6) varies from 0.9 ± 0.3°C (1.2 ± 0.4°C) under
51 RCP2.6 (SSP1-2.6) to 3.5 ± 0.7°C (3.8 ± 0.9°C) under RCP8.5 (SSP5-8.5) (Interactive Atlas). For all Global
52 Warming Levels (GWLs) the land region is projected to warm by a slightly smaller amount than the global
53 average, with 10th–90th percentile ranges for CMIP5 (CMIP6) of 1.2°C to 1.6°C (1.2°C to 1.5°C) for the
54 1.5°C GWL and of 3.3°C to 4.0°C (3.3°C to 3.9°C) for the 4°C GWL relative to the 1850–1900 baseline
55 (calculated from RCP8.5 (SSP5-8.5) projections). Changes for other warming levels, periods, and emissions
Do Not Cite, Quote or Distribute Atlas-60 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 pathways are shown in Figure Atlas.17 and can be explored in the Interactive Atlas.
2
3 Projections of future rainfall changes are highly variable among sub-regions of Southeast Asia and among
4 the models (high confidence). The CMIP5 and CMIP6 ensembles showed an increase in annual mean
5 precipitation over most land areas by the mid and late 21st century, although only with a strong model
6 agreement for higher warming levels (Interactive Atlas, Figure Atlas.17), while CORDEX produces a
7 general decrease in projected precipitation (Figure Atlas.17). Based on CORDEX Southeast Asia multi-
8 model simulations, significant and robust increases of mean rainfall over Indochina and the Philippines were
9 projected while there is a drying tendency over the Maritime Continent during DJF for the early, mid and
10 end of the 21st century periods under both RCP4.5 and RCP8.5 (Tangang et al., 2020) (Figure Atlas.19). At
11 the end of the 21st century during DJF and under RCP8.5, an increase of 20% in mean rainfall is projected
12 over Myanmar, northern central Thailand and northern Laos, and of 5–10% over the eastern Philippines and
13 northern Vietnam. During JJA, significant drier conditions are projected over almost the entire Southeast
14 Asia except over Myanmar and northern Borneo. Over the Indonesian region, especially Java, Sumatra and
15 Kalimantan, as much as a 20–30% decrease in mean rainfall is projected during JJA by the end of the 21st
16 century. The projected drier condition over Indonesia from CORDEX is consistent with that of (Kusunoki,
17 2017; Giorgi et al., 2019; Kang et al., 2019; Supari et al., 2020) and is associated with enhanced subsidence
18 over the region (Kang et al., 2019; Tangang et al., 2020).
19
20
21 [START FIGURE ATLAS.19 HERE]
22
23 Figure Atlas.19: The RCM projected changes in mean precipitation between the early (2011–2040), mid (2041–
24 2070) and late (2071–2099) 21st century and the historical period 1976–2005. Data are obtained
25 from the CORDEX-SEA downscaling simulations. Diagonal lines indicate areas with low model
26 agreement (less than 80%). Adapted from Tangang et al. (2020).
27
28 [END FIGURE ATLAS.19 HERE]
29
30
31 Atlas.5.4.5 Summary
32
33 It is virtually certain that annual mean temperature has been increasing in Southeast Asia in the past decades
34 while changes in annual mean precipitation are less spatially coherent though with some increasing trends
35 over parts of Malaysia, Vietnam and southern Philippines (medium confidence).
36
37 Although various biases still exist, there is high confidence that the models can reproduce seasonal climate
38 patterns well over the different subregions of Southeast Asia. There is medium confidence that the RCMs
39 show added value compared to their host GCMs over the region.
40
41 Projections show continued warming over Southeast Asia, but likely by a slightly smaller amount than the
42 global average. Projected changes in rainfall over Southeast Asia vary, depending on model, subregion and
43 season (high confidence) with consistent projections of increases in annual mean rainfall from CMIP5 and
44 CMIP6 over most land areas (medium confidence) and decreases in summer rainfall from CORDEX
45 projections over much of Indonesia (medium confidence).
46
47
48 Atlas.5.5 Southwest Asia
49
50 Atlas.5.5.1 Key features of the regional climate and findings from previous IPCC assessments
51
52 Atlas.5.5.1.1 Key features of the regional climate
53 Southwest Asia includes the Arabian Peninsula (ARP) and West Central Asia (WCA) reference regions
54 (Figure Atlas.17). ARP has a semi-arid or arid desert climate with very low annual mean precipitation and
55 very high temperature. Its temperature is influenced by SST variations over the tropical ocean (e.g., ENSO)
Do Not Cite, Quote or Distribute Atlas-61 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 and the NAO and AO (Attada et al., 2019) (see Annex IV for these and subsequent modes of variability).
2 Rainfall is influenced by the IOD and ENSO with more rainfall during El Niño (Kang et al., 2015; Kumar et
3 al., 2015; Abid et al., 2018; Kamil et al., 2019) and less during La Niña (Atif et al., 2020). The wet season in
4 ARP is mainly from November to April and the dry season is from June to August. Rainfall is confined
5 mostly to the south-western part of the Peninsula and contribution of extreme events to the total rainfall
6 varies within 20–70% from region to region and season to season (Almazroui, 2020b; Almazroui and Saeed,
7 2020). WCA is separated from Eastern Europe by the Caucasus Mountains, is adjacent to ARP, with South
8 Asia (SAS) to the south and West Siberia (WSB) to the north, and lies between the Mediterranean (MED),
9 Tibetan Plateau (TIB) and East Central Asia (ECA) regions. WCA is heterogeneous in terrain with the
10 Zagros Mountains and Iranian Plateau in the west and southwest, the Caspian Sea and lowland with deserts
11 in the north and northeast. The regional climate of WCA is influenced by the NAO and ENSO and it is
12 typically semi-arid or arid with a strong gradient in both precipitation and temperature from the mountains to
13 the plains and from north to south.
14
15
16 Atlas.5.5.1.2 Findings from previous IPCC assessments
17 The IPCC AR5 established it is very likely that temperatures will continue to increase over WCA in all
18 seasons whilst projections of decreased annual mean precipitation had medium confidence due to medium
19 agreement resulting from model-dependent subregional and seasonal changes (Christensen et al., 2013).
20 AR5 also concluded that for a better understanding of the climate of the region, results of high-resolution
21 regional climate models also need to be assessed and CMIP5 models generally had difficulties simulating the
22 mean temperature and precipitation climatology for Southwest Asia. This is partly related to the poor spatial
23 resolution of the models not resolving the complex mountainous terrain and the influence of different drivers
24 of the European, Asian and African climates. However, observational data scarsity and issues related to the
25 comparison of observations with coarse-resolution models added to the uncertainty and remained poorly
26 analysed in peer-reviewed literature on climate model performance (Christensen et al., 2013).
27
28 SR1.5 stated that even for 1.5°C and 2°C of global warming, Southwest Asia is among the regions with the
29 strongest projected increase in hot extremes with more urban populations exposed to severe droughts in West
30 Asia, while an increase of heavy precipitation events is projected in mountainous regions of Central Asia
31 (Hoegh-Guldberg et al., 2018; IPCC, 2018c). Higher temperatures with less precipitation will likely result in
32 higher risks of desertification, wildfire and dust storms exacerbated by land-use and land-cover changes in
33 the region with consequent effects on human health. Further drying of the Aral Sea in Central Asia will likely
34 have negative effects on the regional microclimate adding to the growing wind erosion in adjacent deltaic
35 areas and deserts that is already resulting in a reduction of the vegetation productivity including croplands.
36 There is also a projected increase of precipitation intensity in the Arabian Peninsula which is likely to lead to
37 higher soil erosion particularly in winter and spring due to floods (Mirzabaev et al., 2019). WCA includes
38 high mountains with enhanced warming above 500 m where, regardless of the emissions scenario, decreases
39 in snow cover are projected due to increased winter snowmelt and more precipitation falling as rain (high
40 confidence). A very strong interannual and decadal variability, as well as scarce in situ records for mountain
41 snow cover, have prevented a quantification of recent trends in High Mountain Asia (Hock et al., 2019b).
42
43
44 Atlas.5.5.2 Assessment and synthesis of observations, trends and attribution
45
46 Since the AR5, there has been an increasing number of studies on past climate change in Southwest Asia
47 though meteorological stations are sparsely scattered in the region. They are mainly located in the plains
48 below 2 km of altitude, very scarce in mountainous areas and have declined in number in WCA since the end
49 of the Soviet Union in 1991. This increases the uncertainty in both temperature and precipitation trends
50 particularly for elevated areas (Christensen et al., 2013; Huang et al., 2014) (high confidence). So researchers
51 use other sources of climate data in the region, particularly freely available gridded data (see Annex I).
52
53 Globally drylands showed an enhanced warming over the past century of 1.2°C to 1.3°C, significantly higher
54 than the warming over humid lands (0.8°C to 1.0°C) (Huang et al., 2017b). A strong increase in annual
55 surface air temperature of 0.27°C to 0.47°C per decade has been found over WCA between 1960 and 2013
Do Not Cite, Quote or Distribute Atlas-62 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 (very high confidence) (Han and Yang, 2013; Li et al., 2013; Hu et al., 2014, 2017; Huang et al., 2014; Deng
2 and Chen, 2017; Zhang et al., 2019a, 2017; Guo et al., 2018b; Haag et al., 2019; Yu et al., 2019). Warming is
3 most prominent in the spring based on the CRU dataset with rates likely ranging from 0.64°C to 0.82°C per
4 decade (Hu et al., 2014). Analysis of seasonal temperature trends based on high-resolution 1 km x 1 km
5 downscaled dataset CHELSA and 20 stations in Uzbekistan has confirmed the maximum significant trend in
6 temperature of from 0.6°C up to 1°C per decade in spring from 1979 to 2013 and no significant trend in
7 winter (Khaydarov and Gerlitz, 2019). There is very high confidence (robust evidence, high agreement) that
8 the shrinking of the Aral Sea has induced an increase in surface air temperature around the Aral Sea region
9 in the range of 2°C to 6°C (Baidya Roy et al., 2014; McDermid and Winter, 2017; Sharma et al., 2018). The
10 plateau of Iran has experienced significant increases in the average monthly values of daily maximum and
11 minimum temperatures with spatially varying rates of 0.1°C–0.3°C up to 0.3°C–0.4°C per decade and
12 greater spatial variation in minimum temperatures (high confidence) (Mahmoudi et al., 2019; Fathian et al.,
13 2020; Sharafi and Mir Karim, 2020).
14
15 Observed warming over northern ARP is higher than over the south, where minimum temperatures are
16 increasing faster than maximum temperatures (Almazroui, 2020a). The rate of mean temperature increase is
17 estimated at 0.10°C per decade over 1901–2010 (Attada et al., 2019), while it has reached 0.63°C (likely in
18 the range of 0.24°C to 0.81°C) per decade for the more recent period of 1978–2019 (Almazroui, 2020a).
19
20 An overall increasing trend of annual precipitation (0.66 mm per decade) was found over Central Asia based
21 on GPCC V7 data for the period 1901–2013 (Hu et al., 2017), but annual trends were found not significant
22 over the shorter period 1960–2013 (see also Figure Atlas.11 and Interactive Atlas). Winter precipitation saw
23 a significant increase of 1.1 mm per decade (Song and Bai, 2016). These estimates have low to medium
24 confidence since the satellite precipitation products have large systematic and random errors in mountainous
25 regions. Moreover CMORPH and TRMM products fail to capture the precipitation events in the ice/snow-
26 covered regions in winter and show a substantial false-alarm percentage in summer, but the gauge-corrected
27 GSMAP performs better than other products (Song and Bai, 2016; Guo et al., 2017b; Hu et al., 2017; Chen et
28 al., 2019b). Over the elevated part of eastern WCA precipitation increases in the range of 1.3–4.8 mm per
29 decade during 1960–2013 were observed (very high confidence) (Han and Yang, 2013; Li et al., 2013; Hu et
30 al., 2014, 2017; Huang et al., 2014; Deng and Chen, 2017; Zhang et al., 2019a, 2017; Guo et al., 2018b;
31 Haag et al., 2019; Yu et al., 2019). Reductions in spring precipitation and increases in winter have been
32 reported for Uzbekistan over the period 1979–2013 based on station data but these are not significant
33 (Khaydarov and Gerlitz, 2019). There is very low confidence of impact of the Aral Sea shrinking on
34 precipitation (Chen et al., 2011; Jin et al., 2017).
35
36 A decreasing trend of precipitation is reported for ARP with the mean value of –6.3 mm per decade (range of
37 –30 mm to 16 mm) for the period 1978–2019 (low confidence) with large interannual variability over Saudi
38 Arabia, which covers 80% of the region (AlSarmi and Washington, 2011; Almazroui et al., 2012; Donat et
39 al., 2014). The same decreasing trend in precipitation totals and an increasing trend in the number of
40 consecutive dry days are found for most of the Iranian plateau (medium confidence) (Rahimi and Fatemi,
41 2019; Fathian et al., 2020; Sharafi and Mir Karim, 2020). January-to-March mean snow cover and depth
42 over mountainous areas decreased between 2000 and 2019 (low to medium confidence due to limited
43 evidence) (Safarianzengir et al., 2020).
44
45
46 Atlas.5.5.3 Assessment of model performance
47
48 There is limited evidence about the performance of GCMs and RCMs in representing the current climate of
49 Southwest Asia due to very few studies evaluating models over this region, but literature is now emerging
50 particularly on CMIP5/CMIP6 and CORDEX simulations.
51
52 Over ARP, surface temperature biases for 18 of 30 CMIP5 models are within one standard deviation of the
53 observed variability (Almazroui et al., 2017). A warm bias in summer and a cold bias for others months
54 along with an underestimation of wet season precipitation and an overestimation in the dry season have been
55 reported in 26 CMIP5 models (Lelieveld et al., 2016). 30 CMIP6 GCMs have limited skill in simulating
Do Not Cite, Quote or Distribute Atlas-63 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 annual precipitation patterns, annual cycle statistics and long-term precipitation trends over Central Asia
2 partially due to considerable wet biases of up to 100% in the Southern Xinjiang and Hexi Corridor regions
3 (Guo et al., 2021). Also, CMIP6 models display a wide range of performance in reproducing ENSO
4 teleconnections that influences the region (Barlow et al., 2021).
5
6 RCM simulations using the CORDEX-MENA domain reproduce the main features of the mean surface
7 climatology over ARP with moderate biases (high confidence). RegCM4 driven by five GCMs (HadGEM2,
8 GFDL, CNRM, CanESM2, and ECHAM6) showed an ensemble mean cold bias of about –0.7°C and dry
9 bias of –13% over ARP (Almazroui, 2016) with a cold (warm) bias over western (south-eastern) areas (Syed
10 et al., 2019). Temperature biases in 30-year historical simulations with WRF using three different radiation
11 parameterizations were within ±2°C and mostly caused by surface long-wave radiation errors which affected
12 night-time minimum temperatures over 70% of the domain (Zittis and Hadjinicolaou, 2017). Mean absolute
13 errors in COSMO-CLM driven by ERA-Interim were about 1.2°C for temperature, 15 mm per month for
14 precipitation and 9% for total cloud cover, and with new parameterizations of albedo and aerosols optimized
15 for the region the RCM simulated the main climate features of this very complex area (Bucchignani et al.,
16 2016). RegCM4.4 also simulated the main features of the observed climatology (especially for dry regions)
17 with temperature biases within ±3.0°C. Annual precipitation was overestimated with winter and spring
18 underestimated (Ozturk et al., 2018).
19
20 Four RCMs (REMO, RegCM4.3.5, ALARO-0, and COSMO-CLM5.0) driven by ERA-Interim, NCEP2
21 reanalyses and two different GCMs reproduced reasonably well the spatio-temporal patterns for temperature
22 and precipitation though underestimated diurnal temperature range and had cold biases over mountainous
23 and high plateau regions in all seasons. There is low confidence in this result because of low station density
24 and a lack of high-elevation stations and with biases dependent on the choice of the observational dataset.
25 However, the performance of both GCMs and RCMs is better than reanalyses when compared to available
26 observations (Mannig et al., 2013; Ozturk et al., 2017; Russo et al., 2019; Top et al., 2021).
27
28
29 Atlas.5.5.4 Assessment and synthesis of projections
30
31 Temperature and precipitation projections from CMIP5/CMIP6 and CORDEX for different global warming
32 levels, SSP and RCP scenarios, time periods and baselines are shown in Figure Atlas.17 and further details
33 can be explored in the Interactive Atlas.
34
35 In WCA, projections for different GWLs are consistent not only in annual and seasonal warming but in
36 ranges of the projections. Under RCP8.5, annual mean temperature will likely exceed 2°C by mid-century
37 (compared with 1995–2014) and reach up to 4.8°C–6°C by the end of the century (Yang et al., 2017) with
38 faster warming projected by the CMIP6 ensemble under SSP5-8.5. In individual county-level studies on
39 GCM future climate projections temperatures increased by up to 7°C by the end of the century, depending on
40 season and emission scenario (Allaberdiyev, 2010; MENRPG, 2015; Vermishev, 2015; Gevorgyan et al.,
41 2016; Osborn et al., 2016; Aalto et al., 2017; IDOE, 2017; Salman et al., 2017). Statistical downscaling of 18
42 CMIP5 GCMs projected an annual temperature increase of 0.37°C per decade (under RCP4.5) with the
43 maximum in northern WCA and warming most conspicuous in summer (Luo et al., 2019). RCM
44 downscaling of GCMs over Central Asia projected a larger increase of temperature under RCP8.5 for the
45 2071–2100 period, ranging from 5°C to 8°C (Ozturk et al., 2017).
46
47 In ARP, the projected change in ensemble mean annual temperature from 30 CMIP6 models is from 1.6°C
48 (SSP1-2.6) to 5.3°C (SSP5-8.5) by 2070–2099 compared to 1981–2010 (Almazroui et al., 2020a). The
49 projected warming is the highest in the north, reaching 5.9°C and lowest in the south (4.7°C). COSMO-CLM
50 projections over the CORDEX-MENA domain show for ARP and WCA a strong warming with marked
51 seasonality for the end of the 21st century, ranging from 2.5°C in winter under RCP4.5 to 8°C in summer
52 under RCP8.5 and with large increases found over high-altitude areas in winter and spring (Bucchignani et
53 al., 2018; Ozturk et al., 2018). The CMIP5 multi-model mean warming in boreal summer in 2070–2099,
54 compared with 1951–1980, is projected to be about 2.5°C and 6.5°C at the 2°C and 4°C global warming
55 levels respectively (Huang et al., 2014).
Do Not Cite, Quote or Distribute Atlas-64 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Future projections of precipitation in Southwest Asia have large uncertainties and thus low confidence. There
2 are few significant changes, little consensus on the sign and with a tendency for reduction in CMIP5 being
3 reversed in CMIP6 across all warming levels (Ozturk et al., 2018). Statistical downscaling of 18 CMIP5
4 GCMs under RCP4.5 projected an increase in precipitation of 4.6 mm per decade in Southwest Asia during
5 2021–2060 relative to 1965–2004 (Luo et al., 2019). CMIP5 simulations project a general decrease in
6 precipitation over lowlands in Turkey, Iran, Afghanistan and Pakistan (Ozturk et al., 2017) and increase over
7 high-mountain regions (Aalto et al., 2017; Salman et al., 2018). At a 4°C global warming level, the multi-
8 model mean annual precipitation for Turkmenistan and parts of Tajikistan and Uzbekistan is projected to
9 decrease by 20%, with somewhat stronger relative decreases in summer (Reyer et al., 2017). Over northern
10 WCA, the CMIP5 ensemble mean projects increases of over 3 mm per decade under RCP2.6 and over 6 mm
11 per decade under RCP4.5 and RCP8.5 over the 21st century (Huang et al., 2014). Mean annual precipitation
12 is projected to rise by 5.2% at the end of the 21st century (2070–2099) under RCP8.5, compared to 1976–
13 2005, while mean annual snowfall is projected to decrease by 26.5% in Central Asia (Yang et al., 2017).
14 However, regardless of the sign of the precipitation change in the high-mountain regions of Central Asia, the
15 influence of the warming on the snowpack will very likely cause important changes in the timing and amount
16 of the spring melt (Diffenbaugh et al., 2013).
17
18 In ARP, the projected change in ensemble mean annual precipitation from 30 CMIP6 models ranges from
19 3.8% (–2.6% to 28.8%) to 31.8% (12.0% to 106.5%) under SSP1-2.6 and SSP5-8.5 emissions for the period
20 2080–2100 compared with 1995–2014 (Almazroui et al., 2020a). Northwest ARP precipitation is projected
21 to decrease between –6% to –27% per decade and in the south precipitation to increase by up to 8.6% per
22 decade. CMIP6 projections are in line with those from CMIP3 and CMIP5, however they are less variable in
23 the central area in CMIP6. The uncertainty associated with precipitation over ARP is large because of very
24 low annual amounts and high variability.
25
26
27 Atlas.5.5.5 Summary
28
29 Increase in annual surface air temperature over Southwest Asia are very likely in the range of 0.24°C to
30 0.81°C per decade over the last 50–60 years. Annual precipitation change over ARP since 1970 is estimated
31 at –6.3 mm per decade (and in the range of –30 to 16 mm per decade) and over WCA is generally not
32 significant except over the elevated part of eastern WCA where increases between 1.3 mm and 4.8 mm per
33 decade during 1960–2013 have been observed (very high confidence). In mountainous areas, the scarcity and
34 decline of the number of observation sites since the end of the former Soviet Union from 1991 increase the
35 uncertainty of the long-term temperature and precipitation estimates (high confidence).
36
37 Mean temperature biases in RCMs are within ±3°C in Southwest Asia, and annual precipitation biases are
38 positive in almost all parts of the region except over the ARP where they are negative in the wet season
39 (November to April) and over WCA in winter and spring (from December to May) (medium confidence).
40 Since regional model evaluation literature has only recently emerged there is medium evidence about the
41 performance of RCMs in Southwest Asia though with medium to high agreement on mean temperature and
42 precipitation biases. RCMs simulate colder temperatures than observed over mountainous and high plateau
43 regions (limited evidence, high agreement).
44
45 Further warming over Southwest Asia is projected in the 21st century to be greater than the global average,
46 with rates varying from 0.25°C to 0.8°C per decade depending on the season and scenario and maximum
47 rates found in the northern part of the region in summer (high confidence). The influence of the warming on
48 the snowpack will very likely cause changes in the timing and amount of the spring melt. CMIP6 projected
49 changes in annual precipitation totals are in the range of –3% to 29% (SSP1-2.6) and 12% to 107% (SSP5-
50 8.5) in ARP (medium confidence). Strong spatio-temporal differences with overall precipitation decreases
51 projected in the central and northern parts of WCA in summer (JJA) with increases in winter (DJF) (medium
52 confidence).
53
54
55
Do Not Cite, Quote or Distribute Atlas-65 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.6 Australasia
2
3 The assessment in this section focuses on changes in average temperature and precipitation (rainfall and
4 snow), including the most recent years of observations, updates to observed datasets, the consideration of
5 recent studies using CMIP5 and those using CMIP6 and CORDEX simulations. Assessment of changes in
6 extremes are in Chapter 11 (Table 11.10–12) and climatic impact-drivers in Chapter 12 (Table 12.5).
7
8
9 Atlas.6.1 Key features of the regional climate and findings from previous IPCC assessments
10
11 Atlas.6.1.1 Key features of the regional climate
12
13 Australasia is divided into five regions for the Atlas (Figure Atlas.21), as follows: New Zealand (NZ), with a
14 varied climate with diverse landscapes, mainly maritime temperate with four distinct seasons; Northern
15 Australia (NAU) which is mainly tropical with monsoonal summer-dominated rainfall (monsoon season
16 December to March, see Annex V), but with a hot, semi-arid climate in the south of the region; Central
17 Australia (CAU) with a predominantly hot, dry desert climate; Eastern Australia (EAU) with a temperate
18 oceanic climate at the coast to semi-arid inland; and Southern Australia (SAU) which ranges from
19 Mediterranean and semi-arid in the west to mainly cool temperate maritime climate in the southeast. Various
20 remote drivers have notable teleconnections to regions within Australasia, including an effect of the El Niño
21 Southern Oscillation and the Indian Ocean Dipole (Table Atlas.1, Annex IV). Much of southern NZ and
22 SAU are affected by systems within the westerly mid-latitude circulation, in turn affected by the Southern
23 Annular Mode. The monsoon and the Madden-Julian Oscillation affect rainfall variability in northern
24 Australia.
25
26
27 Atlas.6.1.2 Findings from previous IPCC assessments
28
29 The AR5 WGI and WGII reports (IPCC, 2013c; Stocker et al., 2013; Reisinger et al., 2014) give very high
30 confidence that air and sea temperatures in the region have warmed, cool extremes have become rarer in
31 Australia and New Zealand since 1950, while hot extremes have become more frequent and intense (e.g., it
32 is very likely that the number of warm days and nights have increased). The AR5 reported it is virtually
33 certain that mean air and sea temperatures will continue to increase, with very high confidence that the
34 greatest increase will be experienced by inland Australia and the smallest increase by coastal areas and New
35 Zealand. The AR5 reported a range of different precipitation trends within the region. For example, while
36 annual rainfall has been significantly increasing in north-western Australia since the 1950s (very high
37 confidence), it has been decreasing in the northeast of the South Island of New Zealand over 1950–2004
38 (very high confidence) and over southwest of Western Australia. In line with these trends, WGI reported it is
39 likely that drought has decreased in northwest Australia. Future projections for precipitation extremes
40 indicate an increase in most of Australia and New Zealand, in terms of rare daily rainfall extremes (i.e.,
41 current 20-year return period events) and of short duration (sub-daily) extremes (medium confidence).
42 Likewise, however, there is a projected increase the frequency of drought in southern Australia (medium
43 confidence) and in many parts of New Zealand (medium confidence). Owing to hotter and drier conditions
44 there is high confidence that the occurrence of fire weather will increase in most of southern Australia, and
45 medium confidence that the fire danger index will increase in many parts of New Zealand.
46
47 AR5 reported mean sea levels have also increased in Australia and New Zealand at average rates of relative
48 sea-level rise of 1.4 ± 0.6 mm yr–1 from 1900 to 2011, and 1.7 ± 0.1 mm yr–1 from 1900 to 2009, respectively
49 (very high confidence). The assessment found that the volume of ice in New Zealand has declined by 36–
50 61% from the mid-late 1800s to the late 1900s (high confidence), while late-season significant snow depth
51 has also declined in three out of four Snowy Mountain sites in Australia between 1957 and 2002 (high
52 confidence). As mean sea-level rise is projected to continue for at least several more centuries, there is very
53 high confidence that this will lead to large increases in the frequency of extreme sea-level events in Australia
54 and New Zealand. On the other hand, the volume of winter snow and the number of days with low-elevation
55 snow cover in New Zealand are projected to decrease in the future (very high confidence), while both snow
Do Not Cite, Quote or Distribute Atlas-66 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 depth and area are projected to decline in Australia (very high confidence).
2
3 SROCC (Hock et al., 2019b) reports on the observed and projected decline in snow cover in Australasia, as
4 well as the retreat of New Zealand glaciers following an advance in 1983–2008 due to enhanced snowfall. It
5 also reports on the vulnerability of some Australian communities and ecosystems to sea level rise, increases
6 in the intensity and duration of marine heatwaves driven by human influence (high confidence), the decrease
7 in frequency of tropical cyclones landfall on eastern Australia since the late 1800s (low confidence in an
8 anthropogenic signal), and presented a case study on the multiple hazards, compound risk and cascading
9 impacts from climate extremes in Tasmania in 2015/2016 (including an attributable human influence on
10 some events). SRCCL (Mirzabaev et al., 2019) found widespread vegetation ‘greening’ has occurred in parts
11 of Australia, and an increase in the desertification and drought risk in future in southern Australia.
12
13
14 Atlas.6.2 Assessment and synthesis of observations, trends and attribution
15
16 Reliable station observations are available from around 1900 in Australasia, but in some regions the
17 coverage was and remains poor. Australia and New Zealand have continued to warm, and many rainfall
18 trends have continued since the AR5. Changes and trends in temperature and precipitation from 1961 to
19 2015 from three different global data sets are displayed in Figure Atlas.11 and the Interactive Atlas and show
20 significant (at 0.1 significance level) warming trends over the southern and eastern Australia. Most of the
21 observed changes in precipitation over the region are not significant over this period. Although observed
22 datasets (e.g. GPCC and GPCP) generally agree on a significant drying trend in the southern regions of New
23 Zealand during the shorter 1980 to 2015 period, this is in fact the reverse of the longer-term trends in 1961 to
24 2015 (Interactive Atlas).
25
26 For a longer-term perspective based on high-quality regional datasets, Figure Atlas.20 shows Australasia has
27 warmed over the last century (very high confidence). Australian mean temperature has increased by 1.44 ±
28 0.24°C during the period 1910–2019 using the updated observed temperature dataset ACORN-SATv2.1,
29 with 2019 Australia’s hottest year on record and nine out of ten warmest years on record occurring since
30 2005 (Trewin et al., 2020). Much of the warming has occurred since 1960, there is clear anthropogenic
31 attribution of this change and emergence of the signal from the1850-1900 climate (BOM and CSIRO, 2020;
32 Hawkins et al., 2020).Warming has been more rapid than the national average in central and eastern
33 Australia, with a warming minimum and non-significant trends since the 1960s in the northwest (CSIRO and
34 BOM, 2015; BOM and CSIRO, 2020). The National Institute of Water and Atmospheric Research
35 temperature record, NIWA NZ, shows a warming of 1.13 ± 0.27°C during the period 1909–2019, although
36 several stations show non-significant trends since 1960 (Figure Atlas.20), including a warming minimum in
37 the southeast at least partly due to a persistent shift in atmospheric circulation (Sturman and Quénol, 2013;
38 MfE and Stats NZ, 2017, 2020).
39
40
41 [START FIGURE ATLAS.20 HERE]
42
43 Figure Atlas.20: Observed trends in mean annual temperature (a–b) and summer (DJF) and inter (JJA)
44 precipitation (c–d) for Australia and New Zealand from high-quality regional datasets. Time
45 series show anomalies from 1961–1990 average and 10-year running mean; maps show annual linear
46 trends for 1960–2019; rainfall trends are shown in % per decade, crosses show areas and stations with
47 a lack of significant trend and regions of seasonally dry conditions (<0.25 mm day–1) are masked and
48 outlined in red. Datasets are Australian Climate Observation Reference Network – Surface Air
49 Temperature version 2.1 (ACORN-SATv2.1) for Australian temperature, the Australian Gridded
50 Climate Data (AGCD) for Australian rainfall (Evans et al., 2020a), and the 30-station high-quality
51 network for New Zealand temperature and rainfall. Further details on data sources and processing are
52 available in the chapter data table (Table Atlas.SM.15).
