Final Government Distribution Annex III IPCC AR6 WGIII Table of Contents Annex III: Scenarios and Modelling Methods ..................................................................................... I-1 Preamble .............................................................................................................................................. I-4 Part I. Modelling methods .............................................................................................................. I-5 1. Overview of modelling tools........................................................................................................ I-5 2. Economic frameworks and concepts used in sectoral models and integrated assessment models I- 8 3. Energy system modelling ........................................................................................................... I-11 3.1. Bottom-up models ............................................................................................................... I-11 3.2. Modelling of energy systems in context of economy ......................................................... I-13 3.3. Hybrid models ..................................................................................................................... I-13 4. Building sector models .............................................................................................................. I-14 4.1. Models purpose, scope and types ........................................................................................ I-14 4.2. Representation of energy demand and GHG emissions ...................................................... I-14 4.3. Representation of mitigation options .................................................................................. I-15 4.4. Representation of climate change impacts .......................................................................... I-15 4.5. Representation of sustainable development dimensions ..................................................... I-16 4.6. Models underlying the assessment in Chapter 9 ................................................................. I-18 5. Transport models ....................................................................................................................... I-22 5.1. Purpose and scope of models .............................................................................................. I-22 5.2. Inventory of transportation models included in AR6 .......................................................... I-24 6. Industry sector models ............................................................................................................... I-24 6.1. Types of industry sector models ......................................................................................... I-24 6.2. Representation of demand for industrial products .............................................................. I-24 6.3. Representation of mitigation options - mitigation options, how their uptake is represented, how potentials and costs are represented ................................................................................... I-25 6.4. Limitations and critical analysis ......................................................................................... I-26 7. Land use modelling .................................................................................................................... I-27 7.1. Modelling of land use and land use change ........................................................................ I-27 7.2. Demand for food, feed, fibre and agricultural trade ............................................................ I-28 7.3. Treatment of land-based mitigation options ....................................................................... I-28 7.4. Treatment of environmental and socio-economic impacts of land use ............................... I-29 8. Reduced complexity climate modelling ..................................................................................... I-29 9. Integrated assessment modelling ............................................................................................... I-30 9.1. Types of Integrated Assessment Models............................................................................. I-31 9.2. Components of integrated assessment models .................................................................... I-32 Do Not Cite, Quote or Distribute I-2 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 9.3. Representation of nexus issues and sustainable development impacts in IAMs ................. I-33 9.4. Policy analysis with IAMs .................................................................................................. I-35 9.5. Limitations of IAMs............................................................................................................ I-37 10. Key characteristics of models that contributed mitigation scenarios to the assessment .......... I-39 11. Comparison of mitigation and removal measures represented by models that contributed mitigation scenarios to the assessment........................................................................................... I-42 Part II. Scenarios ........................................................................................................................... II-48 1. Overview on climate change scenarios ..................................................................................... II-48 1.1. Purposes of climate change scenarios ................................................................................ II-48 1.2. Types of climate change mitigation scenarios ................................................................... II-49 1.3. Scenario framework for climate change research .............................................................. II-52 1.4. Key design choices and assumptions in mitigation scenarios ............................................ II-54 2. Use of scenarios in the assessment ........................................................................................... II-57 2.1. Use of scenario literature and database .............................................................................. II-57 2.2. Treatment of scenario uncertainty...................................................................................... II-58 2.3. Feasibility of mitigation scenarios ..................................................................................... II-58 2.4. Illustrative mitigation pathways ......................................................................................... II-60 2.5. Scenario approaches to connect WG III with the WG I and WG II assessments .............. II-63 3. WG III AR6 scenario database ................................................................................................. II-68 3.1. Process of scenario collection and vetting ......................................................................... II-68 3.2. Global pathways................................................................................................................. II-70 3.3. National and regional pathways ......................................................................................... II-81 3.4. Sector transition pathways ................................................................................................. II-83 References ......................................................................................................................................... II-85 Do Not Cite, Quote or Distribute I-3 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Preamble 2 The use of scenarios and modelling methods are pillars in IPCC WG III Assessment Reports. Past WG 3 III assessment report cycles identified knowledge gaps about the integration of modelling across scales 4 and disciplines, mainly between global integrated assessment modelling methods and bottom-up 5 modelling insights of mitigation responses. The need to improve the transparency of model assumptions 6 and enhance the communication of scenario results was also recognised. 7 This annex on Scenarios and modelling methods aims to address some of these gaps by detailing the 8 modelling frameworks applied in the WG III AR6 chapters and disclose scenario assumptions and its 9 key parameters. It has been explicitly included in the Scoping Meeting Report of the WG III 10 contribution to the AR6 and approved by the IPCC Panel in the 46th Session of the Panel. 11 The annex includes two parts: Part I. on modelling methods summarises methods and tools available to 12 evaluate sectorial, technological and behavioural mitigation responses as well as integrated assessment 13 models (IAMs) for the analysis of “whole system” transformation pathways; Part II on scenarios sets 14 out the portfolio of climate change scenarios and mitigation pathways assessed in the WG III AR6 15 chapters, its underneath principles and interactions with scenario assessments by WG I and WG II. 16 Do Not Cite, Quote or Distribute I-4 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Part I. Modelling methods 2 3 1. Overview of modelling tools 4 Modelling frameworks vary vastly amongst themselves, and several key characteristics can be used as 5 basis for model classification (Scrieciu et al. 2013; Hardt and O’Neill 2017; Capellán-Pérez et al. 2020; 6 Dodds et al. 2015). Broadly, literature characterises models along three dimensions: (i) level of detail 7 and heterogeneity, (ii) mathematical algorithm concepts and (iii) and temporal and spatial system 8 boundaries (Krey 2014). 9 Commonly climate mitigation models are referred to as bottom-up and top-down depending upon their 10 degree of detail (van Vuuren et al. 2009). Generally, bottom-up approaches present more systematic 11 individual technological details about a reduced number of mitigation strategies of a specific sector or 12 sub-sector. These models tend to disregard relations between specific sectors/technologies and miss 13 evaluating interactions with the whole system. On the other hand, top-down approaches present a more 14 aggregated and global analysis, in detriment of less detailed technological heterogeneity. They tend to 15 focus on interactions within the whole system, such as market and policy instrument interactions within 16 the global economy systems. Studies using top-down models are more capable of representing 17 economic structural change than adopting technology-explicit decarbonisation strategies (Kriegler et al. 18 2015a; van Vuuren et al. 2009). Integrated Assessment Models (IAMs) typically use a top-down 19 approach to model sectorial mitigation strategies. 20 Although this dichotomic classification has been while mentioned in the literature, since AR5, climate 21 mitigation models have evolved towards a more hybrid approach incorporating attributes of both 22 bottom-up and top-down approaches. This is partly due to different modelling communities having 23 different understandings of these two approaches principles, which can be misleading. 24 One of the most basic aspects of a modelling tool is how it approaches the system modelled from a 25 solution perspective. A broad interpretation of mathematical algorithm concepts classifies models as 26 simulation and optimisation models. Simulation models are based on the evaluation of the dynamic 27 behaviour of a system (Lund et al. 2017). They can be used to determine the performance of a system 28 under alternative options of key parameters in a plausible manner. Most often, simulation models 29 require comprehensive knowledge of each parameter, in order to choose a specific path under several 30 alternatives. On the other hand, optimisation models seek to maximise or minimise a mathematical 31 objective function under a set of constraints (Iqbal et al. 2014; Baños et al. 2011). Most often, the 32 objective function represents the total cost or revenue of a given system or the total welfare of a given 33 society. One major aspect of optimisation models is that the solution in achieved by simultaneously 34 binding a set of constraints, which can be used to represent real life limitation on the system, such as: 35 constraints on flows, resource and technology availability, labour and financial limitations, 36 environmental aspects, and many other characteristics that the model may require (Fazlollahi et al. 37 2012; Cedillos Alvarado et al. 2016; Pfenninger et al. 2014). Specifically, when modelling climate 38 mitigation responses, limiting carbon budgets is often used to represent future temperature level 39 pathways (Gidden et al. 2019; Rogelj et al. 2016; Millar et al. 2017; Peters 2018). 