53
54 [END FIGURE ATLAS.20 HERE]
55
56
Do Not Cite, Quote or Distribute Atlas-67 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Since 1960, precipitation has increased in much of mainland Australia in austral summer and decreased in
2 many regions of southern and eastern Australia in austral winter (Figure Atlas.20). A detectable
3 anthropogenic signal of increases in precipitation in Australia has been reported particularly for north central
4 Australia and for a few regions along the south-central coast for the period 1901–2010 (Knutson and Zeng,
5 2018). Seasonally, there is a significant decline in winter rainfall in southwest Western Australia (Figure
6 Atlas.20), with an attributable human influence with high confidence (robust evidence and medium
7 agreement) (Delworth and Zeng, 2014, and others)(see Section 10.4). Rainfall trends in the southeast are not
8 significant since 1960 but have shown a notable reduction since the 1990s, and there is limited evidence for
9 the attribution of this change to human influence (e.g., Rauniyar and Power, 2020). In New Zealand between
10 1960 and 2019 in both summer and winter, rainfall increased in some stations in the South Island and
11 decreased at many stations in the North Island, however most station trends are not statistically significant
12 (Figure Atlas.20)(MfE and Stats NZ, 2020). In JJA, Milford Sound (increasing) and Whangaparaoa
13 (decreasing) trends are significant.
14
15 In Australia, there has been a decrease in snow depth and area since the late 1950s, especially in spring
16 (BOM and CSIRO, 2018). Based on a reconstructed snow cover record, the recent rapid decrease in the past
17 five decades has been shown to be larger by more than an order of magnitude than the maximum loss for any
18 5-decade period over the past 2000 years (McGowan et al., 2018). In New Zealand, from 1977 to 2018,
19 glacier ice volume decreased from 26.6 km3 to 17.9 km3 (a loss of 33%) (Salinger et al., 2019).
20
21
22 Atlas.6.3 Assessment of climate model performance
23
24 Most studies assessed in WGII AR5 were based on Coupled Model Inter-comparison Project Phase 3
25 (CMIP3) models and Special Report on Emission Scenarios (SRES) scenarios and CMIP5 models whenever
26 available. WGI AR5 reported that model biases in annual temperature and rainfall are similar to or lower
27 than other continental regions outside the tropics, with temperature biases generally <1°C in the multi-model
28 mean and <2°C in most models over Australia compared to reanalysis, and with a wet bias over the
29 Australian inland region but a dry bias near coasts and mountain regions of both Australia and New Zealand.
30
31 Early results from CMIP6 suggest incremental improvements compared to CMIP5 in the simulation of the
32 mean annual climatology of temperature and precipitation of the Indo-Pacific region surrounding
33 Australasia, the teleconnection between ENSO and IOD and Australian rainfall and other relevant climate
34 features (Grose et al., 2020). These assessments suggest that confidence in projections is similar to AR5 or
35 incrementally improved. The CORDEX Australasia simulations are found to have cold biases in daily
36 maximum temperature and an overestimation of precipitation but overall showed added value in the
37 simulation of the current climate (Di Virgilio et al., 2019; Evans et al., 2020b).
38
39
40 Atlas.6.4 Assessment and synthesis of projections
41
42 Similar to the global average (Chapter 4), mean temperature in Australasia is projected to continue to rise
43 through the 21st century at a magnitude proportional to the cumulative greenhouse-gas emissions (virtually
44 certain, very high confidence, robust evidence), CMIP5 and CMIP6 results are shown in Figure Atlas.21. A
45 higher end to the range of temperature projections is found in CMIP6 compared to CMIP5 (Grose et al.,
46 2020), produced by a group of models with high climate sensitivity (Forster et al., 2019), and this creates a
47 higher multi-model-mean change. For example, projections for Australasia including ocean between 1995–
48 2014 and 2081–2100 are 1.4°C (1.1°C to 1.8°C 10th–90th percentile range) in CMIP5 under RCP4.5, but
49 1.8°C (1.3°C to 2.5°C) in CMIP6 under SSP2-4.5.
50
51 Using warming levels, the results can be directly compared accounting for the different distribution of
52 climate sensitivities in the two ensembles. In this framework, Australasia (land only) is projected to warm by
53 a similar amount to the global average: 1.4°C to 1.8°C for the 1.5°C warming level, through to 3.9°C to
54 4.8°C for the 4°C warming level from the 1850–1900 baseline in CMIP6 using SSP5-8.5 (results using other
55 SSPs and from CMIP5 are similar). Projected warming is greater over land than ocean, greater in Australia
Do Not Cite, Quote or Distribute Atlas-68 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 than in New Zealand, and greater over inland Australia than coastal regions. Due to historical warming,
2 projected temperature change from the AR6 baseline of 1995–2014 is lower: 0.3°C to 1.0°C for the 1.5°C
3 warming level, through to 2.9°C to 4.0°C for the 4°C warming level. Changes for other warming levels,
4 subregions and emissions pathways are shown in Figure Atlas.21 and can be explored in the Interactive
5 Atlas. Regional modelling suggests projected temperature increase is higher in mountainous areas than
6 surrounding low-elevation areas in New Zealand and Australia (Olson et al., 2016; MfE, 2018).
7
8
9 [START FIGURE ATLAS.21 HERE]
10
11 Figure Atlas.21: Regional mean changes in annual mean surface air temperature and precipitation relative to the
12 1995–2014 baseline for the reference regions in Australasia (warming since the 1850–1900 pre-
13 industrial baseline is also provided as an offset). Bar plots in the left panel of each region triplet
14 show the median (dots) and 10th–90th percentile range (bars) across each model ensemble for annual
15 mean temperature changes for four datasets (CMIP5 in intermediate colours; subset of CMIP5 used to
16 drive CORDEX in light colours; CORDEX overlying the CMIP5 subset with dashed bars; and CMIP6
17 in solid colours); the first six groups of bars represent the regional warming over two time periods
18 (near-term 2021–2040 and long-term 2081–2100) for three scenarios (SSP1-2.6/RCP2.6, SSP2-
19 4.5/RCP4.5, and SSP5-8.5/RCP8.5), and the remaining bars correspond to four global warming levels
20 (GWL: 1.5°C, 2°C, 3°C, and 4°C). The scatter diagrams of temperature against precipitation changes
21 display the median (dots) and 10th–90th percentile ranges for the above four warming levels for
22 December-January-February (DJF; middle panel) and June-July-August (JJA; right panel),
23 respectively; for the CMIP5 subset only the percentile range of temperature is shown, and only for
24 3°C and 4°C GWLs. Changes are absolute for temperature (in °C) and relative (as %) for
25 precipitation. See Section Atlas.1.3 for more details on reference regions (Iturbide et al., 2020) and
26 Section Atlas.1.4 for details on model data selection and processing. The script used to generate this
27 figure is available online (Iturbide et al., 2021) and similar results can be generated in the Interactive
28 Atlas for flexibly defined seasonal periods. Further details on data sources and processing are
29 available in the chapter data table (Table Atlas.SM.15).
30
31 [END FIGURE ATLAS.21 HERE]
32
33
34 In line with recent trends, a significant reduction in annual mean rainfall in southwest Australia is projected,
35 with the greatest reduction in winter and spring (very likely, high confidence). There is more than 80% model
36 agreement for projected mean annual rainfall decrease in southwest Western Australia for both mid (2041–
37 2060) and far (2081–2100) future, and for all warming levels (Interactive Atlas). Rainfall decreases, mainly
38 in winter and spring, are also projected for other regions within southern Australia with only medium
39 confidence (medium evidence and medium agreement). Almost all models project continued drying in SAU
40 in winter (JJA) and also in spring (SON), but a few models show little change. CMIP5 and CMIP6 results
41 are similar or with a slightly narrower range in the latter (Figure Atlas.21). CORDEX produces a similar
42 range of change in winter rainfall change for SAU as a whole. Circulation change is the dominant driver of
43 these projected reductions, explaining the range of model results for southern Australia (CSIRO and BOM,
44 2015; Mindlin et al., 2020). Studies of winter rainfall change and circulation in southern Australia suggest
45 the wettest changes in winter rainfall change may possibly be rejected (Grose et al., 2017, 2019a).
46
47 The model mean projection of northern Australian wet season precipitation (a period including DJF) is for
48 little change under all SSPs and warming levels, with low confidence in the direction of change as the
49 projections include both large and significant decrease and increases (Figure Atlas.21, Interactive Atlas).
50 Evidence from warming patterns suggests a constraint on the dry end of projections (Brown et al., 2016), and
51 the CMIP6 ensemble suggests that the projection follows the zonally-averaged rainfall response in the
52 southern hemisphere rather than changes in the western Pacific (Narsey et al., 2020). There is also evidence
53 for a projected increase in rainfall variability in northern Australia in scales from days to decades (Brown et
54 al., 2017). Liu et al. (2018) find that under 1.5°C warming, central and northeast Australia is projected to
55 become wetter, however this projection has low confidence. There are similar projections from CMIP5 and
56 CMIP6 (Figure Atlas.21).
57
Do Not Cite, Quote or Distribute Atlas-69 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Projections for EAU vary by season, with moderate model agreement on a decrease in rainfall in winter and
2 spring, but with lower agreement in CMIP6 compared to CMIP5, and low model agreement on the direction
3 of change in summer (Figure Atlas.21). CAU shows a similar range of change as EAU, with low model
4 agreement on the direction of change in DJF, moderate agreement on direction of change in JJA, but
5 significant changes are projected by some models. Other seasonal and regional rainfall changes in Australia
6 are reviewed in Dey et al. (2019).
7
8 For the NZ reference region, precipitation is projected to increase in winter and annual rainfall, with some
9 differences in magnitude between CMIP5, CMIP6 and CORDEX (Figure Atlas.21). This projection of
10 rainfall increase is a function of changes in the southern extent of the region, and notable regional differences
11 are expected. Regional modelling suggests precipitation increases in the west and south of New Zealand and
12 decreases in the north and east (MfE, 2018), with medium confidence, and with notable differences by
13 season. Liu et al. (2018) project that the North Island will be drier, while the South Island will be wetter
14 under both 1.5°C and 2°C warming levels. The projected increase in precipitation in the far future (2081–
15 2100) for the southern regions of NZ has high agreement (Interactive Atlas). Other seasonal and regional
16 rainfall changes in Australia can be explored in the Interactive Atlas.
17
18 The CORDEX Australasia simulations produce some regional detail in projected precipitation change
19 associated with important features such as orography. Areas where there is coincident ‘added value’ in the
20 simulation of the current climate and ‘potential added value’ as new information in the projected climate
21 change signal (collectively termed ‘realised added value’) in Australia include the Australian Alps, Tasmania
22 and parts of northern Australia (Di Virgilio et al., 2020). There have been several studies of regional climate
23 change for New Zealand and states within Australia at fine resolution (5–12 km) that have produced
24 important insights. One is enhanced drying in cool seasons on the windward slopes of the southern
25 Australian Alps (decreases of 20–30% compared to 10–15% in the driving models), and conversely a chance
26 of enhanced rainfall increase on the peaks of mountains in summer (Grose et al., 2019b), with the summer
27 finding in line with those for the European Alps (Giorgi et al., 2016).
28
29 Under future warming, the snowpack in Australia is projected to decrease by approximately 15% and 60%
30 by 2030 and 2070 respectively under the SRES A2 scenario (Di Luca et al., 2018), while in New Zealand the
31 number of annual snow days is projected to decrease by 30 days or more by 2090 under RCP8.5 (MfE,
32 2018). New Zealand is also projected to lose up to 88 ± 5% of its glacier volume by the end of the 21st
33 century (Chinn et al., 2012; Hock et al., 2019a).
34
35
36 Atlas.6.5 Summary
37
38 There is very high confidence that the climate of Australia warmed by around 1.4°C and New Zealand by
39 around 1.1°C since reliable records began in 1910 and 1909 respectively, with human influence the dominant
40 driver. Warming is virtually certain to continue, with a magnitude roughly equal to the global average
41 temperature. A significant decrease in April to October rainfall in southwest Western Australia is observed
42 from 1910 to 2019, is attributable to human influence with high confidence and is very likely to continue in
43 future noting consistent projections in CMIP5 and CMIP6. Other observed and projected rainfall trends are
44 less significant or less certain. Model representation of the climatology of Australasian temperature and
45 rainfall has improved since AR5, through an incremental improvement between CMIP5 and CMIP6 and the
46 development of coordinated regional modelling through CORDEX-Australasia. Snow cover is likely to
47 decrease throughout the region at high altitudes in both Australia and New Zealand (high confidence).
48
49
50 Atlas.7 Central and South America
51
52 The assessment in this section focuses on changes in average surface temperature and precipitation (rainfall
53 and snow), including the most recent years of observations, updates to observed datasets, the consideration
54 of recent studies using CMIP5 and those using CMIP6 and CORDEX simulations. Assessment of changes in
55 extremes are in Chapter 11 (Table 11.13–15) and climatic impact-drivers in Chapter 12 (Table 12.6). It
Do Not Cite, Quote or Distribute Atlas-70 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 considers climate change over the regions show in Figure Atlas.22, extending to all territories from Mexico
2 to South America, including the Caribbean islands. This figure supports the assessment of regional mean
3 changes over the region which, due to the high climatological and geographical heterogeneity, has been split
4 into two subregions: Central America and the Caribbean, and South America.
5
6
7 [START FIGURE ATLAS.22 HERE]
8
9 Figure Atlas.22: Regional mean changes in annual mean surface air temperature and precipitation relative to the
10 1995–2014 baseline for the reference regions in Central America, the Caribbean and South
11 America (warming since the 1850–1900 pre-industrial baseline is also provided as an offset). Bar
12 plots in the left panel of each region triplet show the median (dots) and 10th–90th percentile range
13 (bars) across each model ensemble for annual mean temperature changes for four datasets (CMIP5 in
14 intermediate colours; subset of CMIP5 used to drive CORDEX in light colours; CORDEX overlying
15 the CMIP5 subset with dashed bars; and CMIP6 in solid colours); the first six groups of bars represent
16 the regional warming over two time periods (near-term 2021–2040 and long-term 2081–2100) for
17 three scenarios (SSP1-2.6/RCP2.6, SSP2-4.5/RCP4.5, and SSP5-8.5/RCP8.5), and the remaining bars
18 correspond to four global warming levels (GWL: 1.5°C, 2°C, 3°C, and 4°C). The scatter diagrams of
19 temperature against precipitation changes display the median (dots) and 10th–90th percentile ranges
20 for the above four warming levels for December-January-February (DJF; middle panel) and June-
21 July-August (JJA; right panel), respectively; for the CMIP5 subset only the percentile range of
22 temperature is shown, and only for 3°C and 4°C GWLs. Changes are absolute for temperature (in °C)
23 and relative (as %) for precipitation. See Section Atlas.1.3 for more details on reference regions
24 (Iturbide et al., 2020) and Section Atlas.1.4 for details on model data selection and processing. The
25 script used to generate this figure is available online (Iturbide et al., 2021) and similar results can be
26 generated in the Interactive Atlas for flexibly defined seasonal periods. Further details on data sources
27 and processing are available in the chapter data table (Table Atlas.SM.15).
28
29 [END FIGURE ATLAS.22 HERE]
30
31
32 Atlas.7.1 Central America and the Caribbean
33
34 Atlas.7.1.1 Key features of the regional climate and findings from previous IPCC assessments
35
36 Atlas.7.1.1.1 Key features of the regional climate
37 The Central America and Caribbean region is assessed considering three reference regions South Central
38 America (SCA), including the isthmus and the Yucatan peninsula; North Central America (NCA), including
39 Mexico (centre and north); and the Caribbean (CAR), including the Greater Antilles, the Lesser Antilles, the
40 Bahamas and other small islands (see Figure Atlas.22); NCA is also covered in Section Atlas.9 North
41 America.
42
43 Precipitation in most of SCA is characterized by two maxima in June and September, an extended dry season
44 from November to May, and a shorter relatively dry season between July and August known as the
45 midsummer drought (MSD) (Magaña et al., 1999; Perdigón-Morales et al., 2018) (see Chapter 10). To some
46 extent, precipitation seasonality is explained by the migration of the Intertropical Convergence Zone (ITCZ)
47 (Taylor and Alfaro, 2005). The climate of NCA is temperate to the north of the Tropic of Cancer, with
48 marked difference between winter and summer, modulated by the North American monsoon (NAmerM,
49 Section 8.3.2.4.4). CAR has two main seasons, characterized by differences in temperature and precipitation.
50 The wet or rainy season, with higher values of temperature and accumulated precipitation, occurs during the
51 boreal summer and part of spring and autumn (Gouirand et al., 2020). The MSD is also present in the Greater
52 Antilles and the Bahamas (Taylor and Alfaro, 2005), influenced by the oscillations of the North Atlantic
53 Subtropical High (NASH), interacting with the Pacific and Atlantic branches of the ITCZ and modulated by
54 the Atlantic Warm Pool and the Caribbean Low-Level Jet (CLLJ), while the Atlantic ITCZ is responsible for
55 the unimodal rainfall cycle of the central and southern Lesser Antilles (Martinez et al., 2019). The CLLJ is a
56 persistent climatological feature of the low-level circulation in the Central Caribbean, with a characteristic
57 semi-annual cycle with maxima in the summer (main) and winter (secondary) (Amador, 1998; Magaña et al.,
Do Not Cite, Quote or Distribute Atlas-71 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 1999; Whyte et al., 2008). Temporal variability is influenced by several large-scale atmospheric modes (see
2 Annex IV and Table Atlas.1). A significant positive correlation between precipitation rates in CAR and the
3 Atlantic Multidecadal Variability was found (Enfield et al., 2001). A similar result was found in southern
4 Mexico (north of SCA) in the MSD region (Méndez and Magaña, 2010; Cavazos et al., 2020) (see case-
5 study discussion in Section 10.4.2.3). On the other hand, ENSO favours wet conditions in NCA, but its effect
6 is modulated by Pacific Decadal Variability (Maldonado et al., 2016).
7
8 One of the most prominent features of the regional climate is the incidence of tropical cyclones (TCs), which
9 represent an important hazard for almost all the countries of the region between June and November. A
10 detailed assessment is given in Chapter 11.
11
12
13 Atlas.7.1.1.2 Findings from previous IPCC assessments
14 According to the AR5 (Christensen et al., 2013), significant positive trends of temperature have been
15 observed in Central America (high confidence), while significant precipitation trends are regionally
16 dependent, especially during the summer. In addition, changes in climate variability and in extreme events
17 have severely affected the region (medium confidence). A decrease in mean precipitation is projected in SCA
18 and NCA. El Niño and La Niña teleconnections are projected to move eastwards in the future (medium
19 confidence), while changes in their effects on other regions, including Central America and the Caribbean is
20 uncertain (medium confidence). There is medium confidence in projections showing an increase in seasonal
21 mean precipitation on the equatorial flank of the ITCZ affecting parts of Central America and the Caribbean.
22
23 In relation to the 1986–2005 baseline period, temperatures are very likely to increase by the end of the
24 century, even for the RCP2.6 scenario, with changes of more than 5°C in some regions for the RCP8.5
25 scenario. Precipitation change is projected to vary between +10% and –25% (medium confidence)
26 (Christensen et al., 2013). SR1.5 (Hoegh-Guldberg et al., 2018) states there is a high agreement and robust
27 evidence that at the 1.5°C global warming level the Caribbean region will experience a 0.5°C to 1.5°C
28 warming compared to the 1971–2000 baseline period, with greatest warming over larger land masses.
29
30
31 Atlas.7.1.2 Assessment and synthesis of observations, trends and attribution
32
33 Significant warming trends between 0.2°C and 0.3°C per decade have been observed in the three reference
34 regions of Central America in the last 30 years (Planos Gutiérrez et al., 2012; Jones et al., 2016c; Hidalgo et
35 al., 2017), with the largest increases in the North America monsoon region (high confidence) (Cavazos et al.,
36 2020) (see also Figure Atlas.11 and the Interactive Atlas). There is high confidence of increasing temperature
37 over parts of NCA, reaching 0.5°C per decade in Mexico and southern Baja California. with a lower rate
38 (0.2°C per decade) in the Yucatan Peninsula and the Guatemala Pacific coastal region (Cueto et al., 2010;
39 García Cueto et al., 2013; Martínez-Austria et al., 2016; Martínez-Austria and Bandala, 2017; Navarro-
40 Estupiñan et al., 2018; Cavazos et al., 2020) and CAR (McLean et al., 2015) over the last 30 to 40 years.
41 Cooling trends have been detected in limited areas of Honduras and northern Panama (Hidalgo et al., 2017).
42
43 Changes in mean precipitation rates are less consistent and long-term trends are generally weak. Different
44 databases show significant differences depending mainly on the type and resolution of data (Centella-Artola
45 et al., 2020). Small positive trends were observed in the total annual precipitation (Stephenson et al., 2014).
46 In SCA and CAR, trends in annual precipitation are generally non-significant, with the exception of small
47 significant positive trends for subregions or limited periods (Planos Gutiérrez et al., 2012; Hidalgo et al.,
48 2017) and the 1970–1999 trends in precipitation in SCA are generally non-significant (Jones et al., 2016a;
49 Hidalgo et al., 2017). Positive trends in the duration of the MSD have been found in this region over the past
50 four decades (Anderson et al., 2019) (low confidence). For CAR see also Section Atlas.10 Small Islands.
51
52
53 Atlas.7.1.3 Assessment of model performance
54
55 The ability of climate models to simulate the climate in this region has improved in many key aspects
Do Not Cite, Quote or Distribute Atlas-72 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 (Karmalkar et al., 2013; Fuentes-Franco et al., 2014, 2015, 2017; Vichot-Llano et al., 2014; Vichot-Llano
2 and Martínez-Castro, 2017; Martínez-Castro et al., 2018a). Particularly relevant for this region are increased
3 model resolution and a better representation of the land-surface processes (high confidence).
4
5 Regional climate models (RCMs) forced with reanalyses and atmosphere-only global climate models provide
6 simulations with a reasonably good performance over the core North American monsoon, mostly in NCA
7 (high confidence) (Bukovsky et al., 2013; Cerezo-Mota et al., 2015). RCMs also reproduce the seasonal
8 spatial patterns of temperature and the bimodal rainfall characteristics of the NCA, SCA and CAR (high
9 confidence) (Karmalkar et al., 2013; Centella-Artola et al., 2015; Martínez-Castro et al., 2018b; Cavazos et
10 al., 2020; Vichot-Llano et al., 2020) though in some subregions specific models overestimate and shift the
11 month of the maxima. RCM simulations in the region do not necessarily improve with the size of the
12 domain, as important features of the regional circulation and key rainfall climate features, such as the CLLJ
13 and MSD, are well represented for a variety of domains of different sizes (Centella-Artola et al., 2015; Cabos
14 et al., 2018; Martínez-Castro et al., 2018b; Cavazos et al., 2020; Vichot-Llano et al., 2020) (high
15 confidence).
16
17
18 Atlas.7.1.4 Assessment and synthesis of projections
19
20 Figure Atlas.22 and the Interactive Atlas synthesize regional mean changes in annual mean surface air
21 temperature and precipitation for the Central America reference regions for CMIP6, CMIP5 and CORDEX
22 for different warming levels and time periods. At the 1.5°C GWL, it is very likely that average annual
23 temperature in Central America over land surpasses 1.3°C (CAR), 1.7°C (NCA) and 1.6°C (SCA). For the
24 3°C GWL, the corresponding projected ensemble mean regional warming values are 2.7°C (CAR), 3.5°C
25 (NCA) and 3.1°C (SCA). CAR average annual warming is below the level of global warming, while the two
26 continental reference regions are close to the global warming level with CMIP6 and CMIP5 showing very
27 consistent results (Figure Atlas.22). However, when focusing in time slices instead of warming levels, the
28 CMIP6 projections show systematically higher median values than CMIP5. CORDEX results are also
29 consistent with the previous findings, though the subset of driving models spans a smaller range of
30 uncertainty, particularly over CAR. Results have also been reported for this region based on CMIP5, CMIP6
31 and downscaled simulations over the CORDEX CAM domain or similar smaller domains (Taylor et al.,
32 2013a; Nakaegawa et al., 2014; Imbach et al., 2018; Vichot-Llano et al., 2019; Almazroui et al., 2021).
33 Statistical downscaling methods have been also applied to CMIP5 projections to obtain bias-adjusted
34 regional projections (Colorado-Ruiz et al., 2018; Taylor et al., 2018; Vichot-Llano et al., 2019).
35
36 Global and regional models consistently project warming in the whole region for the end of the century,
37 under RCP4.5 and RCP8.5 for CMIP5 projections with greater warming for continental compared to insular
38 territories, likely reaching values between 2°C and 4°C (high confidence) (Campbell et al., 2011; Karmalkar
39 et al., 2011; Cavazos and Arriaga-Ramírez, 2012; Cantet et al., 2014; Chou et al., 2014; Coppola et al., 2014;
40 Hidalgo et al., 2017; Colorado-Ruiz et al., 2018; Imbach et al., 2018). The greatest warming of 5.8°C for the
41 end of the century was projected for northern Mexico under RCP8.5 (Colorado-Ruiz et al., 2018), using an
42 ensemble of CMIP5 GCMs (see also the Interactive Atlas).
43
44 Regarding precipitation, it is likely that the annual average precipitation changes for the 1.5°C GWL will be
45 in the ranges of –11% to 0% in CAR, from –12% to 0% in SCA, and from –10% to +3% in NCA (Interactive
46 Atlas). For the 3°C GWL, the corresponding annual average precipitation changes will be from –17% to –2%
47 in CAR, from –16% to +2% in NCA, and from –23% to 0% in SCA. A clear drying tendency is observed for
48 the 3°C GWL relative to the 1.5°C GWL. Maloney et al. (2014) examined 21st-century climate projections
49 of North American climate in CMIP5 models under RCP8.5, including Central America and the Caribbean.
50 Summertime drying was projected in CAR and SCA for most of the models, with good agreement. The
51 strongest drying is projected to occur during July and August which are the months when the MSD occurs in
52 many subregions (Figure Atlas.22 and the Interactive Atlas). Intensification of the MSD in SCA was also
53 projected by using the Rossby Centre Regional Climate Model (RCA4) (Corrales-Suastegui et al., 2020), but
54 with future decrease in area and frequency (see Cross-Chapter Box Atlas.2). They also found a projected
55 intensification of CLLJ and drying for the future time slice of 2071–2095, relative to their baseline of 1981–
Do Not Cite, Quote or Distribute Atlas-73 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 2005. Decreased precipitation was also projected for SCA (Imbach et al., 2018) with the 8-km resolution Eta
2 RCM during the rainy season, including an intensification of the MSD, although no significant change was
3 projected for the CLLJ.
4
5 Colorado-Ruiz et al. (2018) assessed an ensemble of 14 GCMs from CMIP5 for a 1971–2000 baseline
6 period, projecting precipitation decreases of between 5% and 10% by the end of the century for the RCP4.5
7 and RCP8.5 scenarios respectively. The greatest decrease in precipitation is projected during summer
8 reaching 13%, especially in southern Mexico, Central America and the Caribbean. Dynamically downscaled
9 simulations (Bukovsky et al., 2015) also projected a decrease of precipitation for the middle of the century
10 (2041–2069) relative to 1971–1999 for the north of Mexico, though despite good agreement amongst the
11 models, these results must be considered of low confidence, because of their poor simulation of important
12 monsoon physical processes. Vichot-Llano et al. (2021) used a multiparameter ensemble of RegCM4, driven
13 by the CMIP5 global model HagGEM2-ES projections to conclude that, relative to the 1975–2004 baseline,
14 in the near (2020–2049) and more prominently in the far (2070–2099) future, drier conditions will prevail at
15 over the eastern Caribbean. The projected future warming trend was statistically significant at the 95%
16 confidence level over CAR and SCA. Almazroui et al. (2021) used an ensemble of 31 CMIP6 models to
17 estimate climate change signals of temperature and precipitation in six reference regions in North, Central
18 America and the Caribbean, finding a decrease in precipitation (10–30 %) over Central America and the
19 Caribbean under three scenarios with regional and seasonal variations.
20
21 There is high agreement and high confidence in the projected decrease of precipitation by the end of the
22 century for most of the region particularly for annual and summer precipitation, but there is low confidence
23 on the magnitude of this decrease which varies between 5% and 50% for different projections and different
24 subregions (see extended information in the Interactive Atlas).
25
26 The status of climate extreme trends and projections for the region has been assessed in Chapter 11 and the
27 main findings are synthesized here. There is high confidence in the projections of significant heatwave
28 events at the end of the century in SCA (Angeles-Malaspina et al., 2018) and an increase in warm days and
29 warm nights over this region and CAR (Stennett-Brown et al., 2017). For CAR islands, using dynamically
30 downscaled CMIP3 models, Karmalkar et al. (2013) projected increase in drought severity at the end of the
31 century, mainly due to precipitation decrease during the early wet season. In SCA projections suggest an
32 increase in the MSD (Imbach et al., 2018) and increase in consecutive dry days (Chou et al., 2014),
33 consistent with the projections of Stennett-Brown et al. (2017).
34
35
36 Atlas.7.1.5 Summary
37
38 Significant warming trends between 0.2°C and 0.3°C per decade have been observed in the three reference
39 regions of Central America in the last 30 years, with the largest increases in the North America monsoon
40 region (high confidence). Changes in mean precipitation rates are less consistent and long-term trends are
41 generally weak. Small positive trends were observed in the total annual precipitation in part of the region.
42
43 Warming in the continental part of the region is projected to increase in the range of the mean global values
44 for GWL of 1.5°C and 3°C, but in the Caribbean regional warming will be lower. Precipitation is projected
45 to decrease with increasing GWLs especially for CAR and SCA.
46
47 Projected change in mean annual precipitation shows a large spatial variability across Central America and
48 the Caribbean. Under moderate future emissions overall negative but non-significant precipitation trends are
49 projected for the 21st century (low confidence). Under higher emissions scenarios and at higher GWLs,
50 average precipitation is likely to decrease in most of the region, particularly in the north-western and central
51 Caribbean and part of continental Central America, especially in SCA.
52
53
54
55
Do Not Cite, Quote or Distribute Atlas-74 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.7.2 South America
2
3 Atlas.7.2.1 Key features of the regional climate and findings from previous IPCC assessments
4
5 Atlas.7.2.1.1 Key features of the regional climate
6 Regional synthesis of observed and modelled climate in South America is challenging due to the latitudinal
7 extent of the continent, the Andes mountains, and local to regional climatic features, which are influenced by
8 multiple drivers. The main large-scale drivers include many modes of natural variability (Annex IV.2): the
9 interdecadal modes, Atlantic Multidecadal Variability (AMV) and Pacific Decadal Variability (PDV); the
10 interannual-to-annual modes, El Niño-Southern Oscillation (ENSO), the Indian Ocean Dipole (IOD), the
11 Southern Annular Mode (SAM) and the North Atlantic Oscillation (NAO); seasonal variability driven by the
12 meridional migration of the Intertropical Convergence Zone and the timing and intensity of the South
13 American Monsoon System (SAmerM, Section 8.3.2.4.5), the Madden-Julian Oscillation sub-seasonal mode
14 of natural variability (MJO) and the behaviour at finer scales of the tropical easterly waves.
15
16 The regional assessment in this section emphasizes the seven new South American reference regions (Figure
17 Atlas.22) (Iturbide et al., 2020) that have a largely consistent climate and response to climate change and can
18 be used for analysis and impact studies (Solman et al., 2008; Neukom et al., 2010; Barros et al., 2015; Nobre
19 et al., 2016). At the subregional scale, several phenomena drive climate variability. Brazil's northeast (North-
20 Eastern South America; NES) is the most densely populated dryland globally and recurrently affected by
21 climatic extremes. The climate variability, particularly the precipitation, is marked by strong interannual
22 variability related to ENSO, the ITCZ, and the North Tropical Atlantic Ocean SSTs (Marengo et al., 2018a).
23 Northern (NSA) and North-Western South America (NWS) are part of the Amazonia region. Its most
24 recognizable features are the high rainfall, high humidity, and high temperatures that prevail in the region.
25 Rainfall variability in these regions results from the interplay between regional atmospheric circulation, the
26 SSTs variations in both the Pacific and Atlantic ocean, among other regional-to-local interactions (Marengo
27 and Espinoza, 2016; Espinoza et al., 2020). The South America monsoon (SAM) region has distinct wet
28 (summer) and dry (winter) periods. Key drivers include the South Atlantic Convergence Zone (Marengo et
29 al., 2012), the Bolivian High the 40- to 60-day intraseasonal oscillation, and the forcing of the high Andes
30 Mountains to the west (Almeida et al., 2017). The geographic position of South-Western South America
31 (SWS) results in very specific climatic characteristics since SWS contains subtropical climates as well as
32 sub-Antarctic and Antarctic climates. The climate of SWS is driven by seasonal changes in the position of
33 subtropical high-pressure air masses in the South Atlantic and South Pacific ocean, the Southern Annular
34 Mode, the dynamics of the cold Humboldt ocean current, and the icy cold fronts and mid-latitude westerlies
35 (Valdés-Pineda et al., 2016). The densely populated, highly productive subregion of South-Eastern South
36 America (SES) has cool winters and hot summers typical of the temperate zone, and climatic conditions are
37 strongly tied to ENSO, whose influence is moderated by local air-sea thermodynamics in the South Atlantic
38 (Barreiro, 2010). Lastly, the climate of the southern tip of South America (SSA) is cold and dry, and is
39 influenced by the Southern Annular Mode, and the interaction between the wetter Pacific winds and the
40 Andean Cordillera (Aceituno, 1988; Silvestri and Vera, 2009).
41
42
43 Atlas.7.2.1.2 Findings from previous IPCC assessments
44 According to WGII AR5 Chapter 27 (Magrin et al., 2014), during the last decades of the 20th century,
45 observational studies identified significant trends in precipitation and temperature in South America (high
46 confidence). Increasing trends in annual rainfall in south-eastern South America contrast with decreasing
47 trends in central southern Chile and some regions of Brazil. Warming has been detected throughout South
48 America (near 0.7°C to 1°C in the 40 years since the mid-1970s), except for a cooling off the Chilean coast
49 of about –1°C over the same period.
50
51 WGI AR5 (Flato et al., 2013) noted that climate simulations from CMIP3 and CMIP5 models were able to
52 represent well the main climatological features, such as seasonal mean and annual cycle (high confidence),
53 although some biases remained over the Andes, Amazon basin and for the South America Monsoon. On the
54 other hand, climate models from CMIP5 showed better results when compared to CMIP3.
55
Do Not Cite, Quote or Distribute Atlas-75 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 The SR1.5 (Hoegh-Guldberg et al., 2018) assessed that a further increase of 0.5°C or 1°C is likely to have
2 detectable effects on mean temperature and precipitation in South America, particularly in tropical regions
3 (NWS, NAS, SAM and NES), as well as in SES given that changes in mean temperatures and precipitation
4 have already been attributed in the last decades for global warming of less than 1°C.
5
6
7 Atlas.7.2.2 Assessment and synthesis of observations, trends and attribution
8
9 Studies on climatic trends in South America indicate that mean temperature and extremely warm maximum
10 and minimum temperatures have shown an increasing trend (high confidence), particularly for a large region
11 in northern South America and the south-western Andes (NSA, SAM, NES, SWS and the north of SES)
12 (Skansi et al., 2013; de Barros Soares et al., 2017). Also, the trend of the difference between the annual mean
13 of the daily maximum temperature and the annual mean of the daily minimum temperature was positive – up
14 to 1°C per decade – over the extra-tropics with the maximum temperature generally increasing faster than the
15 minimum temperature, while a negative trend – up to –0.5°C per decade – was observed over the tropics.
16
17 Regionally, analyses of temperatures point to an increased warming trend (high confidence) over Amazonia
18 over the last 40 years, which reached approximately 0.6°C–0.7°C (Figure Atlas.11 and the Interactive Atlas)
19 and with stronger warming during the dry season and over the southeast. The analyses also showed that 2016
20 was the warmest year since at least 1950 (Marengo et al., 2018b). Andean temperatures showed significant
21 warming trends, especially at inland and higher-elevation sites, while trends are non-significant or negative
22 at coastal sites (Vuille et al., 2015; Burger et al., 2018; Vicente-Serrano et al., 2018; Pabón-Caicedo et al.,
23 2020) (high confidence). Over central Chile, positive trends are largely restricted to austral spring, summer
24 and autumn seasons for mean, maximum and minimum temperatures (Burger et al., 2018; Vicente-Serrano et
25 al., 2018). Over Peru trends of maximum air temperature were mainly amplified during the austral summer,
26 but trends of cold season minimum air temperature showed an opposite pattern, with the strongest warming
27 being recorded in the austral winter (Vicente-Serrano et al., 2018).
28
29 In general, the spatial patterns of observed trends in temperature are more consistent than for precipitation
30 across the whole South America (Interactive Atlas) (de Barros Soares et al., 2017) (medium confidence). In
31 southeast Brazil there is a region of highly significant decrease of rainfall in both wet and dry seasons
32 recorded in the period 1979–2011 (Rao et al., 2016) (Interactive Atlas). The most consistent evidence of
33 positive rainfall trend occurs in the southern part of the La Plata Basin (southern Brazil, Uruguay, and north-
34 eastern Argentina) (de Barros Soares et al., 2017) (high confidence). By contrast, there is high confidence
35 that annual rainfall has decreased over northeast Brazil during the last decades (Carvalho et al., 2020).
36 Contrary to temperature changes, trends in annual precipitation exhibit different signs across sectors in the
37 Andes. For instance, annual precipitation trends in north tropical (north of 8°S) and south tropical (8°S–
38 27°S) Andes do not show a homogeneous pattern. Over the subtropical Andes, central Chile shows a robust
39 signal of declining precipitation since 1970 (Pabón-Caicedo et al., 2020) (high confidence).
40
41 Observational studies show that the dry-season length over southern Amazonia has increased significantly
42 since 1979 (Fu et al., 2013; Alves, 2016) (high confidence). In the Peruvian Amazon-Andes basin, there is no
43 trend in mean rainfall during the period 1965–2007 (Lavado Casimiro et al., 2012) though statistically
44 significant decreases in total annual rainfall in the central and southern Peruvian Andes from 1966 to 2010
45 were found (Heidinger et al., 2018). Despite that, recent analyses of Amazon hydrological and precipitation
46 data suggest an intensification of the hydrological cycle over the past few decades (Gloor et al., 2015). In
47 general, these changes are attributed mainly to decadal climate fluctuations (high confidence), ENSO, the
48 Atlantic SST north-south gradient, feedbacks between fire and land-use change mainly across southern
49 south-eastern Amazon and changes in the frequency of organized deep convection (Fernandes et al., 2015a;
50 Sánchez et al., 2015; Tan et al., 2015).
51
52 Since AR5, there has been limited attribution literature in the South America. Recent publications based on
53 observational and modelling evidence assessed that anthropogenic forcing in CMIP5 models explains with
54 the overall warming (high confidence) over the entire South American continent, including the increase in
55 the frequency of extreme temperature events (Hannart et al., 2015) such as the Argentinian heatwave of
Do Not Cite, Quote or Distribute Atlas-76 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 December 2013 (Chapter 11). It has a detectable influence in explaining positive and negative precipitation
2 trends observed in regions such as SES and the Southern Andes (Vera and Díaz, 2015; de Barros Soares et
3 al., 2017; Boisier et al., 2018; de Abreu et al., 2019). Despite that, there is limited evidence that human-
4 induced greenhouse gas emissions had an influence on the 2014/2015 water shortage in Southeast Brazil
5 (Otto et al., 2015). Extreme event attribution on sub-continental scales is assessed in Chapter 11 and
6 continental-scale attribution in Chapter 3.
7
8 In summary, analyses of historical temperature time series point strongly to an increased warming trend
9 (high confidence) across many South American regions, except for a cooling off the Chilean coast. Annual
10 rainfall has increased over south-eastern South America and decreased in most tropical land regions,
11 particularly in central Chile (high confidence). The number and strength of extreme events, such as extreme
12 temperatures, droughts and floods, have already increased (medium confidence) (see Table 11.7).
13
14 It is noted that the major barrier to the study of climate change in many regions of South America is still the
15 absence or insufficiency of long time series of observational data (Carvalho, 2020; Condom et al., 2020).
16 Most national datasets were created in the 1970s and 1980s, preventing a more comprehensive long-term
17 trend analysis. To fulfil the users’ demand for climatological and meteorological data products covering the
18 whole region, several interpolation techniques have been used with reanalysis and gridded gauge-analysis
19 products to add the necessary spatial detail to the climate analyses over land and for climate variability and
20 trend studies but these are subject to uncertainties (Skansi et al., 2013; Rozante et al., 2020).
21
22
23 Atlas.7.2.3 Assessment of model performance
24
25 Since AR5 the number of publications on climate model performance and their projections in South America
26 has increased, particularly for regional climate modelling studies (Giorgi et al., 2009; Boulanger et al., 2016;
27 Ambrizzi et al., 2019) and the understanding of their strengths and weaknesses (high confidence).
28
29 Most global and regional climate models can simulate reasonably well the current climatological features of
30 South America, such as seasonal mean and annual cycles. However, significant biases persist mainly at
31 regional scales (high confidence) (Torres and Marengo, 2013; Blázquez and Nuñez, 2013b; Gulizia et al.,
32 2013; Joetzjer et al., 2013; Jones and Carvalho, 2013; Gulizia and Camilloni, 2015; Zazulie et al., 2017a;
33 Abadi et al., 2018; Barros and Doyle, 2018; Solman and Blázquez, 2019; Teichmann et al., 2020; Fan et al.,
34 2020; Rivera and Arnould, 2020). During the dry season, precipitation is underestimated in most models
35 over Amazonia (medium evidence and high agreement) (Torres and Marengo, 2013; Yin et al., 2013; Solman
36 and Blázquez, 2019). Over regions with complex orography, such as the tropical Andes of NWS, CMIP5
37 models tend to underestimate precipitation which is associated with the misrepresentation of the Pacific
38 ITCZ and the local low-level jets (Sierra et al., 2015, 2018), whereas over the subtropical central Andes in
39 SWS, the models are found to overestimate both mean temperature and precipitation values (limited evidence
40 and high agreement) (Zazulie et al., 2017b; Rivera and Arnould, 2020; Díaz et al., 2021). Most models show
41 a dry bias over SES (Díaz and Vera, 2017; Barros and Doyle, 2018; Solman and Blázquez, 2019; Díaz et al.,
42 2021) associated with an underestimation of the northern flow that brings water vapour into the region
43 (medium confidence) (Gulizia et al., 2013; Zazulie et al., 2017a; Barros and Doyle, 2018). The biases in
44 seasonal precipitation, annual precipitation and climate extremes over several regions of South America were
45 reduced, including the Amazon, central South America, Bolivia, eastern Argentina and Uruguay, in the
46 CMIP5 models when compared to those of CMIP3 (medium confidence) (Joetzjer et al., 2013; Gulizia and
47 Camilloni, 2015; Díaz and Vera, 2017). The evidence is still insufficient to determine whether CMIP6 biases
48 are reduced when compared with CMIP5 simulations regarding precipitation and its variability in South
49 America. The temperature and precipitation patterns of anomalies associated with ENSO in the tropical
50 South America (NWS, NSA and NES) are better captured by GCMs in tropical South America (NWS, NSA
51 and NES) than in extra-tropical South America (SES), particularly during austral summer and autumn
52 (limited evidence and high agreement) (Tedeschi and Collins, 2016; Perry et al., 2020).
53
54 Based on regional simulations, studies showed that some RCMs improve the quality of the simulated climate
55 when compared with the driving GCM (medium evidence and high agreement) (Llopart et al., 2014; Sánchez
Do Not Cite, Quote or Distribute Atlas-77 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 et al., 2015; Falco et al., 2019; Solman and Blázquez, 2019; Ciarlo et al., 2020; Teichmann et al., 2020).
2 Regional climate model (RCM) simulations over South America can reproduce the main features of
3 temperature and precipitation in terms of both spatial distributions (Solman et al., 2013; Falco et al., 2019)
4 and seasonal cycles over the different climate regimes, including the main SAmerM features (high
5 confidence) (Jacob et al., 2012; Solman, 2013; Llopart et al., 2014; Reboita et al., 2014a; de Jesus et al.,
6 2016; Lyra et al., 2018; Bozkurt et al., 2019; Ashfaq et al., 2020). However, RCMs showed systematic biases
7 such as precipitation overestimations and temperature underestimations along the Andes throughout the
8 (high confidence) although these biases may be artificially amplified by the lack of a dense observational
9 station network (Jacob et al., 2012; Solman et al., 2013; Bozkurt et al., 2019; Falco et al., 2019). RCMs
10 tended to show dry biases over the Amazon and the northern part of the continent (SAM, NSA) during DJF
11 and during the maximum precipitation associated with the intertropical convergence zone (ITCZ) over NSA
12 during JJA (medium evidence and high agreement) (Solman et al., 2013; Falco et al., 2019). Temperature
13 overestimation and precipitation underestimation over La Plata Basin (in SES) are also RCM common biases
14 with the warm bias amplified for austral summer and the dry bias amplified for the rainy season (high
15 confidence) (Solman et al., 2013; Reboita et al., 2014a; Solman, 2016; Falco et al., 2019). Despite their
16 relevance RCM simulations at very high resolution (less than 10 km) are still few in South America (high
17 confidence) and are mainly designed for specific regions or purposes (Lyra et al., 2018; Bozkurt et al., 2019;
18 Bettolli et al., 2021).
19
20 The evaluation of statistical downscaling models (ESD) in representing regional climate features in South
21 America has increased since the AR5, however there are still few ESD studies over the different subregions.
22 Precipitation simulations based on ESD models are able to reproduce mean precipitation over tropical and
23 subtropical South American regions, especially over maximum precipitation areas in western Colombia,
24 south-eastern Peru, central Bolivia, Chile and the La Plata basin (medium confidence) (Souvignet et al.,
25 2010; Mendes et al., 2014; Palomino-Lemus et al., 2015, 2017, 2018; Troin et al., 2016; Soares dos Santos et
26 al., 2016; Borges et al., 2017; Bettolli and Penalba, 2018; Araya-Osses et al., 2020; Bettolli et al., 2021).
27 Temperature simulations are fewer but show added value to GCM simulations (medium evidence and high
28 agreement) (Souvignet et al., 2010; Borges et al., 2017; Bettolli and Penalba, 2018; Araya-Osses et al.,
29 2020).
30
31 Overall, climate modelling has made some progress in the past decade but there is no model that performs
32 well in simulating all aspects of the present climate over South America (high confidence). The performance
33 of the models varies according to the region, time scale, and variables analysed (Abadi et al., 2018). There is
34 also a fairly narrow spread in the representation of temperature and precipitation over South America by the
35 CMIP5 GCMs and also the RCMs, with biases that can be associated with the parametrizations and schemes
36 of surface, boundary layer, microphysics and radiation used by the models. Finally, observational reference
37 datasets, such as reanalysis products, used in the calibration and validation of climate models also can be
38 quite uncertain and may explain part of the apparent biases present in climate models (high confidence).
39
40
41 Atlas.7.2.4 Assessment and synthesis of projections
42
43 It is very likely that annual mean temperature will increase over South America, with a wide range of
44 projected changes of 1.0°C to 6.0°C by the end of the 21st century (from RCP2.6/SSP1-2.6 to RCP8.5/SSP5-
45 8.5 emissions, Figure Atlas.22). Overall, GCMs project higher temperature change than RCMs in austral
46 summer and winter over all subregions and in winter mainly over the central part of the continent (Coppola
47 et al., 2014; Llopart et al., 2020; Teichmann et al., 2020) (see also the Interactive Atlas). The largest
48 warmings over the South American continent are projected for the Amazon basin (SAM and NSA) and the
49 central Andes range (southern SAM, northern SWS and south-eastern NWS), Figure Atlas.22, especially
50 during the dry and dry-to-wet transition seasons (austral winter and spring) (Blázquez and Nuñez, 2013a;
51 Coppola et al., 2014; Pabón-Caicedo et al., 2020; Teichmann et al., 2020) (high confidence).
52
53 Using warming levels (Figure Atlas.22), the temperature is projected to increase at or above the level of
54 global warming in all regions apart from SSA with additional warming (compared to a 1995–2014 baseline)
55 of over 4°C for the 4°C warming level in NSA and SAM. Changes for other warming levels, subregions and
Do Not Cite, Quote or Distribute Atlas-78 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 emissions pathways are shown in Figure Atlas.22 and can be explored with the Interactive Atlas.
2
3 In general, models show a wide regional range in the direction and the magnitude of mean precipitation
4 change in many South American regions, with large significant increases and decreases (Figure Atlas.22,
5 Interactive Atlas). In the medium and long term, under the high-emission scenario, the CMIP5 multi-model
6 ensemble projected an increase in precipitation (generally greater than 10%) in SES and NWS and a decrease
7 (less 10%) in NSA across seasons (high confidence, robust evidence) (Solman, 2013; Chou et al., 2014;
8 Coppola et al., 2014; Llopart et al., 2014, 2020, Reboita et al., 2014b, 2020; Sánchez et al., 2015; Menéndez
9 et al., 2016; Ruscica et al., 2016; Bozkurt et al., 2018a; Zaninelli et al., 2019). Also, in parts of SWS, annual
10 precipitation is projected to decrease (up 30%) by the late 21st century (Souvignet et al., 2010; Palomino-
11 Lemus et al., 2017, 2018; Bozkurt et al., 2018a). Under high RCPs, the CMIP5 ensemble projects that all
12 Brazilian regions will experience more rainfall variability in the future, so drier dry periods and wetter wet
13 periods on daily, weekly, monthly and seasonal timescales, despite the future changes in mean rainfall being
14 currently uncertain (medium confidence) (Alves et al., 2020). Regarding the SAmerM, it is very likely that
15 the monsoon will experience changes in its life cycle by the end of the 21st century for both RCP4.5 and
16 RCP8.5 emissions and, in particular, delayed onset. However there is low agreement on the projected
17 changes in terms of extreme and total precipitation of the monsoon season in South America (Llopart et al.,
18 2014; Ashfaq et al., 2020). Changes in the SAmerM are assessed in Section 8.3.2.4.5.
19
20 Projected changes in seasonal precipitation and their uncertainties generally agree with the annual changes,
21 particularly for the decreases in SWS (Figure Atlas.22). DJF precipitation changes in NSA and SAM are
22 largely uncertain, with weak agreements in the projections, particularly for CMIP5 and CMIP6 ensembles,
23 which project almost no change and decreasing precipitation for NSA and a narrow range from slight
24 increases to no change respectively for SAM.
25
26
27 Atlas.7.2.5 Summary
28
29 In summary, it is virtually certain that the climate of South America has warmed. Studies on climate trends
30 in South America indicate that mean temperature, maximum and minimum temperatures have increased over
31 the last 40 years. Long-term observed precipitation trends show an increase over south-eastern South
32 America and decreases in most tropical land regions (high confidence).
33
34 Evaluation of global and regional climate model simulations have increased over South America in the past
35 decade and shown improved performance. However, the results reveal that no model performs well in
36 simulating all aspects of the present climate (very likely). On the other hand, there is still a lack of high-
37 quality and high-resolution observational data that may explain part of the important biases present in
38 climate models (high confidence).
39
40 Climate model projections show a general increase in annual mean surface temperature over the coming
41 century for all emission scenarios (RCPs and SSPs) (high confidence) consistent with the observed warming
42 and with all regions except SSA warming faster than the global average. Unlike temperature, annual
43 precipitation has patterns of decrease in North-eastern South America (NES) and South-western South
44 America (SWS) and increase in southern South America (SES) and North-western South America (NWS)
45 (high confidence), with small changes projected under a low-emission scenario. However, there is low
46 confidence in the magnitude because of the large spread among models, both GCMs and RCMs.
47
48
49 Atlas.8 Europe
50
51 The assessment in this section focuses on changes in average temperature and precipitation (rainfall and
52 snow), including the most recent years of observations, updates to observed datasets, the consideration of
53 recent studies using CMIP5 and those using CMIP6 and CORDEX simulations. Assessment of changes in
54 extremes are in Chapter 11 (Table 11.16–11.18) and climatic impact-drivers in Chapter 12 (Table 12.7).
55
Do Not Cite, Quote or Distribute Atlas-79 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.8.1 Key features of the regional climate and findings from previous IPCC assessments
2
3 Atlas.8.1.1 Key features of the regional climate
4
5 Westerly winds and the accompanying Atlantic storm track with cyclones and anticyclones travelling from
6 the Atlantic towards inland Europe are the main climatic features that characterize daily to inter-annual
7 variability in the European region. The Siberian High in winter determines cold weather in East Europe and
8 can affect other regions with cold outbreaks. Intra-seasonal and inter-annual variations are driven by modes
9 of climate variability such as the North Atlantic Oscillation (see Table Atlas.1 and Annex IV.2). Global
10 warming can lead to systematic changes in regional climate variability via thermodynamic responses such as
11 altered lapse rates (Kröner et al., 2017; Brogli et al., 2019) and land-atmosphere feedbacks (Zampieri and
12 Lionello, 2011; Boé and Terray, 2014). Regional feedbacks involving the land-sea contrast, sea surface, land
13 surface, clouds, aerosols, radiation and other processes modulate the regional response to enhanced warming.
14
15 Four climatic regions are defined for Europe (see Figure Atlas.24). The Mediterranean region (MED) in the
16 south is characterized by mild winters and hot and dry summers (Mediterranean climate; see Section
17 10.6.4.2). It covers both Europe and Africa, and MED assessments in this section generally imply the entire
18 MED domain unless stated otherwise. The Western and Central Europe region (WCE) has distinct summer
19 and winter seasons with increasing continentality of climate eastwards. The northern region (NEU), close to
20 the Atlantic Ocean, is characterized by high humidity and relatively mild winters, and strong exposure to the
21 Atlantic storm track. Eastern Europe (EEU) covers the western part of Russia and neighbouring territories
22 and has continental characteristics. Many regional datasets and model projections assessed here do not
23 sufficiently cover the EEU region.
24
25
26 Atlas.8.1.2 Findings from previous IPCC assessments
27
28 AR5 WG II (Kovats et al., 2014) reports with high confidence that observed climate trends show regionally
29 varying changes in temperature and rainfall in Europe. The average temperature in Europe has continued to
30 increase, with seasonally different rates of warming being greatest in high latitudes in Northern Europe.
31 Annual precipitation has increased in Northern Europe and decreased in parts of Southern Europe. SROCC
32 (Hock et al., 2019b) reports with high confidence that a reduction in snow cover at low elevation and glacier
33 extent is observed in recent decades, with consequent changes in annual and seasonal runoff patterns.
34 According to the SRCCL report (IPCC, 2019b) there is high agreement that observed vegetation greening
35 and forestation in the last 30 years cools summertime surface temperature and warms winter temperature due
36 to decreased snow cover and increased snow shading in forested areas. It is very likely that aerosol column
37 amounts have declined over Europe since the mid-1980s.
38
39 AR5 (Collins et al., 2013) reports that the ability of models to simulate the climate in Europe has improved
40 in many important aspects. Particularly relevant for this region are increased model resolution and a better
41 representation of the land-surface processes in many of the models that participated in CMIP5. The spread in
42 climate model projections is still substantial, partly due to pronounced internal variability in this region
43 (particularly NAO and AMO). In the winter half year, NEU and WCE are likely to have increased mean
44 precipitation associated with increased atmospheric moisture and moisture convergence and intensification
45 in extratropical cyclone activity. No change or a moderate reduction is projected for MED. In the summer
46 half year, it is likely that NEU and WCE mean precipitation will have only small changes with a notable
47 reduction in MED. According to SR1.5 (Hoegh-Guldberg et al., 2018), these precipitation changes are more
48 pronounced at 2°C than at 1.5°C of global warming. For a 2°C global warming, an increase in runoff is
49 projected for north-eastern Europe while decreases are projected in the Mediterranean region, where runoff
50 differences between 1.5°C and 2°C global warming will be most prominent (medium confidence). According
51 to SROCC (Hock et al., 2019b) the RCP8.5 projections lead to a loss of more than 80% of the ice mass from
52 small glaciers by the end of century in Central Europe (high confidence). Snow cover and glaciers are
53 projected to decrease all along the 21st century.
54
55
Do Not Cite, Quote or Distribute Atlas-80 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.8.2 Assessment and synthesis of observations, trends and attribution
2
3 To support climatological analyses and model evaluation, national meteorological and hydrological services
4 are increasingly making available high spatial and temporal resolution gridded and in situ homogenized and
5 quality-checked datasets (Déqué and Somot, 2008; Vidal et al., 2010; Rauthe et al., 2013; Noël et al., 2015;
6 Spinoni et al., 2015b; Ruti et al., 2016; Fantini et al., 2018; Lussana et al., 2018; Herrera et al., 2019;
7 Skrynyk et al., 2020). The inclusion of additional station data and data rescue activities lead to a better
8 representation of extreme precipitation statistics than the global- or continental-scale datasets (see Section
9 Atlas.1.4.1). Recent gridded products merging radar and station data allow higher spatial and temporal
10 resolutions to be reached (Haiden et al., 2011; Tabary et al., 2011; Berg et al., 2016; Fumière et al., 2020). A
11 number of regional reanalysis products has become available for the European region (Bollmeyer et al.,
12 2015; Bach et al., 2016; Dahlgren et al., 2016; Landelius et al., 2016). A European ensemble of regional
13 reanalyses from 1961 to 2019 is shown to add accuracy and reliability in comparison to global reanalysis
14 products, but also introduces additional uncertainties, especially for threshold-based climate indices (Kaiser-
15 Weiss et al., 2019). However, gridded European datasets are unreliable over data-sparse regions. Also, many
16 datasets employ different approaches to interpolation and gridding, which adds to their uncertainty and
17 complicates comparative evaluations (Berthou et al., 2018; Fantini et al., 2018; Kotlarski et al., 2019). For
18 some subregions and performance metrics differences between datasets have been shown to be of the same
19 magnitude as errors in regional climate models (Prein et al., 2016; Prein and Gobiet, 2017; Fantini et al.,
20 2018), but observational uncertainty is substantially reduced when datasets of similar nature and
21 representativeness are used (Kotlarski et al., 2019).