40 Another major distinction amongst modelling tools is related to the solution methodology from a 41 temporal perspective. They can have a perfect foresight intertemporal assumption or a recursive- 42 dynamic assumption. Intertemporal optimisation with perfect foresight is an optimisation method for 43 achieving an overall optimal solution over time. It is based on perfect information on all future states 44 of a system and assumptions (such as technology availability and prices) and, as such, today’s and future 45 decisions are made simultaneously, resulting in a single path of optimal actions that lead to the overall 46 optimal solution (Keppo and Strubegger 2010; Gerbaulet et al. 2019). Such modelling approach can Do Not Cite, Quote or Distribute I-5 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 present an optimal trajectory of the set of actions and policies that would lead to the overall first-best 2 solution. However, real-life decisions are not always based on optimal solutions (Ellenbeck and 3 Lilliestam 2019) and, therefore, solutions from perfect foresight models can be challenging to be 4 implemented by policymakers (Pindyck 2013, 2017). For instance, perfect foresight implies perfect 5 knowledge of the future states of the system, such as future demand on goods and products and 6 availability of production factors and technology. 7 Recursive-dynamic models, also known as myopic or limited foresight models, make decisions over 8 sequential periods of time. For each time step, the solution is achieved without information of future 9 time steps. Therefore, the solution path is a series of solutions in short trajectories that, ultimately, are 10 very unlikely to achieve the overall optimal solution over the whole time period considered (Fuso Nerini 11 et al. 2017). Nonetheless, the solution represents a set of possible and plausible policies and behavioural 12 choices of the agents that could be taken in short-term cycles, without perfect information (Hanna and 13 Gross 2020; Heuberger et al. 2018). In between, some models consider imperfect or adaptive 14 expectations, where economic decisions are based on past, current and imperfectly anticipated future 15 information (Keppo and Strubegger 2010; Löffler et al. 2019; Kriegler et al. 2015a). Modelling tools 16 can also be differentiated by their level of representation of economic agents and sectors: they can have 17 a full representation of all agents of the economy and their interactions with each other (general 18 equilibrium) or focus on a more detailed representation of a subset of economic sector and agents 19 (partial equilibrium) (Babatunde et al. 2017; Cheng et al. 2015; Hanes and Carpenter 2017; Sanchez 20 et al. 2018; Guedes et al. 2019; Pastor et al. 2019) (Annex III, I.2). 21 The most basic aspect to differentiate models is their main objective function, which include the detail 22 at which they represent key sectors, systems and agents. This affects the decision on methodology and 23 other coverage aspects. Several models have been developed for different sectorial representation, such 24 as the energy (Annex III.I.3) buildings (Annex III.I.4), transports (Annex III.I.5), industry (Annex 25 III.I.6) and land use (Annex III.I.7). 26 Modelling exercises vary considerable in terms of key characteristics, including geographical scales, 27 time coverage, environmental variables, technologies portfolio, and socioeconomic assumptions. A 28 detailed comparison of key characteristics of global and national models used in this report is presented 29 in Annex III.I.9.Geographical coverage ranges from sub-national (Cheng et al. 2015; Feijoo et al. 2018; 30 Rajão et al. 2020), national (Vishwanathan et al. 2019; Li et al. 2019; Sugiyama et al. 2019; Schaeffer 31 et al. 2020), regional (Vrontisi et al. 2016; Hanaoka and Masui 2020) and global models (McCollum et 32 al. 2018; Gidden et al. 2018; Kriegler et al. 2018a; Rogelj et al. 2019b; Drouet et al. 2021). Even models 33 with the same geographical coverage can still be significantly different from each other, for instance, 34 due to the number of regions within the model. Models can also have spatially implicit and explicit 35 formulations, which in turn can have different spatial resolution. This distinction is especially important 36 for land use models, which account for changes in land use and agricultural practices (Annex III.7. Land 37 use modelling). The time horizon, time steps and time resolution are major aspects that differ across 38 models. Model horizon can range from short- to long-term, typically reaching from a few years to up 39 until the end of the century (Fujimori et al. 2019b; Rogelj et al. 2019a; Ringkjøb et al. 2020; Gidden et 40 al. 2019). Time resolution is particularly relevant for specific applications, such as power sector models, 41 which have detailed representation of power technologies dispatch and operation (Soria et al. 2016; 42 Abujarad et al. 2017; Guan et al. 2020). 