22
23 In addition to the global display of observed temperature and precipitation trends in Figure Atlas.11, annual
24 mean temperature and precipitation trends between 1980 and 2015 calculated from the gridded ensemble E-
25 OBS dataset (Cornes et al., 2018) are shown in Figure Atlas.23, together with time series of temperature and
26 precipitation anomalies relative to the 1980–2015 mean value from E-OBS, CRU, EWEMBI and Berkeley
27 for temperature, and E-OBS, CRU, GPCC and GPCP for precipitation (see also Figure 2.11 for global mean
28 values, and Section Atlas.1.4.1 for description of global datasets).
29
30 In NEU continued warming has been observed, particularly during spring. An annual mean temperature
31 increase of 0.4°C per decade was reported between 1970 and 2008 (Rutgersson et al., 2015). In WCE
32 temperature increases since the mid 20th century have been documented for Poland (Degirmendžić et al.,
33 2004) and Ukraine (Boychenko et al., 2016; Balabukh and Malitskaya, 2017). Land-only observations
34 indicate a rapid increase in summer (JJA) mean surface air temperature since the mid-1990s (Dong et al.,
35 2017). In Eastern Europe no significant trend in winter mean air temperatures was found between 1881 and
36 2016 in Belarus (Loginov et al., 2018). In parts of the European area of the MED spring and summer
37 temperatures are reported to increase faster than in the other seasons (Brunetti et al., 2006; Homar et al.,
38 2009; Lionello et al., 2012; Philandras et al., 2015; Gonzalez-Hidalgo et al., 2016; Vicente-Serrano et al.,
39 2017) (see also the Mediterranean case study in Section 10.6.4 and Figure 10.18). Figure Atlas.23 shows that
40 since 1980 in each European region all datasets show a consistent warming of annual mean temperature of
41 0.04°C yr–1 to 0.05°C yr–1. Trends in European land temperature cannot be explained without accounting for
42 anthropogenic warming offset by anthropogenic aerosol emissions (see Section 8.3.1.1 and Figure 3.8). It is
43 virtually certain that annual mean temperature continues to increase in each European subdomain.
44
45 Multidecadal trends in mean precipitation are generally small and non-significant. Apart from difficulties
46 related to observational coverage (Prein and Gobiet, 2017), gauge undercatch (e.g., Murphy et al., 2019) and
47 data inhomogeneity (e.g., Camuffo et al., 2013), strong interannual and multidecadal variability is dominant
48 over at least the last two centuries. However, significant precipitation trends have been recorded for recent
49 periods, for example in south-western Europe between 1960 and 2000 (Peña-Angulo et al., 2020) and
50 between 1961 and 2015 in NEU (Interactive Atlas). Also, some studies suggest that in the MED precipitation
51 has declined and more frequent and severe meteorological droughts have occurred between 1960 and 2000
52 (Spinoni et al., 2015a; Gudmundsson and Seneviratne, 2016), and in some regions cannot be explained
53 without anthropogenic forcing (Knutson and Zeng, 2018) (see also Section 10.4.1.2). Other studies suggest
54 that this trend can be seen as an expression of multidecadal internal variability driven mainly by the North
55 Atlantic Oscillation (Table Atlas.1) (Kelley et al., 2012; Zittis, 2018). Global dimming and brightening also
Do Not Cite, Quote or Distribute Atlas-81 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 are reported to affect precipitation trends in the Mediterranean region (see Section 8.3.1.6 and Figure 8.7).
2
3 The large-scale spatial patterns of the E-OBS annual mean precipitation trend between 1980 and 2015 shown
4 in Figure Atlas.23 is broadly consistent with trends derived from CRU, GPCP and GPCC (Figure Atlas.11)
5 but with more explicit spatial detail. Trends calculated for regional averages are sensitive to the selection of
6 the time window: for 1980–2015 annual mean precipitation averaged over the regions shows a positive trend
7 (not significant at p = 0.05), while for CRU and GPCC the trend calculated over 1901–2015 is positive for
8 NEU, EEU and WCE, and non-significant for MED. Precipitation trends in the MED are significant only in
9 selected areas (Lionello et al., 2012; MedECC, 2020). Also the NEU trends show large spatial variability and
10 are subject to decadal variability related to NAO (Heikkilä and Sorteberg, 2012), but are generally positive
11 over the 20th century (Figure Atlas.23). There is medium confidence that annual mean precipitation in NEU,
12 WCE and EEU has increased since the early 20th century. In the European Mediterranean observed land
13 precipitation trends show pronounced variability within the region, with magnitude and sign of trend in the
14 past century depending on time period and exact study region (medium confidence).
15
16 Trends in snowfall and snow melt are related to seasonal changes in both temperature and precipitation. In
17 EEU melt onset dates have advanced by 1–2 weeks in the 1979–2012 period (Mioduszewski et al., 2015).
18 Over Eurasia, trends in spring and early summer snow cover extent increased over the 1971–2014 period
19 (Hernández-Henríquez et al., 2015). Between 1966 and 2012, averaged over entire Eurasia, monthly mean
20 snow depth decreased in autumn and increased in winter and spring (Zhong et al., 2018), while the snow
21 cover extent was reported to have decreased during the past 40 years (Bulygina et al., 2011). In NEU late
22 winter and early spring snow depth and snow cover decreases since the early 1960s are reported over Finland
23 (Luomaranta et al., 2019) and Norway (Rizzi et al., 2018) with a dependence on altitude (Skaugen et al.,
24 2012), while winter snow depth increased in northern Sweden (Kohler et al., 2006). It is very likely that since
25 the early 1980s in snow-dominant areas in NEU and EEU the length of the snowfall season is reduced with
26 regional warming, and the melt onset dates have advanced.
27
28
29 [START FIGURE ATLAS.23 HERE]
30
31 Figure Atlas.23: (a) Mean 1980–2015 trend of annual mean surface air temperature (°C per decade) from E-OBS
32 (Cornes et al., 2018). Data for non-European countries in the MED area are masked out. (b) Time
33 series of mean annual temperature anomaly relative to the 1980–2015 period (shown with grey
34 shading) aggregated for the land area in each of the four European subregions, from E-OBS, CRU,
35 Berkeley and ERA5 (see Section Atlas.1.4.1 for description of global datasets). Mean trends for
36 1901–2015, 1961–2015 and 1980–2015 are shown for each data set in corresponding colours in the
37 same units as panel (a) (see legend in upper panel). (c) As panel (a) for annual mean precipitation
38 (mm day–1 per decade). (d) as panel (b) for annual mean precipitation, and data sets E-OBS, CRU,
39 GPCC and GPCP. Note that E-OBS data are not shown in panels b and d for region EEU. For the
40 MED region data are aggregated over the European countries alone. Trends have been calculated
41 using ordinary least squares regression and the crosses indicate non-significant trend values (at the 0.1
42 level) following the method of Santer et al. (2008) to account for serial correlation. Further details on
43 data sources and processing are available in the chapter data table (Table Atlas.SM.15).
44
45 [END FIGURE ATLAS.23 HERE]
46
47
48 The increasing trend in surface shortwave radiation, documented in AR5 (Hartmann et al., 2013) to have
49 occurred since the 1980s and referred to as a brightening effect, is substantiated over Europe and the
50 Mediterranean region (Nabat et al., 2014; Sanchez‐Lorenzo et al., 2015; Cherif et al., 2020). This increasing
51 trend has been attributed to the decrease in anthropogenic sulphate aerosols over the 1980–2012 period
52 (Nabat et al., 2014). In model sensitivity experiments, the aerosol trend has been quantified to explain 81 ±
53 16% of the European surface shortwave trend and 23 ± 5% of the European surface temperature warming. It
54 is likely that trends in anthropogenic aerosols in Europe have generated positive trends in shortwave
55 radiation and surface temperature since the 1980s (see also Sections 6.3.3.1, 8.3.1.6 and 10.6.4).
56
Do Not Cite, Quote or Distribute Atlas-82 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Assessments of observed European trends in meteorological extremes and CIDs are reported elsewhere in
2 this report. Section 11.3.5 documents and attributes an increase in the frequency and extent of heatwaves and
3 daily maximum temperatures, and Section 11.6.2 discusses the uncertainty concerning the detection of trends
4 in meteorological droughts, and the role of increasing atmospheric evaporative demand on hydrological and
5 ecological/agricultural droughts. Section 8.3.1.8 reports on increasing aridity trends in the Mediterranean
6 related to soil moisture declines and increases in atmospheric water vapor demand. Section 11.4.2 reports on
7 the increased likelihood and intensity of daily precipitation extremes, while Sections 11.5.2 and 12.4.5.2
8 discuss implications for peak streamflow. Section 12.4.5.5 discusses the increased likelihood of wildfires,
9 while Section 12.4.5.3 discusses the substantial decadal variability in mean wind speed and the trends in
10 wind storms and gusts. The acceleration of sea level rise in the Atlantic and European seas has been
11 discussed in Section 12.4.5.5.
12
13
14 Atlas.8.3 Assessment of model performance
15
16 A global evaluation of annual mean temperature and precipitation from the CMIP6 ensemble is presented in
17 Sections 3.3.1 and 3.3.2 respectively. In general, annual mean temperature is slightly underestimated at high
18 latitudes and overestimated in the MED area. Temporal evolution of decadal temperature oscillations in
19 Europe simulated by the CMIP6 historical simulations is well reproduced (Fan et al., 2020). Fernandez-
20 Granja et al. (2021) report an overall improvement of CMIP6 compared to CMIP5 to reproduce atmospheric
21 weather patterns over Europe.
22
23 Regional climate models (RCMs; Section 10.3.1.2) have been extensively evaluated for a range of climate
24 features over Europe (Casanueva et al., 2016; Vaittinada Ayar et al., 2016; Ivanov et al., 2017; Krakovska et
25 al., 2017; Terzago et al., 2017; Cavicchia et al., 2018; Drobinski et al., 2018; Fantini et al., 2018; Harzallah
26 et al., 2018; Panthou et al., 2018b). Standard assessments of RCMs driven by reanalyses, typically run at 12–
27 25 km spatial resolution, confirm that the Euro-CORDEX and Med-CORDEX ensembles are capable of
28 reproducing the salient features of European climate (Kotlarski et al., 2014; Krakovska, 2018) and represent
29 European circulation features realistically (Cardoso et al., 2016; Drobinski et al., 2018; Flaounas et al., 2018;
30 Sanchez-Gomez and Somot, 2018). Seasonal and regionally averaged temperature biases generally do not
31 exceed 1.5°C, while precipitation biases can be up to ±40% (Kotlarski et al., 2014). Extensive evaluation of a
32 large collection of RCM-GCM combinations show a general wet, cold and windy bias compared to
33 observations and reanalyses, but none of the models is systematically performing best or worst (Vautard et
34 al., 2020). Higher-resolution simulations do show improved performance in reproducing the spatial patterns
35 and seasonal cycle of not only extreme precipitation but also mean precipitation over all European regions
36 (Mayer et al., 2015; Fantini et al., 2018; Soares and Cardoso, 2018; Ciarlo et al., 2020)(see also Sections
37 10.3.3.4 and 10.3.3.5 for an extensive evaluation of the added value of increased simulation resolution).
38
39 In line with findings reported in Section 10.3.3.8, several studies argue that both GCMs and RCMs
40 underestimate the observed trend in European summer temperature (Dosio, 2016; Boé et al., 2020b),
41 indicating that essential processes are missing or that the natural variability is not correctly sampled
42 (Dell’Aquila et al., 2018). Nabat et al. (2014) argued that including realistic aerosol variations enables
43 climate models to correctly reproduce the summer warming trend (as is required for attributing continental
44 annual temperature trends, Section 8.3.1.1). However, other studies showed models to be sensitive also to
45 local effects, such as land-surface processes, convection, microphysics, and snow albedo (Vautard et al.,
46 2013; Davin et al., 2016). In Euro-CORDEX the warm and dry summer bias over southern and south-eastern
47 Europe is reduced compared to the previous ENSEMBLES simulations (Katragkou et al., 2015; Giot et al.,
48 2016; Prein and Gobiet, 2017; Dell’Aquila et al., 2018). Natural variability has strongly affected the
49 historical warming and large ensembles are necessary for a correct estimation of the forced signal versus
50 natural variability (Aalbers et al., 2018; Lehner et al., 2020).
51
52 Specific assessments of Convection Permitting RCMs (CPRCMs, running at a resolution of typically 1 to 3
53 km and designed for extreme precipitation characteristics) is undertaken in Section 10.3.3.4.1. A unique
54 CPRCM ensemble has been applied over the great Alpine domain and improves representation of mean and
55 extreme precipitation compared to coarser resolution models (Ban et al., 2021; Pichelli et al., 2021).
Do Not Cite, Quote or Distribute Atlas-83 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 The role of aerosol forcing is increasingly analysed as new and more realistic aerosol datasets become
2 available (Nabat et al., 2013; Pavlidis et al., 2020), and as RCMs begin to include interactive aerosols (Nabat
3 et al., 2012, 2015, 2020; Drugé et al., 2019). Explicitly accounting for aerosol effects in RCMs leads to
4 improved representation of the surface shortwave radiation at various scales: long-term means (Gutiérrez et
5 al., 2018), day-to-day variability (Nabat et al., 2015), and long-term trends (Nabat et al., 2014).
6
7 New, or updated, higher-resolution, coupled atmosphere-ocean-ice model systems have been found to
8 improve simulations of observed climate features over the Baltic area compared to atmosphere-only model
9 versions, including correlation between precipitation and SST, between surface heat-flux components and
10 SST, and weather events like convective snow bands over the Baltic Sea (e.g., Tian et al., 2013; Van Pham et
11 al., 2014; Gröger et al., 2015; Wang et al., 2015a; Pham et al., 2017). Coupled atmosphere-land-river-ocean
12 Regional Climate System Models (RCSMs) from Med-CORDEX have similar skill as the ENSEMBLES and
13 the Euro-CORDEX ensembles to represent decadal variability of Mediterranean climate and its extremes
14 (Cavicchia et al., 2018; Dell’Aquila et al., 2018; Gaertner et al., 2018). Panthou et al. (2018b) showed that,
15 over land, differences between atmosphere-only and coupled RCMs are confined to coastal areas that are
16 directly influenced by SST anomalies. In contrast, Van Pham et al. (2014) showed significant differences in
17 seasonal mean temperature across a widespread continental domain.
18
19 Statistical downscaling methods are assessed in Section 10.3.3.7, including the intercomparison and
20 evaluation activities performed in the framework of VALUE and Euro-CORDEX over Europe.
21
22
23 Atlas.8.4 Assessment and synthesis of projections
24
25 Simulations from CMIP5 and CMIP6 indicate pronounced geographical patterns and scenario dependence of
26 the projections of mean temperature and precipitation. Global warming projected under SSP5-8.5 emissions
27 in CMIP6 exceeds the warming projected by RCP8.5 emissions in CMIP5 (Forster et al., 2019) (see also
28 Section 4.3). In selected regions in Europe CMIP6 also projects a systematically higher mean temperature
29 than CMIP5 (Seneviratne and Hauser, 2020). The annual mean projections from CMIP5, CMIP6 and 0.11°
30 resolution EURO-CORDEX contained in the Interactive Atlas are shown for the four European regions in
31 Figure Atlas.24. For each region and season a warming offset between the preindustrial (1850–1900) and the
32 recent past (1995–2014) baselines is also shown. The results confirm higher CMIP6 long-term annual mean
33 warming rates for WCE, EEU and MED and a larger inter-model spread for each region. For given global
34 warming levels, regional annual mean temperature change in CMIP5 and CMIP6 are largely consistent and
35 higher than the global average, most prominently in EEU. For high warming levels the CMIP5 subset of 8
36 GCMs used to drive the EURO-CORDEX simulations show a lower annual mean temperature change than
37 the full CMIP5 ensemble in each of the European subregions. This illustrates the large inter-model spread
38 and implications for subsampling a relatively small subset from the full ensemble. Regional warming is
39 strongest in continental EEU away from the Atlantic and in MED during summer (Lionello and Scarascia,
40 2018). The assessment of EURO-CORDEX projections for levels of global warming of 1.5°C and 2.0°C
41 indicate enhanced local warming even at relatively low global warming levels, particularly towards the north
42 in winter (Schaller et al., 2016; Dosio and Fischer, 2018; Kjellström et al., 2018; Teichmann et al., 2018).
43
44 Some signatures of climate change projected by GCMs are modified by RCMs and CPRCMs. Projections of
45 temperature, precipitation and wind in RCMs may deviate from GCM signals dependent on the dominant
46 atmospheric circulation (Kjellström et al., 2018). In many areas RCMs produce lower warming rates and
47 higher precipitation (less drying) in summer (Fernández et al., 2019; Boé et al., 2020a). Also for mean
48 surface shortwave radiation systematic differences between GCM and RCM outputs are found (Bartók et al.,
49 2017; Gutiérrez et al., 2020). Although RCMs generally have a smaller bias for the present climate (Sørland
50 et al., 2018) and better cloud representation (Bartók et al., 2017), the representation of aerosol forcing (Boé
51 et al., 2020a; Gutiérrez et al., 2020), air-sea coupling (Boé et al., 2020a) or vegetation response to elevated
52 atmospheric CO2 (Schwingshackl et al., 2019) give rise to systematic biases in RCM projections. The
53 comparison between EURO-CORDEX and the CMIP5 subset shown in Figure Atlas.24 illustrates that the
54 RCMs primarily modify the climate change warming signal from the driving GCMs for MED and WCE in
55 summer (Boé et al., 2020a).
Do Not Cite, Quote or Distribute Atlas-84 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Changes in precipitation clearly show a seasonal signature and a meridional gradient over Europe. Mean
2 precipitation increases by 4% to 5% per °C of global warming in NEU, EEU and WCE in DJF, and
3 decreases in summer in WCE and MED (Jacob et al., 2018) (see Figure Atlas.24). CMIP5 projections of
4 precipitation change in MED are strongest in DJF in the south, while changes in JJA are dominant in the
5 northern (European) part of MED, Lionello and Scarascia (2018). The European north-south gradient in
6 precipitation response is confirmed by the EURO-CORDEX experiment (Coppola et al., 2020), but Figure
7 Atlas.24 shows that the JJA precipitation reduction in WCE projected by CMIP5 and CMIP6 at higher
8 warming levels has low confidence in the CORDEX simulations. Precipitation in JJA in EEU is reduced in
9 CMIP6, while little change is shown in CMIP5. Quantitative estimations of climate change features from
10 regional climate projections in Eastern Europe (Partasenok et al., 2015; Kattsov et al., 2017) have low
11 confidence due to the use of relatively small ensembles of GCMs and/or RCMs, and limited evaluation of
12 model performance in the region.
13
14 Over specific geographic features such as high mountains, RCMs further modify the climate change signal of
15 precipitation simulated by the low-resolution GCMs (Giorgi et al., 2016; Torma and Giorgi, 2020). This is
16 especially true for summer precipitation over the Alps where opposite signs of changes in mean and extreme
17 precipitation are generated by the CMIP5 GCM ensemble and the 12-km Med-CORDEX and EURO-
18 CORDEX RCM ensembles (Giorgi et al., 2016) (see also Section 10.6.4.7).
19
20
21 [START FIGURE ATLAS.24 HERE]
22
23 Figure Atlas.24: Regional mean changes in annual mean surface air temperature and precipitation relative to the
24 1995–2014 baseline for the reference regions in Europe (warming since the 1850–1900 pre-
25 industrial baseline is also provided as an offset). Bar plots in the left panel of each region triplet
26 show the median (dots) and 10th–90th percentile range (bars) across each model ensemble for annual
27 mean temperature changes for four datasets (CMIP5 in intermediate colours; subset of CMIP5 used to
28 drive CORDEX in light colours; CORDEX overlying the CMIP5 subset with dashed bars; and CMIP6
29 in solid colours); the first six groups of bars represent the regional warming over two time periods
30 (near-term 2021–2040 and long-term 2081–2100) for three scenarios (SSP1-2.6/RCP2.6, SSP2-
31 4.5/RCP4.5, and SSP5-8.5/RCP8.5), and the remaining bars correspond to four global warming levels
32 (GWL: 1.5°C, 2°C, 3°C, and 4°C). The scatter diagrams of temperature against precipitation changes
33 display the median (dots) and 10th–90th percentile ranges for the above four warming levels for
34 December-January-February (DJF; middle panel) and June-July-August (JJA; right panel),
35 respectively; for the CMIP5 subset only the percentile range of temperature is shown, and only for
36 3°C and 4°C GWLs. Changes are absolute for temperature (in °C) and relative (as %) for
37 precipitation. See Section Atlas.1.3 for more details on reference regions (Iturbide et al., 2020) and
38 Section Atlas.1.4 for details on model data selection and processing. The script used to generate this
39 figure is available online (Iturbide et al., 2021) and similar results can be generated in the Interactive
40 Atlas for flexibly defined seasonal periods. Further details on data sources and processing are
41 available in the chapter data table (Table Atlas.SM.15).
42
43 [END FIGURE ATLAS.24 HERE]
44
45
46 Regional warming is virtually certain to extend the observed downward trends in snow accumulation, snow
47 water equivalent and length of the snow cover season in NEU and low altitudes in mountainous areas in the
48 Alps and Pyrenees (very high confidence). This is supported by regional and global multi-model and/or
49 single-model ensemble projections including CMIP5, PRUDENCE, ENSEMBLES and EURO-CORDEX
50 (Jylhä et al., 2008; Steger et al., 2013; Mankin and Diffenbaugh, 2015; Schmucki et al., 2015; Marty et al.,
51 2017; Frei et al., 2018) and attributed to changes in the snowfall fraction of precipitation and to increased
52 snow melt. In mountain areas a strong dependence of projected snow trends on altitude is shown, with most
53 pronounced effects below 1500 m (López-Moreno et al., 2009). Terzago et al. (2017) showed a large positive
54 bias in the amplitude of the annual snow cycle of EURO-CORDEX 0.11° simulations driven by GCM
55 projections, while reanalysis-driven RCMs showed good agreement with in-situ observations.
56
57 Regional ocean warming in projections with RCSMs for the Baltic and North Sea (Gröger et al., 2015) and
Do Not Cite, Quote or Distribute Atlas-85 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 for the Mediterranean (Darmaraki et al., 2019) is associated with increased intensity and frequency of marine
2 heatwaves in the Mediterranean (see Section 12.4.5.5), strong freshening in the Baltic, and, for some
3 simulations, changes in the circulation in response to non-uniform changes in air-sea interaction (Dieterich et
4 al., 2019). Med-CORDEX RCSM and CMIP5 GCM results agree well on the Mediterranean SST warming
5 rate (Mariotti et al., 2015; Darmaraki et al., 2019); see also the Interactive Atlas.
6
7 Assessments of projected changes in meteorological extremes and CIDs are reported elsewhere in this report.
8 Extreme precipitation and temperature often exhibit a different response to global warming than mean
9 values. Increased intensity and frequency of extreme temperatures and heatwaves is assessed in Sections
10 11.3.5 and 12.4.5.1. Changes in the hydrological cycle include enhanced soil moisture decline in southern
11 Europe, drying in summer and autumn in central Europe, and spring drought due to early snow melt in
12 northern Europe (Sections 8.4.1, 11.6.5 and 12.4.5.2). Changes in mean and extreme wind are very uncertain
13 (Section 12.4.5.3), while sea level rise will increase the frequency of occurrence of extreme sea level at most
14 European coasts (Section 12.4.5.5).
15
16
17 Atlas.8.5 Summary
18
19 An assessment of recent literature largely confirms the findings of previous IPCC reports but with additional
20 detail and (in some cases) higher confidence due to improvements in observations, reanalyses and methods.
21 Observational data sets with global coverage are complemented by the E-OBS gridded ensemble temperature
22 and precipitation dataset, a range of regional observational analyses, and regional reanalysis products. New
23 RCM experiments, including CPRCMs and regional coupled climate system models, mostly coordinated
24 under the umbrella of CORDEX, have generated many new projections and process studies.
25
26 The representation of mean European climate features by GCMs and RCMs is improved compared to
27 previous IPCC assessments (medium confidence), in spite of persisting biases in annual mean and seasonal
28 temperature and precipitation characteristics. The added value of regional downscaling of GCMs by RCM
29 projections for summer mean temperature, precipitation and shortwave radiation is constrained by the
30 representation of processes that lead to a systematic difference between RCM and driving GCM, such as
31 aerosol forcing (medium confidence).
32
33 It is virtually certain that annual mean temperature continues to increase in each European region. There is
34 medium confidence that annual mean precipitation in NEU, WCE and EEU has increased since the early 20th
35 century. In the European Mediterranean trends in annual mean precipitation contain substantial spatial and
36 temporal variability (medium confidence). It is very likely that since the early 1980s in snow-dominant areas
37 in NEU and EEU the length of the snowfall season is reduced with regional warming, and the melt onset
38 dates have advanced. It is likely that decreasing trends in anthropogenic aerosols in Europe have generated
39 positive trends in shortwave radiation and surface temperature since the 1980s.
40
41 At increasing levels of global warming, there is very high confidence that temperature will increase in all
42 European areas at a rate exceeding global mean temperature increases, while increased mean precipitation
43 amounts at high latitudes in DJF and reduced JJA precipitation in southern Europe will occur with medium
44 confidence for global warming levels of 2°C or less, and with high confidence for higher warming levels. At
45 high latitudes and low-altitude mountain areas in Europe strong declines in snow accumulation are virtually
46 certain to occur with further increasing regional temperatures (very high confidence).
47
48
49 Atlas.9 North America
50
51 The assessment in this section focuses on changes in average temperature and precipitation (rainfall and
52 snow) for North America, including the most recent years of observations, updates to observed datasets, the
53 consideration of recent studies using CMIP5 and those using CMIP6 and CORDEX simulations. Assessment
54 of changes in extremes are in Chapter 11 (Table 11.19–21) and climatic impact-drivers in Chapter 12 (Table
55 12.8).
Do Not Cite, Quote or Distribute Atlas-86 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.9.1 Key features of the regional climate and findings from previous IPCC assessments
2
3 Atlas.9.1.1 Key features of the regional climate
4
5 The recent-past climate of North America is characterized by high spatial heterogeneity and by variability at
6 diverse temporal scales. Considering the traditional Köppen-Geiger classification, North America covers all
7 main climate types (see reference region descriptions below). Important geographical features influence local
8 climates over various distances, like the Rocky Mountains through cyclogenesis (Grise et al., 2013) and the
9 Great Lakes through lake-effect snowfall (Wright et al., 2013). The cryosphere is an important component of
10 the climate system in North America, with fundamental roles for sea ice cover, snow cover and permafrost.
11 The ocean surrounding the continent also influence its climate, with water temperatures strongly influencing
12 hurricane activity which impacts the coasts of eastern Mexico and south-eastern USA (Walsh et al., 2010).
13 Temporal variability is influenced by several large-scale atmospheric modes (Annex IV, Table Atlas.1) with
14 the North Atlantic Oscillation (NAO) affecting north-eastern USA and eastern Canada precipitation (Whan
15 and Zwiers, 2017), and El Niño–Southern Oscillation (ENSO) affecting temperature and precipitation in
16 California, although in a complex and not yet fully understood manner (Yoon et al., 2015; Yeh et al., 2018).
17
18 The reference regions defined for summarising North America climate change (Figure Atlas.26) include:
19 North-Western North America (NWN), characterized by a subarctic climate with cool summers and rainfall
20 all year round; North-Eastern North America (NEN), which also has a subarctic climate with sections of
21 tundra climate in the far north (these two northern regions are also discussed in Section Atlas.11.2, Polar
22 Arctic; Western North America (WNA), which has a complex but mainly cold semi-arid climate; the Central
23 North America (CNA) with a mainly by continental climate in the northern part of the region and humid
24 subtropical in the southern portion; Eastern North America (ENA) with a humid continental climate in the
25 northern half and a humid subtropical climate to the south; Northern Central America (northern Mexico)
26 (NCA), has a temperate climate to the north of the Tropic of Cancer, with marked differences between
27 winter and summer, modulated by the North American monsoon (Peel et al., 2007). NCA, also covered in
28 Section Atlas.7.1 Central America, has a temperate climate to the north of the Tropic of Cancer, with marked
29 differences between winter and summer, modulated by the North American monsoon (Peel et al., 2007).
30
31
32 Atlas.9.1.2 Findings from previous IPCC assessments
33
34 The IPCC AR5 (Bindoff et al., 2013; Hartmann et al., 2013) found that the climate of North America has
35 changed due to anthropogenic causes (high confidence), in particular with primarily increasing annual
36 precipitation and annual temperature (very high confidence). Assessment of CMIP5 ensemble projections
37 concluded that mean annual temperature over North America and annual precipitation north of 45N will very
38 likely continue to increase in the future. Also, CMIP5 projects increases in winter precipitation over Canada
39 and Alaska and decreases in winter precipitation over the southwestern USA and much of Mexico.
40
41 The CMIP5 multi-model ensemble generally reproduces the observed spatial patterns but somewhat
42 underestimates the extent and intensity of the North American Monsoon, and also underestimates wetting
43 over Central North America over the period of 1950–2012 during winter season according to AR5 (Flato et
44 al., 2013). In the long-term (2081–2100), the largest changes of precipitation over North America are
45 projected to occur in the mid and high latitudes and during winter (Kirtman et al., 2013).
46
47 SR1.5 (Hoegh-Guldberg et al., 2018) reported a stronger warming compared to the global mean over Central
48 and eastern North America, and a weakening of storm activity over North America under 1.5°C of global
49 warming. SROCC (Hock et al., 2019b) reported that snow depth or mass is projected to decline by 25%
50 mainly at lower elevations over the high mountains in Western North America. SRCCL (Mirzabaev et al.,
51 2019) observed vegetation greening in central North America with high confidence.
52
53
54
55
Do Not Cite, Quote or Distribute Atlas-87 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.9.2 Assessment and synthesis of observations, trends, and attribution
2
3 The observed trends in annual mean surface temperature (see Figure Atlas.11 and the Interactive Atlas)
4 across near-Arctic latitudes are exceptionally pronounced (>0.5°C per decade), significant and consistent
5 across datasets except for far northeast Canada where trends are not significant in the CRU dataset.
6 Significant positive trends are seen across the rest of North America during 1960–2015 (Figure Atlas.11)
7 though over the shorter 1980–2015 period the regional dataset Daymet (Thornton et al., 2016) records non-
8 significant changes over southern Alaska, western and south-central Canada, and north-central USA
9 (Interactive Atlas). An analysis of annual mean surface temperature in the Berkeley Earth dataset aggregated
10 over the reference regions (Figure Atlas.11) demonstrates that a temperature change signal has emerged over
11 all regions of North America. There is a detectable anthropogenic influence (medium confidence) on the
12 observed upward annual temperature trends in western and northern North America (Vose et al., 2017; Wang
13 et al., 2017b; Smith et al., 2018).
14
15 Compared to temperature, trends in annual precipitation over 1960–2015 are generally non-significant
16 though there are consistent positive trends over parts of ENA and CNA (Figure Atlas.11) and Daymet
17 (Interactive Atlas) (high confidence). The global and regional datasets in Figure Atlas.11 and Interactive
18 Atlas also indicate significant decreases in precipitation in parts of south-western US and north-western
19 Mexico (see also Figure 2.15) though these are not all spatially coherent so there is only medium confidence
20 in a drying trend over this region.
21
22 Several factors account for the differences in temperature and precipitation trend significance. Observed
23 trends in precipitation are relatively modest compared to the very large natural interannual variability of
24 precipitation. Furthermore, the precipitation observing network is spatially inadequate (see Section 10.2.2.3)
25 and temporally inconsistent (see Section 10.2.2.2) over some regions of North America, particularly over
26 Arctic and mountainous areas. So detection of multidecadal trends is difficult, especially for regions with
27 summertime convective precipitation maxima that may be spatially patchy (Easterling et al., 2017). See
28 Section 2.3 for further discussion of precipitation trends.
29
30 There is evidence of a recent decline in the overall North American annual maximum snow mass, with a
31 trend for non-alpine regions above 40°N during 1980–2018 estimated from the bias-corrected GlobSnow 3.0
32 data (Pulliainen et al., 2020) (medium confidence). This is despite technical challenges with in situ
33 measurements and remote sensing retrievals of snow variables (Larue et al., 2017; Smith et al., 2017; Wang
34 et al., 2017a; Zeng et al., 2018), spatial heterogeneity and interpolation assumptions that affect gridded
35 reference products, notably over alpine and forested areas (Mudryk et al., 2015; Dozier et al., 2016; Cantet et
36 al., 2019) and breaks in instruments and procedures (Kunkel et al., 2007; Mortimer et al., 2020). Changes in
37 snow cover have evolved in a complex way, with both positive and negative trends and differing from one
38 metric to another (Knowles, 2015; Brown et al., 2019). Evidence of large-area snow cover decline includes
39 decreases in annual maximum snow depth and in snow water equivalent (Vincent et al., 2015; Kunkel et al.,
40 2016; Mote et al., 2018), as well as a shortening of the snow season duration (Knowles, 2015; Vincent et al.,
41 2015). However, reported snow decline trends are statistically significant only for a fraction of the concerned
42 areas or locations (see Figure Atlas.25) (low confidence). See also Sections 2.3.2.2 and 9.5.3.1.
43
44 Rupp et al. (2013) applied a standard finger-printing approach to CMIP5 models and determined that the
45 decline in Northern Hemisphere spring snow cover extent could only be explained by simulations that
46 included natural and anthropogenic forcing. In an attribution study focusing on direct physical causes, it was
47 found that increased spring snowmelt in northern Canada was driven by warming-induced high latitudes
48 changes such as atmospheric moisture, cloud cover, and energy advection (Mioduszewski et al., 2014).
49 In an analysis of drivers of the record low snow water equivalent (SWE) values of spring 2015 in the western
50 US, it was found that the relative importance of greenhouse gases varies spatially (Mote et al., 2016). See
51 also Section 3.4.2 for further discussion of anthropogenic influences on snow extent.
52
53
54
55
Do Not Cite, Quote or Distribute Atlas-88 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 [START FIGURE ATLAS.25 HERE]
2
3 Figure Atlas.25: Grid-box trends (mm yr–1) in annual maximum snow depth for cold season periods of 1960/1961
4 to 2014/2015. (Left) Numbers indicate number of stations available in that grid box. (Right) Boxes
5 with ‘x’ indicate non-significant trends (at the p < 0.05 level of significance) (Kunkel et al., 2016).
6
7 [END FIGURE ATLAS.25 HERE]
8
9
10 Atlas.9.3 Assessment of model performance
11
12 CMIP6 models have been evaluated in the literature, although these studies have not included the full set of
13 CMIP6 simulations. Fan et al. (2020) established on a continental basis for North America that temperature
14 pattern correlations were quite accurate. Thorarinsdottir et al. (2020) compared maximum and minimum
15 temperatures over Europe and North America with several observational datasets and found that the CMIP6
16 ensemble agreed better with ERA5 data than did CMIP5. Srivastava et al. (2020) evaluated historical CMIP6
17 simulations for precipitation, comparing them with several observational datasets over the continental US.
18 Most models show a wet bias over the eastern half of continental USA and the northeast region while dry
19 biases persist in the central part of the country (Akinsanola et al. (2020a) and Almazroui et al. (2021)). The
20 spatial structure of biases is similar in CMIP5 and CMIP6, but with lower magnitudes in CMIP6. Agel and
21 Barlow (2020) examined 16 CMIP6 models over the north-eastern US for precipitation and did not find a
22 distinct improvement over CMIP5, although they did find the higher resolution models tended to perform
23 better. On the basis of the evidence so far, there is medium confidence that CMIP6 models are improved
24 compared to CMIP5 in terms of biases in mean temperature and precipitation over North America.
25
26 North America has been extensively used as a testbed for regional climate model (RCM) experiments, such
27 as the North American Regional Climate Change Assessment Program (NARCCAP) (Mearns et al., 2009),
28 the MultiRCM Ensemble Downscaling (MRED) (Yoon et al., 2012), and NA-CORDEX (Bukovsky and
29 Mearns, 2020). Therefore, much performance evaluation has been conducted with a focus on specific climate
30 features in North America. For the North American monsoon region, multi-model performance evaluation
31 (Bukovsky et al., 2013; Tripathi and Dominguez, 2013; Cerezo-Mota et al., 2015) or a single member
32 performance (Lucas-Picher et al., 2013; Martynov et al., 2013; Šeparović et al., 2013) demonstrated the
33 added value of RCMs, particularly more recent CORDEX simulations, through improved simulation of
34 summertime precipitation and the climatological wintertime storm tracks across the western United States.
35 NA-CORDEX simulations were more successful at reproducing weather types compared to a single model-
36 based large perturbed-physics ensemble (Prein et al., 2019). The application of a complex evaluation tool to
37 the full suite of NA-CORDEX simulations found that the higher resolution simulations (25 km vs 50 km) of
38 precipitation were improved, particularly for daily intensity (Gibson et al., 2019).
39
40 However, deficiencies have also been reported. For example, storm occurrence over the East Coast of North
41 America was found (Poan et al., 2018), and amplitude in the simulated annual cycle was generally excessive
42 in NA-CORDEX simulations. RCMs tend to produce more (less) precipitation over mountains (the coastal
43 plains) (Cerezo-Mota et al., 2015) and winter precipitation in the western US had large positive biases in all
44 RegCM simulations, regardless of the driving GCM (Mahoney et al., 2021).
45
46 Recently, convective-permitting RCMs have been used to simulate North American climate features and
47 generated better simulations of precipitation. For example, summer precipitation over the south-western
48 USA was improved due to better representation of organized mesoscale convective systems at the sub-daily
49 scale (Castro et al., 2012; Liu et al., 2017; Prein et al., 2017b; Pal et al., 2019), the diurnal cycle of
50 convection (Nesbitt et al., 2008), and in terms of means (and extremes) for the north-eastern United States
51 (Komurcu et al., 2018).
52
53 Recent studies have examined RCMs simulation of SWE, a quantity of primary importance notably for
54 hydrological modelling though its ground measurements are restricted by relatively high time and monetary
55 costs (Smith et al., 2017; Odry et al., 2020) which limit model assessment. Also, studies often emphasize that
Do Not Cite, Quote or Distribute Atlas-89 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 a false impression of model skill for SWE can be obtained by compensating temperature and precipitation
2 biases. Assessment frameworks have dealt with these issues by considering observational uncertainty
3 (Mccrary et al., 2017) and by decomposing SWE biases into their contributing processes (Rhoades et al.,
4 2018; Xu et al., 2019). SWE biases exceed observational uncertainty in several 50-km reanalysis-driven
5 NARCCAP simulations over several regions, for all cold months (Mccrary et al., 2017). Analyses of NA-
6 CORDEX simulations show that refining spatial resolution from 50 to 12 km improves certain (but not all)
7 aspects of SWE, stemming from improved mean precipitation and topography-related temperature (Xu et al.,
8 2019). Similarly an assessment of RCM simulations of freezing rain over Eastern Canada found a mix of
9 improved and deteriorated aspects from higher resolution (St-Pierre et al., 2019).
10
11
12 Atlas.9.4 Assessment and synthesis of projections
13
14 [START FIGURE ATLAS.26 HERE]
15
16 Figure Atlas.26: Regional mean changes in annual mean surface air temperature and precipitation relative to the
17 1995–2014 baseline for the reference regions in North America (warming since the 1850–1900
18 pre-industrial baseline is also provided as an offset). Bar plots in the left panel of each region
19 triplet show the median (dots) and 10th–90th percentile range (bars) across each model ensemble for
20 annual mean temperature changes for four datasets (CMIP5 in intermediate colours; subset of CMIP5
21 used to drive CORDEX in light colours; CORDEX overlying the CMIP5 subset with dashed bars; and
22 CMIP6 in solid colours); the first six groups of bars represent the regional warming over two time
23 periods (near-term 2021–2040 and long-term 2081–2100) for three scenarios (SSP1-2.6/RCP2.6,
24 SSP2-4.5/RCP4.5, and SSP5-8.5/RCP8.5), and the remaining bars correspond to four global warming
25 levels (GWL: 1.5°C, 2°C, 3°C, and 4°C). The scatter diagrams of temperature against precipitation
26 changes display the median (dots) and 10th–90th percentile ranges for the above four warming levels
27 for December-January-February (DJF; middle panel) and June-July-August (JJA; right panel),
28 respectively; for the CMIP5 subset only the percentile range of temperature is shown, and only for
29 3°C and 4°C GWLs. Changes are absolute for temperature (in °C) and relative (as %) for
30 precipitation. See Section Atlas.1.3 for more details on reference regions (Iturbide et al., 2020) and
31 Section Atlas.1.4 for details on model data selection and processing. The script used to generate this
32 figure is available online (Iturbide et al., 2021) and similar results can be generated in the Interactive
33 Atlas for flexibly defined seasonal periods. Further details on data sources and processing are
34 available in the chapter data table (Table Atlas.SM.15).
35
36 [END FIGURE ATLAS.26 HERE]
37
38
39 CMIP5 and CMIP6 surface temperature and precipitation projections over the region are similar, with all
40 regions warming more than the global average, most prominently those in the north most (Figure Atlas.26).
41 CMIP6 projects, for scenarios and time-periods, higher temperature changes (see also Chapter 4) with this
42 contrast more accentuated in the long-term future and at higher global warming levels. The higher warming
43 in the north (also see the Interactive Atlas) is clear when comparing NEN, with increases from 2.5°C to over
44 11°C on an annual basis for SSP5-8.5 (near-term to long-term compared to a 1995–2014 baseline), to NCA,
45 where changes range from 1.5°C to 6°C across the same periods. Maps showing changes in temperature and
46 precipitation, and their robustness, are available in the Interactive Atlas. The number of model results (i.e.,
47 ensemble size used to generate these figures) differs, and this sample size difference may affect the results,
48 but the patterns and magnitudes of change are generally consistent and thus it is very likely that temperatures
49 will increase throughout the 21st century in all land areas, with stronger warming in the far north.
50
51 CMIP5 results have been analysed extensively (e.g., Maloney et al., 2014) and used in major climate change
52 assessments. The most recent US National Climate Assessment analysis of CMIP5 focusing on RCP4.5 and
53 RCP8.5 for two future time periods stated that the USA would continue to warm regardless of scenario but is
54 likely to be higher with higher emissions scenarios (e.g., RCP8.5). Projected changes in precipitation are
55 somewhat complex, but increased precipitation dominates in winter and spring, whereas in summer changes
56 are more variable and uncertain. Canada’s Changing Climate Report (Bush and Lemmen, 2019) presents
57 changes in temperature and precipitation, as well as other variables, such as snow, for future periods in
Do Not Cite, Quote or Distribute Atlas-90 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Canada using results from CMIP5. It indicates that annual and winter precipitation is projected to increase
2 everywhere in Canada over the 21st century with larger percentage increases in the north. Temperature is
3 also projected to increase, regardless of scenario, and with larger changes occurring in the north.
4
5 To provide the basis for generating additional information compared to that derived from CMIP5 the NA-
6 CORDEX experiments were designed to involve a GCM-RCM matrix which included multiple GCMs that
7 sampled the full range of climate sensitivity, multiple RCMs, at two different spatial resolutions (25 and 50
8 km) and a range of emissions scenarios (in most cases RCP4.5 and RCP8.5) (Mearns et al., 2017).
9 Karmalkar (2018) noted that the NA-CORDEX model covers subregional ranges of temperature change
10 from the CMIP5 GCMs better than NARCCAP did for the CMIP3 models. This structural design shift
11 provides greater confidence in the NA-CORDEX results in terms of sampling the uncertainty across the
12 CMIP5 models (Figure Atlas.27) (Bukovsky and Mearns, 2020). The pattern of warming is as seen in
13 CMIP5 and CMIP6 which also builds confidence that the RCMs generate high-resolution results consistent
14 with CMIP5 on large scales whilst providing added value over regions such as the complex topography of
15 the Rocky Mountains in the US West, which are not well resolved in the GCMs. There is high confidence
16 that downscaling a subset of CMIP models that spans the range of climate sensitivities in the full ensemble is
17 critical for producing a representative range of dynamically downscaled projections.
18
19
20 [START FIGURE ATLAS.27 HERE]
21
22 Figure Atlas.27: Changes (2070–2099 relative to 1970–1999) in the annual mean surface air temperature by three
23 GCMs (GFDL ESM2M, MPI ESM-LR, HadGEM2-ES) and two RCMs (WRF and RegCM4) nested
24 in the GCMs, for the RCP8.5 scenario (after Bukovsky and Mearns, 2020).
25
26 [END FIGURE ATLAS.27 HERE]
27
28
29 There are striking contrasts in the seasonal results for precipitation for the subregions (Figure Atlas.26). The
30 northern regions and ENA all show steady increases with the global warming levels (very high confidence).
31 For example, the projected increases in NEN region range from 7% in the near term to 40% at the end of the
32 21st century for the SSP5-8.5 scenario. In contrast, projected changes for NCA are for significant decreases
33 both on an annual basis (Interactive Atlas) and in winter and which become greater as warming increases
34 (Akinsanola et al., 2020b; Almazroui et al., 2021). The other two regions (WNA and CNA) exhibit mainly
35 increases in winter. In summer, distributions are in general less uniform except for NWN and NEN which
36 display steady increases with global warming levels (but smaller than in winter). WNA and CNA mainly
37 show decreases (based on the median values) but with some models projecting increases. Projections from
38 the NA-CORDEX ensemble are consistent with those from the GCMs whilst providing greater detail of
39 precipitation changes over the mountains and along the coasts (Bukovsky and Mearns, 2020)(Interactive
40 Atlas). Similar results are found in other analyses of RCM projections (Wang and Kotamarthi, 2015; Ashfaq
41 et al., 2016; Teichmann et al., 2020). Also, further analysis of the NA-CORDEX projections showed
42 substantial changes in weather types related to increased monsoonal flow frequency and drying of the
43 northern Great Plains in summer (Prein et al., 2019).
44
45 In summary, NEN, NWN and most of ENA will very likely experience increased annual mean precipitation,
46 with greater increases at higher levels of warming (very high confidence). In NCA decreases predominate on
47 an annual basis and particularly in winter (high confidence). Projected changes in summer are highly
48 uncertain throughout other regions apart from the far northern parts of NEN and NWN which will likely
49 experience increases (high confidence).
50
51 As discussed in Section 10.3.3.4 an important advance in regional modelling over the past decade or so is the
52 use of convection-permitting regional models (CPMs) (Prein et al., 2015, 2017b). There have been a number
53 of experiments using CPMs over North America (e.g., Rasmussen et al., 2014; Prein et al., 2015, 2019; Liu
54 et al., 2017; Komurcu et al., 2018). A CPM study over North America that investigated changes in
55 Mesoscale Convective Systems projected that by the end of the century, assuming an RCP8.5 scenario, their
Do Not Cite, Quote or Distribute Atlas-91 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 frequency more than tripled and associated precipitation increased by 80% (Prein et al., 2017a). A multiple
2 nesting of WRF over north-eastern US downscaling to 3 km a CESM GCM climate projection assuming an
3 RCP8.5 scenario, found a different pattern of precipitation change of mixed increases and decreases
4 compared to the GCM projection of increases every month (Komurcu et al., 2018). These investigations
5 demonstrate the potential of very high-resolution simulations to add important dimensions to our
6 understanding of regional climate change, though not necessarily to reduce uncertainty (high confidence).
7
8 It is virtually certain that snow cover will experience a general decline across North America during the 21st
9 century, in terms of extent, annual duration and SWE, based on CMIP5 (Maloney et al., 2014), CMIP6
10 (Mudryk et al., 2020), NA-CORDEX (Mahoney et al., 2021) and NARCCAP (e.g., McCrary and Mearns,
11 2019) simulations. For some regions the decline could be discernible over the next few decades, for example
12 in western USA (Fyfe et al., 2017). It is, however, likely that some high-latitude regions will rather
13 experience an increase in certain winter snow cover properties (Mudryk et al., 2018; McCrary and Mearns,
14 2019), due to snowfall increase (Krasting et al., 2013a) prevailing over the warming effect. Discussion of
15 changes in snow in the future is also covered in Section 9.5.3, but for larger regions.
16
17 The fraction of precipitation falling as snow is projected to decrease practically everywhere over North
18 America, including over western USA and south-western Canada (Mahoney et al., 2021), and in the Great
19 Lakes basin where lake-effect precipitation is important (Suriano and Leathers, 2016). In this basin, the
20 frequency of heavy lake-effect snowstorms is expected to decrease during the 21st century, except for a
21 possible temporary increase around Lake Superior by mid-century, if local air temperatures remain low
22 enough (Notaro et al., 2015). CMIP5 simulations of the periods 1981–2000 and 2081–2100 over central and
23 eastern US suggest a northward shift in the transition zone between rain-dominated and snow-dominated
24 areas, by about 2° latitude under the RCP4.5 scenario and 4° latitude under the RCP8.5 scenario (Ning and
25 Bradley, 2015). Rain-on-snow event properties over North America should also evolve during the 21st
26 century, with non-trivial dependencies on the positioning relative to the freezing line (Jeong and Sushama,
27 2018) and on elevation (Musselman et al., 2018).
28
29
30 Atlas.9.5 Summary
31
32 Across North America it is very likely that positive surface temperature trends are persistent. Across near-
33 Arctic latitudes of North America increases are exceptionally pronounced, greater than 0.5°C per decade
34 (high confidence). In parts of eastern and central North America it is likely that annual precipitation has
35 increased over the period of 1960 to 2015 but with no clear trends in other regions except for parts of south-
36 western US and north-western Mexico where there is medium confidence in drying.
37
38 Model representation of the climatology of mean temperature and precipitation has likely improved
39 compared to the AR5 over North America. This is aided by continuous model development, and the
40 existence of new coordinated modelling initiatives such as NA-CORDEX. There is high confidence that
41 downscaling a subset of CMIP models that spans the range of climate sensitivities in the full ensemble is
42 critical for producing a representative range of dynamically downscaled projections.
43
44 It is virtually certain that annual and seasonal surface temperatures over all of North America will continue
45 to increase at a rate greater than the global average, with greater increases in the far north. It is very likely,
46 based on global and regional model future projections, that on an annual time scale precipitation will
47 increase over most of North America north of about 45°N and in eastern North America, and it is likely that
48 it will decrease in the southwest US and northern Mexico, particularly in winter. Elsewhere the direction of
49 change of precipitation is uncertain. It is virtually certain that snow cover will experience a decline over
50 most regions of North America during the 21st century, in terms of water equivalent, extent and annual
51 duration. It is however very likely that some high-latitude regions will rather experience an increase in winter
52 snow water equivalent, due to the snowfall increase prevailing over the warming effect.
53
54
55
Do Not Cite, Quote or Distribute Atlas-92 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.10 Small islands
2
3 Atlas.10.1 Key features of the regional climate and findings from previous IPCC assessments
4
5 Atlas.10.1.1 Key features of the regional climate
6
7 Many small islands lie in tropical regions and their climate varies depending on a range of factors with
8 location, extent and topography having major influences. In general, their climate is determined by that of
9 the broader region in which they lie as they have little influence on the regional climate although steep
10 topography can induce higher rainfall totals locally. Temperature variability tends to be low due to the
11 influence of the surrounding ocean, most marked in the tropics where oceanic temperature ranges are small.
12 However, seasonal rainfall variability can often be significant, both through the annual cycle and also
13 interannually through the influence of many modes of variability (for more details see Cross-Chapter Box
14 Atlas.2:, Annex IV and Section Atlas.7.1 for the Caribbean). Many small islands are exposed to tropical
15 cyclones and the associated hazards of high winds, storm surges and extreme rainfall, and many low-lying
16 islands are exposed to regular flooding from natural high tide and wave activity. In the Pacific, phases of the
17 El Niño Southern Oscillation result in periods of warmer or cooler than average temperatures following the
18 upper ocean warming of El Niño events or cooling of La Niña events, and respectively weaker and stronger
19 trade winds. El Niño conditions also lead to drought in Melanesian islands and increased typhoons and storm
20 surges in French Polynesia with La Niña conditions causing drought in Kiribati. Other islands experience
21 increased rainfall during these periods.
22
23
24 Atlas.10.1.2 Findings from previous IPCC assessments
25
26 The AR5 noted observed temperature increases of 0.1°C–0.2°C per decade in Pacific Islands and that
27 warming was very likely to continue across all Small Island regions (Christensen et al., 2013; IPCC, 2013a).
28 It also reported decreased rainfall over the Caribbean, increases over the Seychelles, streamflow reductions
29 over the Hawaiian Islands and projections of reduced rainfall over the Caribbean and drier rainy season for
30 many of the southwest Pacific Islands (Christensen et al., 2013; IPCC, 2013a; Nurse et al., 2014). The
31 remaining findings are derived from the SROCC (IPCC, 2019a). Ocean warming rates have likely increased
32 in recent decades with marine heatwaves increasing and very likely to have become longer-lasting, more
33 intense and extensive as a result of anthropogenic warming. Open ocean oxygen levels have very likely
34 decreased and oxygen minimum zones have likely increased in extent. There is very high confidence that
35 global mean sea level rise has accelerated in recent decades which, combined with increases in tropical
36 cyclone winds and rainfall and increases in extreme waves, has exacerbated extreme sea level events and
37 coastal hazards (high confidence). It is virtually certain that during the 21st century, the ocean will transition
38 to unprecedented conditions with further warming and acidification virtually certain, increased upper ocean
39 stratification very likely and continued oxygen decline (medium confidence). There is very high confidence
40 that marine heatwaves and medium confidence that extreme El Niño and La Niña events will become more
41 frequent. It is very likely that these changes will be smaller under scenarios with low greenhouse-gas
42 emissions. Global mean sea level will continue to rise and there is high confidence that the consequent
43 increases in extreme levels will result in local sea levels in most locations that historically occurred once per
44 century occurring at least annually by the end of the century under all RCP scenarios (high confidence). In
45 particular, many small islands are projected to experience historical centennial events at least annually by
46 2050 under RCP2.6 and higher emissions. The proportion of Category 4 and 5 tropical cyclones, and
47 associated precipitation rates and storm surges, along with average tropical cyclone intensity are projected to
48 increase with a 2°C global temperature rise, thereby exacerbating coastal hazards.
49
50
51 Atlas.10.2 Assessment and synthesis of observations, trends and attribution
52
53 Significant positive trends in temperature ranging from 0.15°C per decade (over the period 1953–2010) to
54 0.18°C per decade (over the period 1961–2011) are noted in the tropical Western Pacific, where the
55 significant increasing trends in the warm and cool extremes are also spatially homogeneous (Jones et al.,
Do Not Cite, Quote or Distribute Atlas-93 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 2013; Whan et al., 2014; Wang et al., 2016). Similarly, much of the Caribbean region showed statistically
2 significant warming (at the 95% level) over the 1901–2010 period (Jones et al., 2016b). Observation records
3 in the Caribbean region indicate a significant warming trend of 0.19°C per decade and 0.28°C per decade in
4 daily maximum and minimum temperatures, respectively, with statistically significant increases (at the 5%
5 level) in the number of warm days and warm nights during 1961–2010 (Taylor et al., 2012b; Stephenson et
6 al., 2014; Beharry et al., 2015).
7
8 A weather station-based annual precipitation trend analysis over 1901–2010 in the Caribbean region
9 indicated some locations with detectable decreasing trends (Knutson and Zeng, 2018), which were
10 attributable in part to anthropogenic forcing. These include southern Cuba, the northern Bahamas, and the
11 Windward Islands although significant trends were not found over the shorter periods of 1951–2010 and
12 1981–2010. In Caribbean Islands, a Palmer Drought Severity Index dataset from 1950 to 2016 showed a
13 clear drying trend in the region (Herrera and Ault, 2017). The 2013–2016 period showed the most severe
14 drought during the period and was strongly related to anthropogenic warming, which would have increased
15 the severity of the event by 17% and its spatial extent by 7% (Herrera et al., 2018). However, a seasonal
16 analysis of observations grouped into large subregions of the Caribbean revealed no significant long-term
17 trends in rainfall over 1901–2012 but significant inter-decadal variability (Jones et al., 2016b). Declines in
18 summer rainfall (–4.4% per decade) and maximum five-day rainfall (–32.6 mm per decade) over 1960–2005
19 were reported for Jamaica (CSGM, 2012) and an insignificant decrease in summer precipitation was
20 observed for Cuba for 1960 to 1995 (Naranjo-Diaz and Centella, 1998). Three of four stations examined for
21 Puerto Rico exhibited declining JJA rainfall over 1955–2009 with the trend statistically significant at the
22 95% level for Canóvana (Méndez-Lázaro et al., 2014). In the Caribbean, positive regional trends in
23 precipitation of 2% and trends in extremes during 1961 to 2010 were found to be not statistically significant
24 (at the 5% level) (Stephenson et al., 2014; Beharry et al., 2015). Positive trends in JJA rainfall over Cuba and
25 Jamaica are seen in the CRU, whereas they are negative over Cuba for GPCC; over eastern Hispaniola they
26 are positive in CRU and negative in CHIRPS (Cavazos et al., 2020).
27
28 In Hawai’i, between 1920 and 2012, over 90% of the islands showed reduced rainfall and streamflow, an
29 increase in the frequency of days with zero flow (Strauch et al., 2015; Frazier and Giambelluca, 2017) and
30 robust positive trends in the drought frequency and severity (McGree et al., 2016). Over the western Pacific,
31 interannual and decadal variabilities also drive long-term trends in rainfall. Recent analysis of station data
32 showed spatial variations in the mostly decreasing but non-significant trends in annual and extreme rainfall
33 over the Western Pacific from 1961–2011 (low confidence) (McGree et al., 2014). Over the southern
34 subtropical Pacific decreases in annual, JJA, SON and extreme rainfall and increasing drought frequency in
35 western region has been observed since 1951 (Jovanovic et al., 2012; McGree et al., 2016, 2019).
36
37 Over the Western Indian Ocean significant warming trends have been reported for Mauritius (1.2°C during
38 1951–2016 (MESDDBM, 2016)), La Réunion (0.18°C per decade over 1968–2019 (Météo-France, 2020))
39 and Maldives (MEE, 2016). Both Mauritius and La Réunion have experienced rainfall decreases of 8%
40 during 1951–2016 and 1.2% per decade during 1961–2019 with generally weak, non-significant rainfall
41 trends during 1967–2012.
42
43 Assessing observed climate change for Small Islands is often constrained by low station density (Ryu and
44 Hayhoe, 2014; Jones et al., 2016c), digitization requirements or data sharing limitations (Jones et al., 2016c).
45 Station data typically have longer temporal coverage relative to satellite products but are limited in spatial
46 coverage (Cavazos et al., 2020). For Small Island nations, spatial gaps between observations can be very
47 large due to isolation of the islands (Wright et al., 2016). Additionally over past decades, the number of
48 station observations has declined substantially in Mauritius (Dhurmea et al., 2019), Hawai‘i (Bassiouni and
49 Oki, 2013; Frazier and Giambelluca, 2017) and most Pacific Island countries since the 1980s (Jones et al.,
50 2013; McGree et al., 2014, 2016). In Fiji, meteorological stations were located on or by the coast and sparse
51 in the interior (Kumar et al., 2013). Notable topography and land use may result in changes in climatic
52 conditions over small distances (Foley, 2018), making the observational density particularly relevant.
53
54 Moreover, many stations have little metadata available, including those in Vanuatu, the Solomon Islands and
55 Papua New Guinea (Whan et al., 2014). Compared to earlier decades, little metadata are currently being
Do Not Cite, Quote or Distribute Atlas-94 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 documented in the western Pacific islands (McGree et al., 2014), which will challenge the homogenization of
2 long-term observational records. Challenges in the Caribbean include maintaining continuous daily time
3 series with metadata, converting climatological data into digital formats and making them freely available
4 (Stephenson et al., 2014; Beharry et al., 2015; Jones et al., 2016c). This is also an issue in the Pacific as
5 many data are kept in national (local) databases, with only a fraction having been incorporated into global
6 datasets (Whan et al., 2014).
7
8 Because of the small number of stations used for interpolation and the complex mountainous topography,
9 gridded product for these small islands should be interpreted with caution (Frazier and Giambelluca, 2017).
10 For the Antilles, the error in estimating CRU2.0 monthly precipitation can stand locally between 20% and
11 40%. Over the Caribbean, (Cavazos et al., 2020) found a discrepancy across gridded observational datasets
12 (CRU, CHIRPS, and GPCP) in detecting orographic precipitation especially during boreal summer making
13 their use in climate model evaluation challenging (Herrera and Ault, 2017). Furthermore, some reanalysis
14 products such as the 0.7° x 0.7° ERA-Interim reanalysis are not adequate as many of the smaller Caribbean
15 islands are not represented as land (Jones et al., 2016c).
16
17
18 Atlas.10.3 Assessment of model performance
19
20 An assessment of model performance for the Caribbean region is contained in Section Atlas.7.1 on Central
21 America. In summary, the ability of climate models to simulate the climate over the region has improved in
22 many key respects with the application of increased model resolution and a better representation of the land-
23 surface processes of particular importance in these advances (high confidence). Regional climate models
24 (RCMs) simulate realistically seasonal surface temperature and precipitation patterns including the bimodal
25 rainfall in the precipitation annual cycle although with some timing biases in some regions (high confidence).
26 The important regional circulation and precipitation features, the Caribbean Low-Level Jet and the mid-
27 summer drought, are well represented over a variety of RCM domains covering the region (high confidence).
28
29 Over the tropical Pacific, surface temperature biases in CMIP6 models remain similar to those in CMIP5,
30 although are reduced in the higher resolution models in the HiResMIP ensemble. CMIP6 models generally
31 represents trends in sea surface temperatures better than CMIP5 (see Section 9.2.1 for more details). For
32 precipitation, the persistent tropical Pacific bias of the double ITCZ (erronious bands of excessive rainfall
33 both sides of the equatorial Pacific) is still present in CMIP6 models although is slightly improved compared
34 to those in CMIP3 and CMIP5 models (Section 3.3.2.2). Application of downscaling techniques (RCMs and
35 stretched-grid GCMs) using resolutions finer than 10 km over the Pacific can capture topographic influences
36 on wind and rainfall to generate realistic simulations of island climates – for example over Fiji and New
37 Caledonia, (Chattopadhyay and Katzfey, 2015; Dutheil et al., 2019). In both cases applying bias adjustment
38 to the sea surface temperatures used as a lower boundary condition for the downscaling models was
39 important to generate realistic simulations.
40
41
42 Atlas.10.4 Assessment and synthesis of projections
43
44 Projected median temperature increases for Small Islands from the CMIP5 ensemble range from 1°C
45 (RCP4.5) to 1.5°C (RCP8.5) in the period 2046–2065, and from 1.3°C (RCP4.5) to 2.8°C (RCP8.5) by
46 2081–2100 relative to 1986–2005 (Harter et al., 2015). Spatial variations in the warming trend are projected
47 to increase by the end of the 21st century, with relatively higher increases in the Arctic and sub-Arctic
48 islands, and in the equatorial regions compared with islands in the Southern Ocean (Harter et al., 2015). In
49 the western Pacific, temperatures are projected to increase by 2.0°C to 4.5°C by the end of the 21st century
50 relative to 1961–1990 (Wang et al., 2016). The warming over land in the Lesser Antilles is estimated to be
51 about 1.6°C (3.0°C) by 2071–2100 for the RCP4.5 (RCP8.5) scenario, relative to 1971–2000 (Cantet et al.,
52 2014). Projections from the CMIP6 ensemble support these findings (Figure Atlas.28) and across global
53 warming levels from 1.5°C to 4°C CMIP5 and CMIP6 consistently project lower levels of warming for
54 Small Islands than the global average (see also the Interactive Atlas).
55
Do Not Cite, Quote or Distribute Atlas-95 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 [START FIGURE ATLAS.28 HERE]
2
3 Figure Atlas.28: Regional mean changes in annual mean surface air temperature, precipitation and sea level rise
4 relative to the 1995–2014 baseline for the reference regions in Small Islands (warming since the
5 1850–1900 pre-industrial baseline is also provided as an offset). Bar plots in the left panel of each
6 region triplet show the median (dots) and 10th–90th percentile range (bars) across each model
7 ensemble for annual mean temperature changes for four datasets (CMIP5 in intermediate colours;
8 subset of CMIP5 used to drive CORDEX in light colours; CORDEX overlying the CMIP5 subset with
9 dashed bars; and CMIP6 in solid colours); the first six groups of bars represent the regional warming
10 over two time periods (near-term 2021–2040 and long-term 2081–2100) for three scenarios (SSP1-
11 2.6/RCP2.6, SSP2-4.5/RCP4.5, and SSP5-8.5/RCP8.5), and the remaining bars correspond to four
12 global warming levels (GWL: 1.5°C, 2°C, 3°C, and 4°C). Bar plots in the right panel show the median
13 (dots) and 5th–95th percentile range (bars) sea level rise from the CMIP6 ensemble (see Chapter 9 for
14 details) for the same time periods and scenarios. The scatter diagrams of temperature against
15 precipitation changes display the median (dots) and 10th–90th percentile ranges for the above four
16 warming levels for December-January-February (DJF; middle panel) and June-July-August (JJA;
17 right panel), respectively; for the CMIP5 subset only the percentile range of temperature is shown,
18 and only for 3°C and 4°C GWLs. Changes are absolute for temperature (in °C) and relative (as %) for
19 precipitation. See Section Atlas.1.3 for more details on reference regions (Iturbide et al., 2020) and
20 Section Atlas.1.4 for details on model data selection and processing. The script used to generate this
21 figure is available online (Iturbide et al., 2021) and similar results can be generated in the Interactive
22 Atlas for flexibly defined seasonal periods. Further details on data sources and processing are
23 available in the chapter data table (Table Atlas.SM.15).
24
25 [END FIGURE ATLAS.28 HERE]
26
27
28 The CMIP5 ensemble median projected a precipitation decreases of up to –16% over the Caribbean, parts of
29 the Atlantic and Indian Oceans and southern sub-tropical and eastern Pacific Ocean, and increases of up to
30 10% over parts of the western Pacific and Southern Ocean, and up to 55% in the equatorial Pacific islands
31 under RCP6.0 in the period 2081–2100 relative to 1986–2005 (Harter et al., 2015). A projected decrease in
32 annual precipitation is also noted over the Lesser Antilles under the RCP4.5 and RCP8.5 scenarios (Cantet et
33 al., 2014). Seasonal rainfall is projected to decrease in most areas in Hawai’i, except for the climatically wet
34 windward side of the mountains, which would increase the wet and dry gradient over the area (Timm et al.,
35 2015). The average precipitation changes in Hawai’i are estimated to be about –11% to –28% under RCP4.5
36 during the wet season, and about –4% to –28% under RCP4.5 during the dry season in the period 2041–2071
37 relative to 1975–2005, with larger changes under RCP8.5 (Timm et al., 2015). There are still uncertainties in
38 the projected changes, which have been attributed to factors including insufficient model skill in representing
39 topography in the small islands, and high variability in climate drivers. However, the broad-scale pattern of
40 projected wetter conditions in the western and equatorial Pacific, North Indian and Southern Oceans and of
41 drier conditions over the Caribbean, parts of the Atlantic, West Indian and the southern sub-tropical and
42 eastern Pacific Oceans are further strengthened in the CMIP6 ensemble (Figure Atlas.28) which are thus
43 likely regional responses as the climate continues to warm.
44
45 The negative trend in future summer rainfall in the Caribbean and Central America is projected to be
46 strongest during the mid-summer (June–August) based on studies using GCMs (Rauscher et al., 2008;
47 Karmalkar et al., 2013; Karmacharya et al., 2017; Taylor et al., 2018). The future summer drying over the
48 Caribbean is associated with a projected future strengthening of the Caribbean Low-Level Jet (Taylor et al.,
49 2013b). Rauscher et al. (2008) hypothesized that the simulated 21st century drying over Central America
50 represents an early onset and intensification of the mid-summer drought. The westward expansion and
51 intensification of the NASH associated with the mid-summer drought occurs earlier under A1B, with
52 stronger low-level easterlies. Rauscher et al. (2008) further suggested that the eastern Pacific ITCZ is also
53 located further southward and that there are some indications that these changes could be forced by ENSO-
54 like warming of the tropical eastern Pacific and increased land-ocean heating contrasts over the North
55 American continent. Other studies also suggest a future intensification of the NASH due to changes in land-
56 sea temperature contrast resulting from increased greenhouse-gas concentrations (Li et al., 2012b).
57
Do Not Cite, Quote or Distribute Atlas-96 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.10.5 Summary
2
3 It is very likely that all Small Island regions have warmed with significant trends recorded from at least the
4 1960s in all territories or nations. Trends include 0.15°C–0.18°C per decade in the tropical Western Pacific
5 (1953 to 2011), significant warming over the Caribbean (1901–2010) with trends of 0.19°C (0.28°C) per
6 decade in daily maximum (minimum temperatures) (1961–2010) and in La Réunion of 0.18°C per decade
7 (1968–2019). There are fewer significant trends in precipitation in these regions though several locations in
8 the Caribbean have detectable decreasing trends (high confidence), in part attributable to anthropogenic
9 forcing (limited evidence). Also, it is likely that drying has occurred since the mid-20th century in some parts
10 of the Western Indian Ocean, and in the Pacific poleward of 20° latitude in both the northern and southern
11 hemispheres.
12
13 It is very likely that Small Island regions will continue to warm in the coming decades at a level slightly
14 lower than the global mean. Small Island regions in the Western and equatorial Pacific, North Indian and
15 Southern Ocean are likely to be wetter in the future and those in the Caribbean, parts of the Atlantic and
16 West Indian Oceans and the southern sub-tropical and eastern Pacific Ocean drier.
17
18
19 [START CROSS-CHAPTER BOX ATLAS.2 HERE]
20
21 Cross-Chapter Box Atlas.2: Climate information relevant to water resources in Small Islands
22
23 Coordinators: Tannecia Stephenson (Jamaica), Faye Abigail Cruz (Philippines)
24 Contributors: Donovan Campbell (Jamaica), Subimal Ghosh (India), Rafiq Hamdi (Belgium), Mark Hemer
25 (Australia), Richard G. Jones (UK), James Kossin (USA), Simon McGree (Australia/Fiji), Blair Trewin
26 (Australia), Sergio M. Vicente-Serrano (Spain)
27
28 Constructing regional climate information for Small Islands involves synthesis from multiple sources. This
29 cross-chapter box presents information relevant to water resources, drawing on several chapters in AR6 and
30 Section Atlas.10. It introduces the context and current evidence base followed by an assessment of trends
31 and projections in rainfall, temperature and sea levels across Small Islands and highlight key findings.
32
33 Regional context
34
35 Small Islands are predominantly located in the Pacific, Atlantic and Indian Oceans, and in the Caribbean
36 (Nurse et al., 2014; Shultz et al., 2019). They are characterized by their small physical size, being surrounded
37 by large ocean expanses, vulnerability to natural disasters and extreme events and relative isolation (Nurse et
38 al., 2014) (Section 12.4.7, Section Atlas.10, Glossary). These and nearby larger islands (e.g., Madagascar,
39 Cuba) are often water-scarce with low water volumes due to increasing demand (from population growth and
40 tourism), aging and poorly designed infrastructure (Burns, 2002) and decreasing supply (from pollution,
41 changes in precipitation patterns, drought, saltwater intrusion, regional sea level rise, inadequate water
42 governance (Belmar et al., 2016; Mycoo, 2018) and competing and conflicting uses (Cashman, 2014;
43 Gheuens et al., 2019) (Section 8.1.1.1). In the Caribbean, groundwater is the main freshwater source and
44 depends strongly on rainfall variability (Post et al., 2018) while rain, ground or surface water are the primary
45 sources for the Pacific islands depending on island type (volcanic or atoll), size and quality of groundwater
46 reserves (Burns, 2002). Groundwater pumping and increasing sea levels also affect water availability by
47 increasing the salinity of the aquifer (e.g., Bailey et al., 2015, 2016) thus reinforcing negative drought effects
48 from reduced rainfall and increased evaporative demand from higher temperatures. For example, in 54% of
49 the Marshall Islands, groundwater is highly vulnerable to droughts (Barkey and Bailey, 2017).
50
51 The climate of Small Islands and findings from previous IPCC assessments
52
53 Intraseasonal to interannual rainfall in the Caribbean and in the Indian and Pacific Ocean is influenced by the
54 trade winds, the passage of tropical cyclones (TCs), Madden-Julian Oscillation (MJO), easterly waves,
55 migrations of the Inter-Tropical Convergence Zone (ITCZ) and the North Atlantic Subtropical High (NASH)
Do Not Cite, Quote or Distribute Atlas-97 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 for the Caribbean, the South Pacific Convergence Zone (SPCZ) and western North Pacific summer monsoon
2 for the Pacific and the South Asian monsoons for the Indian Ocean. The relevant dominant modes of climate
3 variability (Annex IV, Section 8.3.2.9) are El Niño-Southern Oscillation (ENSO) and the Indian Ocean
4 Dipole (IOD) which have been associated with extreme events in the islands (Stephenson et al., 2014; Kruk
5 et al., 2015; Frazier et al., 2018)(Annex IV). The modes of climate variability are modulated by Pacific
6 Decadal Variability (PDV), Interdecadal Pacific Oscillation (IPO) and Atlantic Multidecadal Variability
7 (AMV). These modes show no sustained trend since the late 19th century (high confidence) (Section 2.4).
8
9 The AR5 WGI reports observed temperature increases of 0.1°C–0.2°C per decade in Pacific Islands with
10 these trends very likely to continue under high emissions, and projects a drier rainy season for many islands
11 in the southwest Pacific (Christensen et al., 2013). AR5 WGII reports rainfall reductions over the Caribbean,
12 increases over the Seychelles, streamflow reductions over the Hawaiian Islands and salt-water intrusion into
13 groundwater reserves in Pacific Islands resulting from storm surges and high tides (Nurse et al., 2014).
14 SROCC (IPCC, 2019a) finds very high confidence that global mean sea level rise has accelerated in recent
15 decades which has exacerbated extreme sea level events and flooding (high confidence). It will continue to
16 rise with consequent increases in extreme levels so that the historical one-in-a-century extreme local sea
17 level will become an annual event by the end of the century under all RCP scenarios (high confidence). In
18 particular, many Small Islands are projected to experience historical centennial events at least annually by
19 2050 under RCP2.6, RCP4.5 and RCP8.5 emissions. The proportion of Category 4 and 5 TCs and associated
20 precipitation rates along with their average intensity are projected to increase with a 2°C global temperature
21 rise which will further increase the magnitude of resultant storm surges and flooding. The SROCC cross-
22 chapter box on Low-lying Islands and Coasts (Magnan et al., 2019) focused on sea level rise and oceanic
23 changes and their impacts, therefore the assessment presented here on climate changes relevant to water
24 resources, including precipitation and temperature, is complementary.
25
26 Observations and attribution of changes
27
28 Cross-Chapter Box Atlas.2: presents an overview of observed subregional trends relevant to water resources
29 in some Small Islands and island regions largely from 1951. Some general observed climate trends include
30 higher magnitude and frequency of temperatures including warm extremes (medium to high confidence;
31 Table 11.7, Sections 12.4.7.1, Atlas.10.2), declines in high intensity rainfall events (low to medium
32 confidence; Table 11.7), regional sea level rise with strong storm surge and waves resulting in increased
33 coastal flood intensity (high confidence, Section 12.4.7.4, Section Atlas.10.2), and increased intensity and
34 intensification rates of tropical cyclones at global scale (medium confidence, Sections 11.7.1.2, 12.4.7.3) and
35 ocean acidification (virtually certain, Chapters 2, 6 and 9, Section Atlas.3.2).
36
37 No significant long-term trends are observed for annual Caribbean rainfall over the 20th century (low
38 confidence; Section Atlas.10.2). Over the western Pacific, generally decreasing but non-significant trends are
39 noted in annual total rainfall from 1961 to 2011 (low confidence; Section Atlas.10.2; Table 11.5). June-July-
40 August (JJA) rainfall over the Caribbean shows some drying tendencies that may be linked to the combined
41 effect of warm ENSO events and a positive NAO phase (Giannini et al., 2000; Méndez-Lázaro et al., 2014;
42 Fernandes et al., 2015b), or to warm ENSO events and a positive PDV (Maldonado et al., 2016). The work
43 of Herrera et al. (2018) however suggests that anthropogenic influences may also be possible though
44 proposed mechanisms to date have not decoupled the influence of anthropogenic trends versus natural
45 decadal variability (Vecchi et al., 2006; Vecchi and Soden, 2007; DiNezio et al., 2009).
46
47 Southern hemisphere subtropical Pacific June–November drying has been associated with intensification of
48 the subtropical ridge and associated declines in baroclinicity (Whan et al., 2014). Austral summer drying in
49 the southwest French Polynesia subregion has been linked with increased greenhouse-gas and ozone changes
50 (Fyfe et al., 2012). The southern hemisphere jet stream has likely shifted polewards (Section 2.3.1.4.3) which
51 is attributed largely to a trend in the Southern Annular Mode (Section 3.7.2).
52
53 These assessments are constrained by limited availability of observational datasets and of scientific studies.
54 Assessment of observed climate change for Small Islands is often constrained by low station density (Ryu
55 and Hayhoe, 2014; Jones et al., 2016c), short periods of record, digitization requirements or data sharing
Do Not Cite, Quote or Distribute Atlas-98 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 limitations (Jones et al., 2016c), availability of metadata (McGree et al., 2014; Stephenson et al., 2014; Jones
2 et al., 2016b), challenges in some gridded product representations of variability, for example for complex
3 topography (Frazier and Giambelluca, 2017), and challenges characterizing the impact of vertical land
4 motion on sea level rise (Wöppelmann and Marcos, 2016) (see also Section Atlas.10.2).
5
6
7 [START CROSS-CHAPTER BOX ATLAS.2, TABLE 1 HERE]
8
9 Cross-Chapter Box Atlas.2, Table 1: Summary of observed trends for Small Island regions. SLR = sea level rise; TC
10 = tropical cyclone; SPCZ = South Pacific Convergence Zone.
11
Region Subregion Temperature Rainfall Other
Caribbean Low confidence in drought intensity
increasing over 2013–2016 (Herrera
and Ault, 2017; Herrera et al., 2018)
Jamaica, Low confidence in declining JJA No attributable JJA
Cuba, rainfall (CSGM, 2012) and rainfall trends 1951–
Puerto Rico decreasing trend Puerto Rico 1955– 2010 (Knutson and
2009 (Méndez-Lázaro et al., 2014). Zeng, 2018)
Mixed trends 1980–2010 (Cavazos et
al., 2020);
Eastern Medium confidence Low confidence in an increase in Medium confidence in
Caribbean in increased periods of drought since 1999 (Van SLR of 1–2.5 mm yr–1
frequency of hot Meerbeeck, 2020) since 1950 (Van
extremes Meerbeeck, 2020)
Pacific Midway- High confidence in Medium confidence in rainfall Medium confidence in
Hawaiian the increase in mean decreasing since 1920, drought relative SLR of 2.1
Islands temperature since frequency and severity increasing mm yr–1 (Mokuoloe Is.
1917 and stronger since 1951 and exceptional aridity and Honolulu, Oahu
increase in minimum since 2008; (McGree et al., 2016; Is.) over 1993–2017
temperature since Frazier and Giambelluca, 2017; Luo
1905 (Keener et al., et al., 2020b).
2018; McKenzie et
al., 2019; Kagawa- Low confidence in extreme rainfall
Viviani and increasing (Kruk et al., 2015)
Giambelluca, 2020)
Northwest High confidence in Low confidence in JJA and SON total Low confidence in
Tropics the increase in mean and extreme rainfall decreasing, decrease in total TC
and extreme increasing drought in east Micronesia numbers. Depends on
temperature and marginal increase in rainfall for dataset/period (Choi
at most locations western islands since 1951 (Kruk et and Cha, 2015; Lee et
since 1951 (Whan et al., 2015; McGree et al., 2019) al., 2020)
al., 2014; McGree et
al., 2019) Medium confidence in
relative SLR of 2.8
mm yr–1 (Majuro,
Marshall Is.) over
1994–2015 (Ford et
al., 2018)
Equatorial Low confidence in increasing annual Low to medium
Pacific and JJA extreme rainfall, decreasing confidence in relative
consecutive dry days in the central SLR of 5.3 (Nauru),
region since 1951 (McGree et al., 0.8 (Kanton, Kiribati)
Do Not Cite, Quote or Distribute Atlas-99 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
2019) and increasing DJF total mm yr–1 over 1993–
rainfall (BOM and CSIRO, 2014) 2015 (Albrecht et al.,
2019; Martínez-
Low confidence in decreasing SON Asensio et al., 2019)
total rainfall, increasing JJA and
SON extreme rainfall and fewer
consecutive wet days in western
region since 1951 (BOM and CSIRO,
2014; McGree et al., 2019)
Southwest Low confidence in change in mean Medium confidence in
SPCZ and extreme rainfall at most locations decrease in total TC
since 1951 (Keener et al., 2012; numbers and low
McGree et al., 2016, 2019) confidence in decrease
Northeast Low confidence in change in mean in numbers of intense
SPCZ and extreme rainfall at most locations TCs since 1981
since 1951 (BOM and CSIRO, 2014; (Kuleshov et al.,
McGree et al., 2016, 2019) 2020).
Southern Medium confidence in annual, JJA Low confidence in shift
Subtropics and SON total and extreme rainfall in mean SPCZ position
decreasing and increasing drought since 1911 (Salinger et
frequency in western region since al., 2014)
1951 (Jovanovic et al., 2012; McGree
et al., 2016, 2019) Medium confidence in
increase in relative
Low confidence in annual, SON, DJF, SLR of 1.7–7.7 mm yr–
MAM total and extreme rainfall 1
across southern
decreasing, increases in drought, JJA Pacific Islands over
rain days and consecutive dry days in period 1993–2015
Southwest French Polynesia since (Martínez-Asensio et
1951 (McGree et al., 2016, 2019) al., 2019)
Western Mauritius Warming of 1.2°C Rainfall decrease of 8% over 1951– Relative SLR at 5.6
Indian over 1951–2016 2016 (MESDDBM, 2016) mm yr–1 over 2007–
Ocean (MESDDBM, 2016) 2016 (MESDDBM,
2016)
La Réunion Temperature increase Rainfall decrease of 1.2% per decade
0.18°C per decade over 1961–2019 (Météo-France,
over 1968–2019 2020)
(Météo-France,
2020)
Maldives Generally warming Generally weak, non-significant SLR of 2.9–3.7 mm yr–
1
trends from the 1970s rainfall trends over 1967–2012 over 1991–2012
to 2012 (MEE, 2016) (MEE, 2016) (MEE, 2016)
1
2 [END CROSS-CHAPTER BOX ATLAS.2, TABLE 1 HERE]
3
4
5 Information on future climate changes
6
7 Small Islands will very likely continue to warm this century, though at a rate less than the global average
8 (Figure Atlas.28), with consequent increased frequency of warm extremes for the Caribbean and Western
9 Pacific islands, and heatwave events for the Caribbean (high confidence) (Table 11.7). Annual and JJA
10 rainfall declines are likely for some Indian and southern Pacific regions with drying over southern French
Do Not Cite, Quote or Distribute Atlas-100 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Polynesia (attributed partially to greenhouse-gas increases) and farther east clearly evident in CMIP5 and
2 CMIP6 projections (Figure Atlas.28) (high confidence). See also Section Atlas.10.4.
3
4 Rainfall is very likely to decline over the Caribbean, in the annual mean and especially in JJA, with a
5 stronger and more coherent signal in CMIP6 compared to CMIP5 (Figure Atlas.28, Interactive Atlas) and
6 reductions of 20–30% by the end of the century under high future emissions (SSP5-8.5). This JJA drying has
7 been linked to a future strengthening of the Caribbean Low Level Jet (CLLJ) (Taylor et al., 2013b), a
8 westward expansion and intensification of the NASH, stronger low-level easterlies over the region, a
9 southwardly placed eastern Pacific ITCZ (Rauscher et al., 2008), and changing dynamics due to increased
10 greenhouse-gas concentrations (Li et al., 2012b) (very high confidence). Projections from 15 GCM and two
11 RCM experiments for 2080–2089 relative to 1970–1989 were for a generally drier Caribbean and a robust
12 summer drying (Karmalkar et al., 2013). More recent downscaling studies (e.g., Taylor et al., 2018; Vichot-
13 Llano et al., 2021) also project a drier Caribbean and longer dry spells (Van Meerbeeck, 2020).
14
15 Sea level rise is very likely to continue in all Small Island regions (Figure Atlas.28, see also Sections 9.6.3.3
16 and 12.4.7.4) and its the effects will be compounded by TC surge events. In general, the most intense TCs
17 are likely to intensify and produce more flood rains with warming, however detailed effects of climate
18 change on TCs will vary by region (Knutson et al., 2019)(Section 11.7.1). Bailey et al. (2016) projected a
19 20% decline in groundwater availability by 2050 in Coral Atoll islands of the Federal States of Micronesia
20 and stressed that under higher sea level rise the decrease could be higher than 50% due to marine water
21 intrusion into aquifers, as well as drought events.
22
23 Summary of information distilled from multiple lines of evidence
24
25 It is very likely that most Small Islands have warmed over the period of instrumental records. The clearest
26 precipitation trend is a likely decrease in JJA rainfall over the Caribbean since 1950. There is limited
27 evidence and low agreement for the cause of the observed drying trend, whether it is mainly caused by
28 decadal-scale internal variability or anthropogenic forcing, but it is likely that it will continue over coming
29 decades. It is likely that drying has occurred since the mid-20th century in some parts of the Pacific poleward
30 of 20° latitude in both the northern and southern hemispheres and that these will continue over coming
31 decades. Rainfall trends in most other Pacific and Indian Ocean Small Islands are mixed and largely non-
32 significant. It is very likely that sea levels will continue to rise in all Small Island regions and this will result
33 in increased coastal flooding with the potential to increase salt-water intrusion into aquifers in Small Islands.
34
35 Whilst this assessment demonstrates that the climate of Small Islands has and will continue to change in
36 diverse ways, constructing climate information for Small Islands is challenging. This is due to observational
37 issues, incomplete understanding of some modes of variability and their representation by climate models
38 and the lack of availability of large ensembles of regional climate model simulations and limited studies to
39 decouple internal variability and anthropogenic influences.
40
41 [END CROSS-CHAPTER BOX ATLAS.2 HERE]
42
43
44 Atlas.11 Polar regions
45
46 The assessment in this section focuses on changes in average temperature, precipitation (rainfall and snow)
47 and surface mass balance over the polar regions, Antarctica and the Arctic, including the most recent years
48 of observations, updates to observed datasets, the consideration of recent studies using CMIP5 simulations
49 and those using CMIP6 and CORDEX simulations. Findings are presented for West Antarctica (WAN), East
50 Antarctica (EAN) and three Arctic regions, Arctic Ocean (ARO), Greenland/Iceland (GIC) and Russian
51 Arctic (RAR) (Figure Atlas.29) with some reference also to North-Eastern North America (NEN), North-
52 Western North America (NWN) and Northern Europe (NEU), which are covered more extensively in
53 Sections Atlas.9 and Atlas.8 respectively. Subregional changes are discussed when relevant, for example the
54 Antarctic Peninsula (AP) as a subregion of WAN. The Southern Ocean (SOO) region is assessed in Chapter
55 9 with changes in climatic impact-drivers assessed in Chapter 12 (12.4.9, Table 12.11) and some extremes in
Do Not Cite, Quote or Distribute Atlas-101 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Chapter 11 (Table 11.7–9 for RAR). Chapter 9 provides an overall assessment of the ice sheet processes and
2 changes, as part of the cryosphere, ocean and sea level change assessment.
3
4
5 [START FIGURE ATLAS.29 HERE]
6
7 Figure Atlas.29: Regional mean changes in annual mean surface air temperature and precipitation relative to the
8 1995–2014 baseline for the reference regions in Arctic and Antarctica (warming since the 1850–
9 1900 pre-industrial baseline is also provided as an offset). Bar plots in the left panel of each region
10 triplet show the median (dots) and 10th–90th percentile range (bars) across each model ensemble for
11 annual mean temperature changes for four datasets (CMIP5 in intermediate colours; subset of CMIP5
12 used to drive CORDEX in light colours; CORDEX overlying the CMIP5 subset with dashed bars; and
13 CMIP6 in solid colours); the first six groups of bars represent the regional warming over two time
14 periods (near-term 2021–2040 and long-term 2081–2100) for three scenarios (SSP1-2.6/RCP2.6,
15 SSP2-4.5/RCP4.5, and SSP5-8.5/RCP8.5), and the remaining bars correspond to four global warming
16 levels (GWL: 1.5°C, 2°C, 3°C, and 4°C). The scatter diagrams of temperature against precipitation
17 changes display the median (dots) and 10th–90th percentile ranges for the above four warming levels
18 for December-January-February (DJF; middle panel) and June-July-August (JJA; right panel),
19 respectively; for the CMIP5 subset only the percentile range of temperature is shown, and only for
20 3°C and 4°C GWLs. Changes are absolute for temperature (in °C) and relative (as %) for
21 precipitation. See Section Atlas.1.3 for more details on reference regions (Iturbide et al., 2020) and
22 Section Atlas.1.4 for details on model data selection and processing. The script used to generate this
23 figure is available online (Iturbide et al., 2021) and similar results can be generated in the Interactive
24 Atlas for flexibly defined seasonal periods. Further details on data sources and processing are
25 available in the chapter data table (Table Atlas.SM.15).
26
27 [END FIGURE ATLAS.29 HERE]
28
29
30 Atlas.11.1 Antarctica
31
32 Atlas.11.1.1 Key features of the regional climate and findings from previous IPCC assessments
33
34 Atlas.11.1.1.1 Key features of the regional climate
35 The Antarctic region, covered by an ice sheet and surrounded by the Southern Ocean, is characterized by
36 polar climate. It is the coldest, windiest and driest continent on Earth and plays a pivotal role in regulating
37 the global climate and hydrological cycle. Antarctica has a mean temperature of –35°C (Lenaerts et al.,
38 2016) and receives 171 mm yr–1 water equivalent of snowfall (north of 82°S, estimate based on satellite
39 measurements during 2006–2011) (Palerme et al., 2014). Precipitation in Antarctica occurs mostly in the
40 form of snowfall and diamond dust, with sporadic coastal rainfall during the summer over the Antarctic
41 Peninsula and sub-Antarctic islands. Drizzle events sometimes occur during warm air intrusions (Nicolas et
42 al., 2017) at relatively low temperatures (Silber et al., 2019). Precipitation constitutes the largest component
43 of the surface mass balance (SMB), which also includes sublimation (from the surface or drifting snow),
44 meltwater runoff and redistribution by wind (Lenaerts et al., 2019). SMB can be considered as a proxy of
45 precipitation if averaged over an annual cycle (Gorodetskaya et al., 2015; Bracegirdle et al., 2019).
46 Precipitation and SMB exhibit spatial and temporal variability controlled by atmospheric large-scale low-
47 pressure systems and moisture advection from lower latitudes. SMB is an important component of the total
48 ice-sheet mass balance (Section 9.4.2.1). The Antarctic contribution to sea level results from the imbalance
49 between net snow accumulation and ice discharge into the ocean (Box 9.1). Ice shelves buttress the ice sheet
50 and are influenced by oceanic and atmospheric drivers (Box 9.1).
51
52 Antarctic climate variability is influenced by the Southern Annular Mode (SAM) and regionally by other
53 modes, including ENSO, Pacific-South American Pattern, Pacific Decadal Variability, Indian Ocean Dipole
54 and Zonal Wave 3 (Annex IV). Climate change in Antarctica and the Southern Ocean is influenced by
55 interactions between the ice sheet, ocean, sea ice, and atmosphere (Meredith et al., 2019) (Sections 9.2.3.2,
56 9.3.2 and 9.4.2). In addition to Chapter 9, Antarctica is discussed across the report: global climate links
57 (Chapters 2 and 10), attribution (Chapter 3), global water cycle (Chapter 8), extremes (Chapter 11), and
Do Not Cite, Quote or Distribute Atlas-102 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 climatic impact-drivers (Chapter 12).
2
3
4 Atlas.11.1.1.2 Findings from previous IPCC assessments
5 AR5 (Vaughan et al., 2013) reported warming over Antarctica since 1950s, mostly over AP and WAN,
6 attributed to the positive trend in the SAM. These trends in the Antarctic temperature were given low
7 confidence due to substantial multi-annual to multi-decadal variability, as well as uncertainties in magnitude
8 and spatial trend structure. AR5 reported low confidence that anthropogenic forcing has contributed to the
9 temperature change in Antarctica. AR5 highlighted a large interannual variability in snow accumulation with
10 no significant trend since 1979 around Antarctica, and high confidence in the overall mass loss from
11 Antarctica, accelerated since 1990s.
12
13 In this and the following paragraphs findings are from SROCC (Meredith et al., 2019) unless otherwise
14 stated. Warming trends were reported over parts of WAN with record surface warmth over WAN during the
15 1990s compared to the past 200 years, and AP surface melting intensifying since the mid-20th century. No
16 significant temperature trends were reported over EAN and there was low confidence in both WAN and EAN
17 trend estimates due to sparse in situ records and large interannual to interdecadal variability. In the AP,
18 concomitant increase in temperature and foehn winds due to positive SAM caused increased surface melting
19 over the Larsen ice shelves (medium confidence). Strong warming between the mid-1950s and the late 1990s
20 led to the collapse of the Larsen B ice shelf in 2002, which had been intact for the 11,000 years (medium
21 confidence).
22
23 Snowfall increased over the Antarctic ice sheet over AP and WAN, offsetting some of the 20th century sea
24 level rise (medium confidence). Longer records suggest either a decrease in snowfall over the Antarctic ice
25 sheet over the last 1000 years or a statistically negligible change over the last 800 years (low confidence).
26
27 Recent warming in the AP and consequent ice-shelf collapse are likely linked to anthropogenic ozone and
28 greenhouse-gas forcing via the SAM and anthropogenically-driven Atlantic sea surface. Also, there is high
29 confidence in the influence of tropical sea surface temperature on the Antarctic temperature and Southern
30 Hemisphere mid-latitude circulation, as well as the SAM. There is medium agreement but limited evidence
31 of an anthropogenic forcing effect on Antarctic ice-sheet mass balance (low confidence) and partitioning
32 between natural and human drivers of atmospheric and ocean circulation changes remains very uncertain.
33
34 In AR5 Church et al. (2013) gave medium confidence in model projections of a future Antarctic SMB
35 increase, implying a negative contribution to global mean sea level rise, consistent with a projection of
36 significant Antarctic warming. Church et al. (2013) also gave high confidence to the relationship between
37 future temperature and precipitation increases in Antarctica on physical grounds and from ice core evidence.
38 In Meredith et al. (2019), the total mass balance projections derived from ice sheet models were reported
39 without separating the SMB though projections were reported of increased precipitation and continued
40 strengthening of the westerly winds in the Southern Ocean.
41
42
43 Atlas.11.1.2 Assessment and synthesis of observations, trends and attribution
44
45 Figure Atlas.30 (Antarctic map inset) shows near surface air temperature trends for 1957-2016 and 1979-
46 2016 at the stations where observations are available for at least 50 years and the detected trends have
47 statistical significance of at least 90% according to the most recent (after SROCC) studies (Jones et al., 2019;
48 Turner et al., 2020). It is very likely that the western and northern AP has been warming significantly since
49 1950s (0.49 ± 0.28°C per decade during 1957–2016 and 0.46 ± 0.15°C during 1951–2018 at Faraday-
50 Vernadsky station; 0.29 ± 0.16°C per decade during 1957–2016 at Esperanza station), with no significant
51 trends reported in the eastern AP during the same period (Gonzalez and Fortuny, 2018; Jones et al., 2019;
52 Turner et al., 2020). Short-term cooling trends, strongest during austral summer, have been reported at AP
53 stations during 1999–2016 but the absence of warming and cooling at some stations during 1999–2016 is
54 consistent with natural variability and there is no evidence of a shift in the overall warming trend observed
55 since 1950s (Turner et al., 2016, 2020; Gonzalez and Fortuny, 2018; Jones et al., 2019; Bozkurt et al., 2020).
Do Not Cite, Quote or Distribute Atlas-103 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Significant warming at the Byrd station (0.29 ± 0.19°C per decade during 1957–2016) confirms and extends
2 earlier trend estimates (0.42 ± 0.24°C per decade during 1958–2010) and is representative of the entire WAN
3 warming (0.22 ± 0.12°C per decade from 1958 to 2012 averaged over WAN excluding AP, medium
4 confidence due to lack of observations) (Bromwich et al., 2013, 2014; Jones et al., 2019). WAN and AP
5 show statistically significant warming in the HadCRUTv5 observational dataset (Figure 2.11b). There is a
6 high confidence in the long-term warming trend at the AP and WAN also at the century scale based on
7 reconstructions (Zagorodnov et al., 2012; Stenni et al., 2017; Lyu et al., 2020) confirming the trends
8 estimated by earlier studies assessed in the SROCC (Meredith et al., 2019). The century-scale warming trend
9 in the AP is very likely an emerging signal compared to natural variability, while the WAN warming trend
10 falls in the high end of century-scale trends over the last 2000 years (Stenni et al., 2017) (medium
11 confidence).
12
13 In EAN, during 1957-2016, three stations showed significant warming (Scott base 0.22 ± 0.15,
14 Novolazarevskaya 0.13 ± 0.09 and Vostok 0.15 ± 0.13°C per decade), while other stations with long-term
15 observations indicated no statistically significant trends (Figure Atlas.30). During 1979-2016, three coastal
16 stations showed cooling, while at South Pole a warming trend was detected, increasing to 0.61±0.34 °C per
17 decade during 1989-2018) (Jones et al., 2019; Clem et al., 2020; Turner et al., 2020) (Figure Atlas.30). The
18 century-scale warming in Queen Maud Land Coast based on the ice-core reconstructions is within the range
19 of centennial internal variability (Stenni et al., 2017).
20
21 While a trend towards a positive phase of the SAM since the 1970s likely explains a significant part of the
22 warming at the northern AP, it had a cooling effect on the continental WAN and EAN (particularly strong in
23 DJF, Table.Atlas.1). Warming in western AP and over WAN during 1957–2016 (Figure.Atlas.30) and
24 through to 2020 (Figure 2.11) is likely due to significant contribution of other factors, such as tropical Pacific
25 forcing through PDV, ENSO, Amundsen Sea Low position/strength and also anthropogenic climate change
26 (Jones et al., 2019; Scott et al., 2019; Wille et al., 2019; Donat-Magnin et al., 2020; Turner et al., 2020).
27 Since SROCC, new studies confirmed the influence of foehn wind and cloud radiative forcing on Larsen C
28 surface melt (Elvidge et al., 2020; Gilbert et al., 2020; Turton et al., 2020). In the WAN, summer surface-
29 melt occurrence over ice shelves may have increased since the late 2000s (Scott et al., 2019). It is likely that
30 increased meltwater ponding and resulting hydrofracturing have been important mechanisms of the rapid
31 disintegration of the Larsen B ice shelf (Banwell et al., 2013; MacAyeal and Sergienko, 2013; Robel and
32 Banwell, 2019). Ice shelf disintegration and relevant processes are discussed in Sections 9.4.2.1 and 9.4.2.3.
33
34 Direct observations of snowfall in Antarctica using traditional gauges are highly uncertain and records from
35 precipitation radars (Gorodetskaya et al., 2015; Grazioli et al., 2017; Scarchilli et al., 2020) are not long
36 enough to assess trends. Estimates of precipitation and SMB are largely model-based due to the paucity of in
37 situ observations in Antarctica (Lenaerts et al., 2019; Hanna et al., 2020). Antarctic SMB is dominated by
38 precipitation and removal by sublimation with very small amounts of melt mostly important only on the ice
39 shelves. Climate models and satellite records (IMBIE team et al., 2018; Rignot et al., 2019; Mottram et al.,
40 2021) suggest that strong interannual variability of Antarctic-wide SMB over the satellite period currently
41 masks any existing trend (Figure Atlas.30) in spite of a possible ozone depletion-related precipitation
42 increase over the 1991–2005 period (Lenaerts et al., 2018). No significant Antarctic-wide SMB trend
43 continent-wide SMB trend is inferred since 1979 (IMBIE team et al., 2018; Medley and Thomas, 2019).
44 While ice core reconstructions show a significant increase in the western AP SMB since 1950s (Thomas et
45 al., 2017; Medley and Thomas, 2019; Wang et al., 2019), this trend is not reproduced by regional climate
46 models or reanalyses used to drive them (Figure Atlas.30) (van Wessem et al., 2016; Wang et al., 2019).
47
48 According to the ice core reconstructions, SMB over WAN (including AP) has likely increased during the
49 20th century with trends of 5.4 ± 2.9 Gt yr–1 per decade (1900–2010) (Wang et al., 2019) mitigating global
50 mean sea level rise by, respectively, 0.28 ± 0.17 mm per decade (WAN excluding AP, during 1901–2000)
51 and 0.62 ± 0.17 mm per decade (AP, during 1979–2000) (Medley and Thomas, 2019). Significant spatial
52 heterogeneity in SMB trends has been observed over AP and WAN:
53 • western AP has likely experienced a significant increase in SMB beginning around 1930 and
54 accelerating during 1970–2010, which is outside of the natural variability range of the past 300 years
55 (Thomas et al., 2017; Medley and Thomas, 2019; Wang et al., 2019);
Do Not Cite, Quote or Distribute Atlas-104 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 • eastern AP has no significant SMB trends during the same period (low confidence, observations
2 limited to one ice core and large interannual variability) (Thomas et al., 2017; Engel et al., 2018);
3 • overall WAN SMB (excluding AP) was stable during 1980–2009 but exhibited high regional
4 variability (Medley et al., 2013): significant increases (5–15 mm per decade during 1957–2000) to
5 the east of the West Antarctic ice sheet divide and significant decrease (–1 to –5 mm per decade
6 during 1901–1956, and –5 to –15 mm per decade during 1957–2000) to the west (Medley and
7 Thomas, 2019; Wang et al., 2019).
8
9 The SMB of EAN increased during the 20th century which mitigated global mean sea level rise by 0.77 ±
10 0.40 mm per decade during 1901–2000 (Medley and Thomas, 2019) (medium confidence). EAN SMB has
11 been increasing at a much lower rate since 1979 as shown by observations, while regional climate models
12 show strong interannual variability masking any trend (Medley and Thomas, 2019; Rignot et al., 2019)
13 (Figure Atlas.30) (low confidence due to limited observations). EAN SMB changes during the 20th century
14 and recent decades showed large spatial heterogeneity:
15 • With significant increases likely in Queen Maud Land (QML): 5.2 ± 3.7% per decade during 1920–
16 2011 measured in ice core near the Kohnen station (Medley et al., 2018), an increase on the plateau
17 (Altnau et al., 2015), and stable conditions during 1993–2010 along the annual stake line from
18 Syowa (coast) to Dome F (plateau) (Wang et al., 2015b); increases during 1911–2010 (Thomas et
19 al., 2017) with anomalously high SMB observed in 2009 and 2011 (Boening et al., 2012; Lenaerts et
20 al., 2013; Gorodetskaya et al., 2014);
21 • increases in Wilkes Land and Queen Mary Land during 1957–2000 (Thomas et al., 2017; Medley
22 and Thomas, 2019) (low confidence due to limited observations and strong spatial variability);
23 • a likely stable SMB in the interior of the East Antarctic plateau during the 1901–2000 period and the
24 last decades (Thomas et al., 2017; Medley and Thomas, 2019);
25 • stable in Adelie Land (annual stake line during 1971–2008) (Agosta et al., 2012) (low confidence
26 due to limited evidence).
27
28 Regional trends of the last 50 years (1961–2010) and 100 years (1911–2010) are within centennial variability
29 of the past 1000 years, except for coastal QML (unusual 100-year increase in accumulation) and for coastal
30 Victoria Land (unusual 100-year decrease in accumulation) (Thomas et al., 2017). Nevertheless, the current
31 EAN SMB is not unusual compared to the past 800 years (Frezzotti et al., 2013).
32
33 The geographic pattern of accumulation changes since the 1950s bears a strong imprint of a trend towards a
34 more positive phase of the SAM (e.g., Medley and Thomas, 2019), which could be linked to ozone depletion
35 (Lenaerts et al., 2018) or large-scale atmospheric warming (Frieler et al., 2015; Medley and Thomas, 2019).
36 More evidence has emerged showing the importance of the Pacific-South American Pattern, ENSO and
37 Pacific Ocean convection, and large-scale blocking causing warm-air intrusions and both extreme
38 precipitation and melt events, responsible for large interannual SMB variability (high confidence)
39 (Gorodetskaya et al., 2014; Bodart and Bingham, 2019; Scott et al., 2019; Turner et al., 2019; Wille et al.,
40 2019; Adusumilli et al., 2021). This strengthens evidence for an important connection between Antarctic
41 climate and tropical sea surface temperature stated by SROCC (Meredith et al., 2019). Section 3.4.3 and
42 SROCC (Meredith et al., 2019) provide a discussion of attribution of Antarctic ice sheet changes.
43
44
45 Atlas.11.1.3 Assessment of model performance
46
47 This section provides evaluation of atmospheric global and regional climate models, including reanalyses.
48 Evaluation of the ice sheet models and relevant processes, including selection of the atmospheric models
49 used to drive ice sheet models, is given in Section 9.4.2.2.
50
51 One of the major systematic biases in CMIP5 and earlier GCMs was an equatorward bias in the latitude of
52 the Southern Hemisphere mid‐latitude westerly jet, which is significantly reduced in the CMIP6 ensemble
53 (Bracegirdle et al., 2020a). GCM Southern Ocean sea-ice biases are also of importance as they influence 21st
54 century temperature projections in Antarctica and simulation of present day temperatures are highly sensitive
55 to these biases (Agosta et al., 2015; Bracegirdle et al., 2015). A positive bias in near-surface temperature
Do Not Cite, Quote or Distribute Atlas-105 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 over the Antarctic plateau is seen in CMIP5 models (Lenaerts et al., 2016).
2
3 CMIP6 GCMs showed an improved representation of the Antarctic near-surface temperature compared to
4 CMIP5 but little improvement (maintaining positive bias) in Antarctic precipitation estimates (Palerme et al.,
5 2017; Roussel et al., 2020). An analysis of the 1850–2000 SMB mean, trends and interannual and spatial
6 variability suggests slightly worse agreement with ice-core-based reanalyses in CMIP6 than CMIP5 (Gorte
7 et al., 2020). Comparison of CMIP5 models with CloudSat satellite products and an ice-core-based SMB
8 reconstruction showed almost all the models overestimate current Antarctic precipitation, some by more than
9 100% (Palerme et al., 2017; Gorte et al., 2020). GCM simulations of surface snow-melt processes are of
10 variable quality, with extremely simple representatons, or non-existent (Agosta et al., 2015; Trusel et al.,
11 2015). Though most meltwater refreezes in the snowpack in current climate simulations this may be an issue
12 in the future climate simulations under global warming as run-off is projected to increase (Kittel et al., 2021).
13 Since CMIP5, representation of snow (Lenaerts et al., 2016) and stable surface boundary layers (Vignon et
14 al., 2018) has improved in some atmospheric GCMs. In one example, the CMIP6 model CESM2 simulation
15 of cloud and precipitation showed substantial improvements (Schneider et al., 2020) though surface melting
16 is still considerably overestimated compared to RCMs and satellite products (Trusel et al., 2015; Lenaerts et
17 al., 2016).
18
19 Assimilation of observations in reanalysis products yields realistic temperature patterns and seasonal
20 variations, with the recent ERA5 reanalysis showing improved performance compared to others for mean
21 and extreme temperature, wind and humidity, though a warm bias in the near-surface air temperatures
22 remains (Retamales-Muñoz et al., 2019; Tetzner et al., 2019; Dong et al., 2020; Gorodetskaya et al., 2020).
23 The ability of the reanalyses to simulate precipitation and SMB is more variable; they generally overestimate
24 the latter (Gossart et al., 2019; Roussel et al., 2020), but are well suited to provide atmospheric and sea
25 surface boundary conditions to drive RCMs.
26
27 Recent higher-resolution simulations covering the entire Antarctic ice sheet with a grid spacing of 12 to 50
28 km include five Polar-CORDEX RCMs – RACMO2 (van Wessem et al., 2018), MAR (Agosta et al., 2019;
29 Kittel et al., 2021), COSMO-CLM2 (Souverijns et al., 2019), HIRHAM5 (Lucas-Picher et al., 2012), and
30 MetUM (Walters et al., 2017; Mottram et al., 2021) – and one stretched-grid GCM – ARPEGE (Beaumet et
31 al., 2019). RCM simulations forced by ERA-Interim agree well with automatic weather station temperatures,
32 with high correlation (R2 > 0.9) and low bias (<1.5°C) except for high resolution HIRHAM5 (–2.1°C) and
33 MetUM (–3.4°C), which are not internally nudged models (Mottram et al., 2021). RCMs generally
34 underestimate the observed SMB but with biases lower than 20%, except for COSMO-CLM2 at lower
35 elevations (<1200 m) and HIRHAM5 and MetUM at higher elevations (>2200 m) (Mottram et al., 2021).
36 These RCM simulations lead to estimates of the grounded Antarctic ice sheet SMB ranging from 2133 Gt yr–
37 1
to 2328 Gt yr–1 when considering the four simulations compatible with the IMBIE2 Antarctic total mass
38 budget (IMBIE team et al., 2018; Mottram et al., 2021). However, the simulated spatial pattern of SMB
39 differs widely between models suggesting the importance of missing or under-represented processes in the
40 models, such as drifting-snow transport and sublimation (Agosta et al., 2019), cloud-precipitation
41 microphysical processes (van Wessem et al., 2018) and snowpack modelling (Mottram et al., 2021).
42 Comparisons of integrated SMB estimates between models are also complicated by different resolutions and
43 continental ice masks, with models showing large differences in the absolute SMB (Mottram et al., 2021) but
44 better agreement for SMB annual rates (Figure Atlas.30)
45
46 Finer resolution RCM studies demonstrate improved representation of precipitation and temperature
47 gradients (van Wessem et al., 2018; Bozkurt et al., 2020; Donat-Magnin et al., 2020; Elvidge et al., 2020),
48 and strength of katabatic winds (Bintanja et al., 2014; Souverijns et al., 2019) in coastal and mountainous
49 regions. Adequate representation of some processes is still lacking, including drifting snow, sublimation of
50 falling snow or the spectral dependency of snow albedo (Lenaerts et al., 2019). Non-hydrostatic regional
51 models, for example Polar-WRF, MetUM or HARMONIE-AROME at spatial resolutions up to 2 km further
52 improve regional RCM simulations, but are still often unable to resolve relevant feedbacks and foehn
53 processes (Grosvenor et al., 2014; Elvidge et al., 2015, 2020; Elvidge and Renfrew, 2016; King et al., 2017;
54 Turton et al., 2017; Bozkurt et al., 2018b; Hines et al., 2019; Vignon et al., 2019; Gilbert et al., 2020).
55
Do Not Cite, Quote or Distribute Atlas-106 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Existing uncertainties in the Antarctic climate representation by both GCMs and RCMs cause significant
2 spread in the future Antarctic climate and SMB projections (Gorte et al., 2020; Kittel et al., 2021). Run-time
3 bias adjustment in atmospheric GCMs (Krinner et al., 2019, 2020) (see also Cross-Chapter Box 10.2) has
4 been proposed to provide low-bias present and consistently-corrected future RCM forcing (reducing the need
5 for coupled model selection), which could be used directly for Antarctic climate projections (Krinner et al.,
6 2019).
7
8
9 [START FIGURE ATLAS.30 HERE]
10
11 Figure Atlas.30: (Upper panels) Time series of annual surface mass balance (SMB) rates (in Gt a –1) for Greenland ice
12 sheet and its regions (shown in the inset map) for the periods 1972–2018 (Mouginot et al., 2019) and
13 1980–2012 (Fettweis et al., 2020) using 13 different models. (Lower panels) Time series of annual
14 SMB rates (in Gt a–1) for the grounded Antarctic ice sheet (excluding ice shelves) and its regions
15 (shown in the inset map) for the periods 1979–2019 (Rignot et al., 2019) and 1980–2016 (Mottram et
16 al., 2021) using five Polar-CORDEX regional climate models. The Antarctic inset map also shows the
17 location of the stations discussed in section Atlas.11.1.2 where observations are available for at least
18 50 years. Colors indicate near surface air temperature trends for 1957-2016 (circles) and 1979-2016
19 (diamonds) statistically significant at 90% (Jones et al., 2019; Turner et al., 2020). Stations with an
20 asterisk (*) are where significance estimates disagree between the two publications. Further details on
21 data sources and processing are available in the chapter data table (Table Atlas.SM.15).
22
23 [END FIGURE ATLAS.30 HERE]
24
25
26 Atlas.11.1.4 Assessment and synthesis of projections
27
28 This section provides an assessment of projections in temperature, precipitation and SMB. See Section 9.4.2
29 for projected changes in the ice sheet total mass balance and relevant processes and see Section 4.3.1 (Table
30 4.2) and Section 4.5.1 for Antarctic temperature projections relative to other regions.
31
32 The Antarctic region is very likely to experience a significant increase in annual mean temperature and
33 precipitation by the end of this century under all emission scenarios used in CMIP5 and CMIP6 (Bracegirdle
34 et al., 2015, 2020b; Frieler et al., 2015; Lenaerts et al., 2016; Previdi and Polvani, 2016; Palerme et al.,
35 2017)(Figure Atlas.29). Ensemble means (and 10th–90th percentile ranges) of end-of-century (2081–2100)
36 projected Antarctic surface air temperature change from 35 CMIP6 models and relative to 1995–2014 are
37 1.2°C (0.5°C–2.0°C) for the SSP1-2.6 emissions scenarios, 2.3°C (1.3°C–3.4°C) for SSP2-4.5, 3.5°C
38 (2.0°C–5°C) for SSP3-7.0 and 4.4°C (2.8°C–6.4°C) for SSP5-8.5 (Interactive Atlas). Both temperature and
39 precipitation projections are characterised by a relatively large multi-model range (Figure Atlas.29,
40 Interactive Atlas). A strong regional variability is present with the projected changes over coastal Antarctica
41 not scaling linearly with global forcing. While continental mean temperatures are linearly related to global
42 mean temperatures in CMIP6 models, the relative increase in coastal temperatures are higher for low-
43 emissions scenarios due to stronger relative Southern Ocean warming and relatively stronger effects of ozone
44 recovery (Bracegirdle et al., 2020b). A higher multi-model average increase in temperature is projected by
45 CMIP6 models compared to CMIP5, with a 1.3°C higher mean Antarctic near-surface temperature at the end
46 of 21st century (Kittel et al., 2021). While similar median temperature changes are projected for WAN and
47 EAN, the former shows larger spread and higher projected temperature range in both CMIP5 and CMIP6
48 models and for all scenarios (Figure Atlas.29). CORDEX-Antarctica simulations show a mean and range in
49 the future temperature changes similar to the subset of CMIP5 models used to drive them for the RCP8.5
50 scenario and 1.5°C, 2°C and 3°C global warming levels (GWL) (Figure Atlas.29).
51
52 There is high confidence that projected future surface-air temperature increase over Antarctica will be
53 accompanied by precipitation increase (Figure Atlas.29). CMIP6 models show a similar or larger but more
54 constrained increase in precipitation (more models agreeing with larger precipitation increase) for the same
55 GWLs compared to CMIP5. For example, over WAN during JJA for 3°C GWL, CMIP6 and CMIP5 models
56 project a median 15% increase in precipitation with a 10th–90th percentile range of 7–25% in CMIP6
Do Not Cite, Quote or Distribute Atlas-107 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 models and of 3–24% in CMIP5. Average precipitation changes relative to 1995–2014 over WAN and EAN
2 are largely similar, and show projected increases for SSP2-4.5 (SSP5-8.5) of around 5% (5%) for 2021–
3 2040, 7% (10%) for 2041–2060, and 12% (25%) for 2081–2100 with smaller increases projected for SSP1-
4 2.6 emissions, reaching around 5% in 2081–2100. Regionally, the largest relative precipitation increase is
5 projected (under all scenarios) for the eastern part of WAN, the western AP, large parts of EAN plateau and
6 over coastal EAN within 0°E–90°E longitudinal sector (Interactive Atlas). The largest increase in absolute
7 precipitation amount is projected along the coastal regions, with the largest increase over coastal WAN and
8 the western AP, and is projected to be largely driven by the increase in maximum 5-day precipitation
9 (Interactive Atlas), which is in line with the dominant contribution of extreme snowfall events to the total
10 annual precipitation in the present Antarctic climate (Boening et al., 2012; Gorodetskaya et al., 2014; Turner
11 et al., 2020). Under all emission scenarios, the coastal precipitation increase corresponds to the snowfall
12 increase, except for the northern and central part of the western AP, where snowfall is projected to decrease
13 and rainfall to increase (similarly to the tendency towards increased precipitation, decreased snowfall and
14 increase in rainfall over the Southern Ocean) (Interactive Atlas).
15
16 From 2000 to 2100, the grounded Antarctic SMB is projected to mitigate sea level rise for RCP4.5 (RCP8.5)
17 by the following sea level equivalents (SLE), 0.03 ± 0.02 m (0.08 ± 0.04 m SLE) from 30 CMIP5 models
18 and for SSP2-4.5 (SSP5-8.5) by 0.03 ± 0.03 m SLE (0.07 ± 0.04 m SLE) from 24 CMIP6 models (Gorte et
19 al., 2020). Subsets or downscaling of CMIP AOGCMs lead to 21st century cumulative projections in the
20 range of 0.05 ± 0.03 m SLE for CMIP5 RCP8.5 and 0.08 ± 0.04 m SLE for CMIP6 SSP5-8.5 (Gorte et al.,
21 2020; Nowicki et al., 2020; Seroussi et al., 2020; Kittel et al., 2021). Use of model subsets reduces spread
22 leading to either lower or higher climate sensitivity in the Antarctic depending on the selection method. For
23 example, models selected by Gorte et al. (2020) based on SMB ice-core reconstruction from Medley and
24 Thomas (2019) tend to underestimate strongly winter sea ice area (Agosta et al., 2015; Roach et al., 2020)
25 and show reduced 21st century increase in Antarctic SMB compared to the full ensembles (Agosta et al.,
26 2015; Bracegirdle et al., 2015). A different subset of models is used for ISMIP6 (Section 9.4.2.3) which
27 gives a lower increase in Antarctic SMB than the full ensemble for CMIP5 but a larger increase for CMIP6.
28
29 Polar-CORDEX RCMs show higher variability in precipitation projections compared to CMIP5 models with
30 a similar spatial pattern of the areas with precipitation increase over continental Antarctica but with higher
31 local magnitude and also showing a larger increase over the Weddell Sea ice shelves (Interactive Atlas).
32 CMIP5 and CMIP6 models, bias-adjusted based on regional climate model simulations, showed that the
33 projected warming is expected to result in increased surface melting over the Antarctic ice shelves, with
34 meltwater runoff under RCP8.5 and SSP5-8.5 becoming larger than precipitation over ice shelves over the
35 period 2045–2050, surpassing intensities that were linked with the collapse of Larsen A and B ice shelves
36 (Trusel et al., 2015; Kittel et al., 2021). Given the existing uncertainty in the present precipitation and SMB
37 simulations and the significant range in the projected precipitation increase under various emissions
38 scenarios in CMIP5, CMIP6 and CORDEX models, there is medium confidence that the future Antarctic
39 SMB will have a negative contribution to sea level during the 21st century under all emissions scenarios (see
40 Section 9.4.2.3 for assessment of the drivers of future Antarctic ice sheet change and Section 9.4.2.6 for
41 longer time scales).
42
43
44 Atlas.11.1.5 Summary
45
46 Observations show a very likely widespread, strong warming trend starting in 1950s in the Antarctic
47 Peninsula. Significant warming trends are observed in other West Antarctic regions and at selected stations
48 in East Antarctica (medium confidence). Antarctic precipitation and SMB showed a significant positive trend
49 over the 20th century according to the ice cores, while large interannual variability masks any existing trend
50 over the satellite period since the end of 1970s (medium confidence).
51
52 An assessment of model performance for the present day shows that high-resolution regional climate models
53 with polar-optimised physics are important for estimating SMB and generating climate information, and
54 show improved realizations compared to reanalyses and GCMs when evaluated against observations. At the
55 same time, CMIP6 GCMs showed an improved representation of the Antarctic near-surface temperature
Do Not Cite, Quote or Distribute Atlas-108 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 compared to CMIP5, though still struggle with the representation of precipitation. There is therefore medium
2 confidence in the capacity of climate models to simulate Antarctic climate and SMB changes.
3
4 Under all assessed emission scenarios, both West and East Antarctica are very likely to have higher annual
5 mean surface air temperatures and more precipitation, which will have a dominant influence on determining
6 future changes in the SMB (high confidence). However, due to the challenges of model evaluation over the
7 region and the possibility of increased meltwater runoff described above, there is only medium confidence
8 that the future contribution of the Antarctic SMB to sea level this century will be negative under all
9 greenhouse-gas emission scenarios.
10
11
12 Atlas.11.2 Arctic
13
14 Atlas.11.2.1 Key features of the regional climate and findings from previous IPCC assessments
15
16 Atlas.11.2.1.1 Key features of the regional climate
17 The Arctic region comprises the Arctic Ocean (ARO), Russian Arctic (RAR), Greenland and Iceland (GIC)
18 and other surrounding land areas in Europe (NEW) and North America (NEN, NWN) (Figure Atlas.29). The
19 region is one of the coldest and driest regions on Earth and plays a key role influencing global and regional
20 climates and the hydrological cycle. A number of physical processes contribute to amplified Arctic
21 temperature variations as compared to the global temperature, in particular thermodynamic changes that
22 include the increase in surface absorption of solar radiation due to surface albedo feedbacks related with sea-
23 ice, ice, and snow-cover retreat as well as poleward energy transports, water-vapour-radiation and cloud-
24 radiation feedbacks (Screen and Simmonds, 2010; Serreze and Barry, 2011; Pithan and Mauritsen, 2014;
25 Bintanja and Krikken, 2016; Graversen and Burtu, 2016; Franzke et al., 2017; Stuecker et al., 2018).
26 Precipitation in the Arctic is dominated by snowfall, with rainfall present mostly during the summer period.
27 Arctic climate is influenced by the North Atlantic Oscillation, the leading mode of atmospheric variability in
28 the North Atlantic basin with a northward extension into the Arctic affecting temperature, precipitation and
29 sea ice over the region, with ENSO and Atlantic Multidecadal Variability also affecting parts of the region
30 (Annex IV). Further, the Greenland Ice Sheet contribution to sea-level results from the imbalance between
31 mass gain by net snow accumulation and mass loss by meltwater runoff and ice discharge into the ocean
32 (IMBIE team, 2020), highlighting that the ice sheet is a major contributor to sea level changes.
33
34
35 Atlas.11.2.1.2 Findings from previous IPCC assessments
36 The following summary from previous IPCC reports is derived from the SROCC (IPCC, 2019a) unless
37 otherwise stated. Arctic surface air temperatures have increased from the mid-1950s, with feedbacks from
38 loss of sea ice and snow cover contributing to the amplified warming (high confidence) (IPCC, 2018c), and
39 have likely increased by more than double the global average over the last two decades (high confidence).
40 Arctic snow cover in June has declined from 1967 to 2018 (high confidence). Arctic glaciers are losing mass
41 (very high confidence) and this along with changes in high-mountain snow melt have caused changes in
42 hydrology, including river runoff, that are projected to continue in the near term (high confidence). The rate
43 of ice loss from the Greenland Ice Sheet has increased; during 2006–2015 the loss was 278 ± 11 Gt yr-1 with
44 the rate for 2012–2016 higher than for 2002–2011 and several times higher than during 1992–2001 (high
45 confidence).
46
47 The Arctic sea-ice area is declining in all months of the year (very high confidence) with the September sea-
48 ice minimum very likely having reduced by 12.8 ± 2.3% per decade during the satellite era (1979 to 2018) to
49 levels unprecedented for at least 1000 years (medium confidence).
50
51 The high latitudes are likely to experience an increase in annual mean precipitation under RCP8.5 (IPCC,
52 2013c). Further, changes in precipitation will not be uniform. Autumn and spring snow cover duration are
53 projected to decrease by a further 5–10% from current conditions in the near term (2031–2050). No further
54 losses are projected under RCP2.6 whereas a further 15–25% reduction in snow cover duration is projected
55 by the end of century under RCP8.5 (high confidence).
Do Not Cite, Quote or Distribute Atlas-109 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Atlas.11.2.2 Assessment and synthesis of observations, trends and attribution
2
3 The Arctic has warmed at more than twice the global rate over the past 50 years with the greatest warming
4 during the cold season (Davy et al., 2018; Box et al., 2019; Przybylak and Wyszyński, 2020; Xiao et al.,
5 2020) (high confidence). This is based on various Arctic amplification processes, in particular the combined
6 effect of several related feedback processes including sea-ice and snow-cover albedo and water-vapour-
7 cloud-radiation feedbacks as well as poleward energy transports. The annual average Arctic surface air
8 temperature increased by 2.7°C from 1971 to 2017, with a 3.1°C increase in the cold season (October–May)
9 and a 1.8°C increase in the warm season (June–September) (AMAP, 2019). Satellite-based data estimate the
10 rate of annual warming for 1981–2012 over sea-ice covered regions to be 0.47°C per decade, whereas the
11 trend was significantly higher at 0.77°C per decade over Greenland and amplified in the northern Barents
12 and Kara seas (Comiso and Hall, 2014). The largest Arctic warming in 2003–2017 was reported over the
13 Barents and Kara Seas with trends larger than 2.5°C per decade (Susskind et al., 2019) and Arctic
14 temperatures from 2014 to 2018 have exceeded all previous records since 1900 (Blunden and Arndt, 2019).
15
16 Over the ARO, long-term temperature records are available from Spitsbergen (Svalbard Airport). For the
17 period 1898–2018, the annual mean warming was 0.32°C per decade, about 3.5 times the global mean
18 temperature for the same period and since 1991, 1.7°C per decade or about seven times the global average
19 for the same period (Nordli et al., 2020). There is a positive trend in the annual temperature for all stations
20 across Svalbard (Gjelten et al., 2016; Hanssen-Bauer et al., 2019; Dahlke et al., 2020) of 0.64°C–1.01°C per
21 decade for 1971–2017 (Hanssen-Bauer et al., 2019) co-varying with regional changes in sea-ice conditions
22 (Dahlke et al., 2020). The largest temperature trends very likely occur in winter, with Svalbard airport
23 warming at 0.43°C per decade during 1898–2018 and 3.19°C per decade during 1991–2018 (Nordli et al.,
24 2020) and Isaksen et al. (2016) reporting on substantial warming in western Spitsbergen, particularly in
25 winter, while the summer warming is moderate.
26
27 A multi-dataset analysis for NEN shows a consistent warming (Rapaić et al., 2015), with the largest annual
28 temperature trend greater than 0.3°C per decade during 1981 to 2010 over eastern NEN and also significant
29 warming over northern Quebec and most of the Canadian Arctic north of the treeline. For the longer 1950–
30 2010 period, a consistent warming is found over central and western NEN, but no trend or no consensus is
31 found over the Labrador coast. The latter is related with cooling of the North Atlantic region during the
32 1970s. For western Greenland, however, summer temperatures increased (2.2°C in June, 1.1°C in July) from
33 1994 to 2015 (Saros et al., 2019). For neighbouring Arctic regions of NEU, WSE and ESB, datasets show a
34 consistent warming of annual mean temperature since the mid-1970s and 1980 (see Sections Atlas.8 and
35 Atlas.5.2).
36
37 Along with the amplified warming, the Arctic has become moister (Rinke et al., 2019; Nygård et al., 2020).
38 AMAP reported Arctic precipitation increases of 1.5–2.0% per decade, with the strongest increase in the cold
39 season (October–May) (AMAP, 2019) (medium confidence). Also, for neighbouring Arctic regions for
40 example NEU, EEU and Northern Asia, mean annual precipitation has increased since the early 20th century
41 (see Sections Atlas.8 and Atlas.5.2). Estimated trends for precipitation and snowfall fraction are mixed for
42 Arctic, with increases and decreases for different regions and seasons (Vihma et al., 2016). However, annual
43 precipitation trends derived from different reanalyses do not agree, differ in sign and have low significance
44 (Lindsay et al., 2014; Boisvert et al., 2018). Direct precipitation measurements are difficult and include
45 uncertainties (among others measuring frozen precipitation), therefore precipitation estimates in the Arctic
46 rely on climate models and reanalyses.
47
48 An average of five reanalyses for 2000–2010 suggests around 40% of Arctic Ocean precipitation falls as
49 snow, though there is large uncertainty in this estimate (Boisvert et al., 2018). Rainfall frequency is
50 estimated to have increased over the Arctic by 2.7–5.4% over 2000–2016 (Boisvert et al., 2018) with more
51 frequent rainfall events reported for NEU and ARO (Svalbard) (Maturilli et al., 2015; AMAP, 2019), and
52 winter rain totals and frequency have increased in Svalbard since 2000 (Łupikasza et al., 2019) (medium
53 confidence). Rain-free winters have rarely occurred since 1998 (Peeters et al., 2019).
54
55 Observational records (1966 to 2010) for the RAR region show changing precipitation characteristics (Ye et
Do Not Cite, Quote or Distribute Atlas-110 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 al., 2016), with higher precipitation intensity but lower frequency and little change in annual precipitation
2 total. Precipitation intensity is reported to have increased in all seasons, strongest in winter and spring,
3 weakest in summer, and at a rate of about 1–3% per degree Celsius of air temperature increase.
4
5 Atlas.11.2.3 Assessment of model performance
6
7 Evaluating simulated temperature and precipitation is problematic in the Arctic due to sparse weather station
8 observations. The lack of reliable observed precipitation datasets for the Arctic thus makes it very unlikely to
9 be able to evaluate objectively the skill of models to reproduce precipitation patterns (Takhsha et al., 2018).
10
11 The CMIP5 models reproduce the observed Arctic warming over the past century (Chylek et al., 2016; Hao
12 et al., 2018; Huang et al., 2019) (medium confidence). The simulated mean Arctic warming for 1900–2014
13 averaged over 40 CMIP5 models is 2.7°C compared to the observed values of 2.2°C (NASA GISS data
14 smoothed using a 1200-km radius) or 1.7°C (using a 250-km smoothing radius) (Chylek et al., 2016).
15 However, there are large inter-model differences in the simulated warming which ranges from 1.2°C to
16 5.0°C. Although the CMIP5 models reproduce the spatially averaged observed warming over the past 50 to
17 100 years, the pattern is different from that of observations and reanalysis (Xie et al., 2016; Franzke et al.,
18 2017; Hao et al., 2018). Zonal mean temperature trends in the CMIP5 models overestimate the warming in
19 the cold season over high latitudes in the northern hemisphere (Xie et al., 2016). Overall, the amplified
20 Arctic warming in the recent decades is overestimated by CMIP5 models (Huang et al., 2019). Possible
21 reasons are modelled sea surface temperature biases and an overestimated temperature response to the Arctic
22 sea-ice decline. Furthermore, some models, which have a warm or weak bias in their Arctic temperature
23 simulations, closely relate the Arctic warming to changes in the large-scale atmospheric circulation. In other
24 models, which show large cold biases, the albedo feedback effect plays a more important role for the
25 temperature trend magnitude. This implies that the dominant simulated Arctic warming mechanism and trend
26 may be dependent on the bias of the model mean state (Franzke et al., 2017). Compared to CMIP5 models,
27 Davy and Outten (2020) found lower biases in CMIP6 models’ representation of sea-ice extent and volume
28 with improved extents linked to a better seasonal cycle in the Barents Sea.
29
30 Rapid temperature changes, such as the pronounced increase of 2°C yr–1 during 2003–2012 over the Kara
31 and Barents Seas in March is well captured in Arctic CORDEX simulations (Kohnemann et al., 2017). The
32 models show adequate skill in capturing the general temperature patterns (Koenigk et al., 2015; Matthes et
33 al., 2015; Hamman et al., 2016; Cassano et al., 2017; Brunke et al., 2018; Diaconescu et al., 2018; Takhsha
34 et al., 2018), but tend to show a cold temperature bias which is largest in winter and depends on the reference
35 dataset. Cassano et al. (2017) showed a large sensitivity of the simulated surface climate to changes in
36 atmospheric model physics. In particular, large changes in radiative flux biases, driven by changes in
37 simulated clouds, lead to large differences in temperature and precipitation biases.
38
39 The CMIP5 models perform well in simulating 20th-century snowfall for the northern hemisphere, although
40 there is a positive bias in the multi-model ensemble relative to the observed data in many regions (Krasting
41 et al., 2013b). Lack of sufficient spatial resolution in the model topography has a serious impact on the
42 simulation of snowfall. The patterns of relative maxima and minima of snowfall, however, are captured
43 reasonably well by the models.
44
45 Arctic CORDEX RCMs reproduce the dominant features of regional precipitation patterns and extremes
46 (e.g., Glisan and Gutowski, 2014; Hamman et al., 2016). Due to their higher spatial resolution, RCMs
47 simulates larger amounts of orographic precipitation compared to reanalyses. Overall, the simulated
48 precipitation is within the reanalysis and global model ensemble spread, but the Arctic river basin
49 precipitation is closer to observations (Brunke et al., 2018). However, Takhsha et al. (2018) show that the
50 RCMs precipitation bias highly depends on the observational reference dataset used.
51
52 The annual mean precipitation pattern of ensemble global atmospheric simulations with a high horizontal
53 resolution agrees well with the observations, with precipitation maxima over the Greenland and Norwegian
54 Seas (Kusunoki et al., 2015). However, the simulated Arctic average annual precipitation shows a positive
55 bias with excessive precipitation over Alaska and the western Arctic (Kattsov et al., 2017).
Do Not Cite, Quote or Distribute Atlas-111 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 Regarding the Greenland Ice Sheet (region GIC), modelled surface mass balance (SMB) has decreased since
2 the end of the 1990s (Fettweis et al., 2020). A multi-model inter-comparison study (Fettweis et al., 2020)
3 emphasized a simulated positive mean annual SMB of 338 ± 68 Gt yr-1 between 1980 and 2012, with a
4 decreasing average rate of 7.3 ± 2.0 Gt yr-2, mainly driven by an increase in meltwater runoff. Mouginot et
5 al. (2019) stated that SMB played a strong role in the ice sheet mass loss, where SMB dominated in the last
6 two decades. Mottram et al. (2019) found that SMB processes dominate the ice sheet mass budget over most
7 of the interior, highlighting that the ice sheet is a contributor to global mean sea level rise between 1991 and
8 2015. More specifically, SMB models have improved (Fettweis et al., 2020; Hanna et al., 2021) due to
9 increased availability and quality of remotely sensed (Koenig et al., 2016; Overly et al., 2016) and in-situ
10 observations (Machguth et al., 2016; Fausto et al., 2018; Vandecrux et al., 2019, 2020). Fettweis et al. (2020)
11 showed that the models’ ensemble mean provides the best estimate of the present-day SMB relative to
12 observations. This is the case for the patterns in all seven regions (regional division after Mouginot et al.,
13 2019) apart from the SE accumulation zone where large discrepancies in modelled snowfall accumulation
14 occurred where the spread can reach 2 m water equivalent per year. Montgomery et al. (2020) confirmed this
15 highlighting that RCMs (MAR and RACMO) are underestimating accumulation in southeast Greenland and
16 that models misrepresent spatial heterogeneity due to an orographically forced bias in snowfall near the
17 coast. Further, for northeast Greenland, Karlsson et al. (2020) found RCMs underestimate snow
18 accumulation rates by up to 35%. The regional time series show that SMB has been gradually decreasing in
19 all seven regions (1979–2017), although the trend is less strong in central-eastern and southeast regions. In
20 the southwest, northeast and northwest, SMB turns negative or close to zero after 2000 and remains above
21 zero in other regions (Figure Atlas.30) (medium confidence).
22
23
24 Atlas.11.2.4 Assessment and synthesis of projections
25
26 Mean temperature in the Arctic is projected to continue to rise through the 21st century significantly higher
27 than the global average (Figure Atlas.29 and Interactive Atlas). For the regions NWN and NEN, see Section
28 Atlas.9. The Arctic is projected to reach a 2°C annual mean warming above the 1981–2005 baseline about 25
29 to 50 years before the globe as a whole under RCP8.5 and RCP4.5 (Post et al., 2019). The Arctic warming
30 may be as much as 4°C in the annual mean and 7°C in late autumn under 2°C global warming, regardless of
31 which scenario is considered (Post et al., 2019) (high confidence).
32
33 Projections from 40 CMIP5 models of the 2014–2100 Arctic annual warming under RCP4.5 vary from 0.9°C
34 to 6.7°C, with a multi-model mean of 3.7°C (Chylek et al., 2016). All models agree on a projected Arctic
35 amplification (of at least 1.5 times), but they disagree on the magnitude and spatial patterns. Arctic warming
36 trends projected by models that include a full direct and indirect aerosol effect (‘fully aerosol-cloud
37 interactive’) are significantly higher than those projected by models without a full indirect aerosol effect
38 (Chylek et al., 2016).
39
40 Projected Arctic warming exhibits a very pronounced seasonal cycle, with exceptionally strong warming in
41 the winter. In projections from 30 CMIP5 models, winter warming over ARO varies regionally from 3°C to
42 5°C by mid-century and 5°C to 9°C by late-century under RCP4.5 (AMAP, 2017) (high confidence).
43 Averaged over the Arctic and based on 36 CMIP5 models, winter warming is 5.8 ± 1.5°C by mid-century
44 and 7.1 ± 2.3°C by 2100 under RCP4.5 (Overland et al., 2019), and an exceptionally strong warming of up to
45 14.1 ± 2.9°C is projected in December under RCP8.5 (Bintanja and Krikken, 2016). Bintanja and Van Der
46 Linden (2013) estimated the Arctic winter warming over the 21st century to exceed the summer warming by
47 at least a factor of four, irrespective of the magnitude of the climate forcing.
48
49 Overland et al. (2014) highlighted the difference between the near-term ‘adaptation timescale’ and the long-
50 term ‘mitigation timescale’ for the Arctic. Only in the latter half of the century do the projections under
51 RCP4.5 and RCP8.5 noticeably separate. End-of-the-century warming is approximately twice as large under
52 RCP8.5 demonstrating the impact of the lower emissions under RCP4.5 (AMAP, 2017) (high confidence).
53 More specifically under the strong forcing scenario, annual mean surface air temperature in the Arctic is
54 projected to increase by 8.5 ± 2.1°C over the course of the 21st century (Bintanja and Andry, 2017), and
55 emerges as a ‘new Arctic’ climate being significantly different from that of the mid-20th century (Landrum
Do Not Cite, Quote or Distribute Atlas-112 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 and Holland, 2020). The end-of-the-century warming (2080–2099 relative to 1980–1999, RCP8.5) can
2 exceed 15°C in autumn and winter over the Arctic Ocean (Koenigk et al., 2015). Projections averaged over
3 the four best-performing CMIP5 models show an Arctic annual warming of 4.1°C (RCP2.6), 5.7°C
4 (RCP4.5), and 10.6°C (RCP8.5) by 2100 compared to 1951–1980 (Hao et al., 2018). Also, for neighbouring
5 Arctic regions, for example NEU, WSB and ESB, temperature is projected to increase towards the end of the
6 century under both RCP4.5 and RCP8.5 (see Sections Atlas.8 and Atlas.5.2).
7
8 The ensemble of CMIP6 shows likely greater warming compared to CMIP5 (Figure Atlas.29). There is weak
9 agreement among the models in projected temperature change over the Arctic North Atlantic under SSPs
10 until the mid-century, but a robust warming signal clearly emerges even there by 2100 (Interactive Atlas).
11 Generally, the largest annual warming is simulated over the Arctic Ocean, particularly over the Barents and
12 Kara Seas. Future warming in CORDEX RCMs and the CMIP5 models are similar (Spinoni et al., 2020).
13 The RCM warming over the AO is smaller, while the warming over land is larger in winter and spring but
14 smaller in summer, compared with CMIP5 (Koenigk et al., 2015).
15
16 Mean precipitation in ARO, GIC and RAR is projected to rise in a warming climate (Figure Atlas.29), with
17 different rates for the different seasons and scenarios. For NWN and NEN, see Section Atlas.9. The CMIP5
18 multi-model mean projected precipitation increase in the Arctic is likely of the order of 50% under RCP8.5
19 by the end of 21st century, which is among the highest globally (Bintanja and Selten, 2014). Over 70°N–
20 90°N, the precipitation increase is likely 62 ± 20% and 56 ± 13% for RCP4.5 and RCP8.5 respectively. For
21 ARO (Svalbard), the increase in annual precipitation by 2100 is estimated to be about 45% for RCP4.5 and
22 65% for RCP8.5 (CMIP5 ensemble median) (Van der Bilt et al., 2019). However, importantly the simulated
23 Arctic precipitation increase varies by a factor of three to four between models (Bintanja and Selten, 2014).
24 The projected increase is strongest in late autumn and winter (Vihma et al., 2016). The interannual
25 variability of Arctic precipitation will likely increase markedly (up to 40% over the 21st century), especially
26 in summer (Bintanja et al., 2020) (medium confidence)
27
28 The CMIP6 projections confirm precipitation will likely increase almost everywhere in the Arctic (see the
29 Interactive Atlas). The largest increase is simulated over the Barents and Kara Seas and East Siberian Sea
30 regions, and over northeast Greenland. A pronounced uncertainty in the projection exists over the Arctic
31 North Atlantic and south Greenland. There, the precipitation signal is not significant even by the end of the
32 21st century and under high-emission scenarios (RCP8.5, SSP5-8.5). Consistent with the generally higher
33 warming in CMIP6, compared to CMIP5, the projected precipitation increase is also higher (Figure Atlas.29)
34 (high confidence).
35
36 The Arctic mean annual precipitation sensitivity has been estimated at 4.5% increase per degree Celsius of
37 temperature rise, compared to a global average of 1.6–1.9% per degree Celsius of temperature rise (Bintanja
38 and Selten, 2014) based on a set of 37 CMIP5 GCMs. Koenigk et al. (2015) stress the different precipitation
39 sensitivity in winter (0.8 mm per month per degree Celsius of temperature rise) and summer (2 mm per
40 month per degree Celsius of temperature rise). Dobler et al. (2016) support the high precipitation sensitivity
41 for the projected temperature changes. The pattern and amplitude of precipitation changes agree in
42 CORDEX simulations with their driving CMIP5 models (Koenigk et al., 2015; Spinoni et al., 2020) (high
43 confidence). However, more small-scale variations over land and coastlines and significantly larger
44 precipitation changes in summer are obvious in the downscaling.
45
46 Rain is projected to become the dominant form of precipitation in the Arctic region by the end of the 21st
47 century (Bintanja, 2018). The CMIP5 models show a decrease in annual Arctic snowfall under both RCP4.5
48 and RCP8.5 (Krasting et al., 2013b; Bintanja and Andry, 2017) (high confidence). In the central Arctic, the
49 snowfall fraction barely remains larger than 50%, with only Greenland still having snowfall fractions larger
50 than 80% (Bintanja and Andry, 2017). The most dramatic reductions in snowfall fraction are projected to
51 occur over the North Atlantic and especially the Barents Sea.
52
53 With ongoing warming and polar amplification in the Arctic, the Greenland Ice Sheet SMB will inevitably
54 continue to change (Lenaerts et al., 2019) (high confidence). For the ice sheet, despite large differences
55 between model scenarios, future projections and regions agree that increasing temperatures will increase
Do Not Cite, Quote or Distribute Atlas-113 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 runoff which will in turn dominate the future decrease of SMB (Rae et al., 2012; van Angelen et al., 2014;
2 Mottram et al., 2017; Hofer et al., 2020), confirming the high sensitivity of the SMB to atmospheric
3 warming. Changes in SMB will continue to dominate future mass loss from the ice sheet, and likely even
4 more when marine terminating glaciers retreat onto land, and solid ice discharge is reduced (Vizcaino, 2014;
5 Lenaerts et al., 2019).
6
7
8 Atlas.11.2.5 Summary
9
10 It is very likely that the Arctic has warmed at more than twice the global rate over the past 50 years and likely
11 that annual precipitation has increased with the highest increases during the cold season. This is based on
12 various Arctic amplification processes, in particular combination of several feedback-related processes such
13 as sea-ice and snow-cover albedo, poleward energy transports, and water-vapour-cloud-radiation feedbacks.
14 The frequency of rainfall increased over the Arctic by 2.7 to 5.4% over the 2000–2016 period with more
15 frequent rainfall events being reported for northern Europe and Svalbard (medium confidence).
16
17 The CMIP5 models reproduce the observed Arctic warming over the past century but overestimate the
18 amplified Arctic warming in the recent decades (medium confidence). Arctic CORDEX simulations show
19 adequate skill in capturing regional temperature and precipitation patterns and precipitation extremes (high
20 confidence). SMB models have improved due to increased availability and quality of remotely sensed and in-
21 situ observations and an ensemble mean of SMB model simulations provides the best estimate of the present-
22 day SMB (medium confidence).
23
24 It is very likely that the Arctic annual mean temperature and precipitation will continue to increase, reaching
25 around 11.5 ± 3.4°C and 49 ± 19% over the 2081–2100 period (with respect to a 1995–2014 baseline) under
26 the SSP5-8.5 scenario or 4.0 ± 2.5°C and 17 ± 11% under the SSP1-2.6 scenario. These CMIP6 results show
27 likely higher Arctic annual mean temperatures compared to CMIP5 for a given time-period and emissions
28 scenario, though the projections are consistent for global warming levels.
29
30
31 Atlas.12 Final remarks
32
33 Developing from the WGI AR5 Atlas Annex (IPCC, 2013a), the Atlas is an innovation in the WGI
34 contribution to the AR6, providing a region-by-region assessment of new knowledge on changes in mean
35 climate and an online interactive tool, the Interactive Atlas. It demonstrates the diversity in the climate
36 changes across these regions, in the evidence base for generating information on what changes have already
37 occurred and why, and what further changes each region is likely to face in the future based on different
38 emission scenarios. Finally, the Interactive Atlas allows for further exploration of the data underpinning
39 assessment material generated by many of the other chapters.
40
41 The foundation of the regional information generated by the Atlas chapter is an assessment of the significant
42 body of new literature on regional climate change though noting substantial heterogeneity in both its
43 availability and the involvement of regional expertise. In many regions this allows for an in-depth
44 assessment though in some the range of information that can be provided and/or the level of confidence in
45 the information is limited. There is similar heterogeneity in the availability of observations to assess recent
46 trends and evaluate model performance, with a lack of curated regional datasets in polar regions, northern
47 South America and Africa.
48
49 Internal variability is a large contributor to the climate uncertainty at regional scales. Recent work has
50 combined outputs of Single Model Initial-conditions Large Ensembles (SMILEs) with CMIP5 and CMIP6 to
51 partition and gain insights on the modelled range and uncertainty arising from internal variability and from
52 model response uncertainty for a given emission scenario (Deser et al., 2020; Lehner et al., 2020; Maher et
53 al., 2021). The work highlights the notable role for internal variability at regional scales, particularly for
54 precipitation in regions with weaker forced response, where internal variability can remain larger than model
55 uncertainty or scenario uncertainty throughout the whole century. The Atlas (similar to the other regional
Do Not Cite, Quote or Distribute Atlas-114 Total pages: 196
Final Government Distribution Atlas IPCC AR6 WGI
1 chapters) uses a single realization per model (CMIP6 models provide multiple realizations, but it is not the
2 case for CORDEX and less so for CMIP5) which allows for the comparison of the different lines of evidence
3 but at the expense of internal variability having a larger influence on the ability to detect or quantify changes.
4
5 The assessment produced in the Atlas is based on the individual results from the different lines of global and
6 regional evidence and the consistency amongst them, as there is a lack of literature on methodologies that
7 combine multiple lines of evidence to distil regional climate change information.
8
9
Do Not Cite, Quote or Distribute Atlas-115 Total pages: 196