43 Life Cycle Assessment (LCA) is an integrated technique to evaluate the sustainability of a product 44 throughout its life cycle. It quantifies the environmental burdens associated with all stages from the 45 extraction of raw materials, through the production of the product itself, its utilisation, and end-life, 46 either via reuse, recycling or final disposal (Rebitzer et al. 2004; Finnveden et al. 2009; Guinée et al. 47 2011; Curran 2013; Hellweg and Milà i Canals 2014). The environmental impacts covered include all 48 types of loads on the environment through the extraction of natural resources and emission of hazardous Do Not Cite, Quote or Distribute I-6 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 substances. For this reason, LCA has the flexibility to evaluate an entire product system hence avoiding 2 sub-optimisation in a single process and identifying the products/processes that result in the least 3 environmental impact. Thus, it allows for the quantification of possible trade-offs between different 4 environmental impacts (e.g. eliminating air emissions by increasing non-renewable energy resources) 5 (Gibon et al. 2017; Nordelöf et al. 2014; Hawkins et al. 2013) and/or from one stage to other (e.g. reuse 6 or recycling a product to bring it back in at the raw material acquisition phase) (Hertwich and Hammitt 7 2001a,b). It gives a holistic view of complex systems and reduces the number of parameters for which 8 decisions have to be taken, while not glossing over technical and economical details. In recent years, 9 LCA has been widely used in both retrospective and prospective analysis of product chains in various 10 climate mitigation fields, namely comparing existing energy technologies with planned alternatives 11 (Portugal-Pereira et al. 2015; Cetinkaya et al. 2012), product innovation and development (Wender et 12 al. 2014; Sharp and Miller 2016; Portugal-Pereira et al. 2015), certification schemes (Prussi et al. 2021), 13 or supply chain management (Hagelaar 2001; Blass and Corbett 2018). 14 15 Two different types of LCA approaches can be distinguished: the Attributional Life Cycle Assessment 16 (ALCA) and the Consequential Life Cycle Assessment (CLCA). The Attributional Life Cycle 17 Assessment (ALCA) aims at describing the direct environmental impacts of a product. It typically uses 18 average and historical data to quantify the environmental burden during a products’ life cycle, and it 19 tends to exclude market effects or other indirect effects of the production and consumption of products 20 (Baitz 2017). CLCA, on the other hand, focus on the effects of changes due to product life cycle, 21 including both consequences inside and outside the product life cycle (Earles and Halog 2011). Thus, 22 the system boundaries are generally expanded to represent direct and indirect effects of products’ 23 outputs. CLCA tends to describe more complex systems than ALCA that are highly sensitive to data 24 assumptions (Plevin et al. 2014; Weidema et al. 2018; Bamber et al. 2020). 25 Integrated Assessment Model (IAM) are simplified representations of the complex physical and social 26 systems, focusing on the interaction between economy, society and the environment (Annex III.I.9). 27 They represent the coupled energy-economy-land-climate system to varying degrees. In a way, IAM 28 differ themselves in all the topics discussed in this section: significant variation in geographical, 29 sectorial, spatial and time resolution; rely greatly on socioeconomic assumptions; different 30 technological representation; partial or general equilibrium assumptions; differentiated between perfect 31 foresight or recursive-dynamic methodology. The difficulty in fully representing the extent of climate 32 damages in monetary terms may be the most important and challenging limitation of IAMs and it is 33 mostly directed to cost benefit IAMs. However, both categories of IAMs present important limitations 34 (Annex III.I.9). 35 Following this brief synopsis of modelling taxonomies, Section 2. Economic frameworks and concepts 36 used in sectoral models and integrated assessment models details key aspects of economic frameworks 37 and principles used to modelling climate mitigation responses and estimates its costs. Sections I.3, I.4, 38 I.5, I.6, I.7 present key aspects of sectorial modelling approaches in energy systems, buildings, 39 transports, industry, and land use, respectively. Interactions between WG I climate emulators and WG 40 III mitigation models are described in Section I.8 A review of integrated assessment model (IAM) 41 approaches, their components and limitations are present in Section I.9. Sections I.10 and I.11 present 42 comparative tables of key characteristics and measures of national and global models that contributed 43 to the WG III AR6 scenario database. 44 Do Not Cite, Quote or Distribute I-7 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2. Economic frameworks and concepts used in sectoral models and 2 integrated assessment models 3 Several types of ‘full economy’ frameworks are used in integrated assessment models. The general 4 equilibrium framework – often referred to as Computable General Equilibrium (CGE) – represents the 5 economic interdependencies between multiple sectors and agents, and the interaction between supply 6 and demand on multiple markets (Robinson et al. 1999). It captures the full circularity of economic 7 flows through income and demand relationships and feedbacks including the overall balance of 8 payments. Most CGE approaches used are neoclassical supply-led models with market clearing based 9 on price adjustment. Representative agents usually minimize production costs or maximize utility under 10 given production and utility function, although optimal behaviours are no a precondition per se. Most 11 CGE models also include assumptions of perfect markets with full employment of factors although 12 market imperfections and underemployment of factors (e.g. unemployment) can be assumed (Babiker 13 and Eckaus 2007; Guivarch et al. 2011). CGE frameworks can either be static or dynamic and represent 14 pathways as a sequence of equilibria in the second case. 15 Macro-econometric frameworks represent similar sectoral interdependence with balance of payments 16 as general equilibrium, and are sometimes considered a subset of it. They differ from standard 17 neoclassical CGE models in the main aspect that economic behaviours are not micro-founded 18 optimizing behaviours but represented by macroeconomic and sectoral functions estimated through 19 econometric techniques (Barker and Scrieciu 2010). In addition, they usually adopt a demand-led post- 20 Keynesian approach where final demand and investment determine supply and not the other way 21 around. Prices also do not instantaneously clear markets and adjust with lag. 22 Macro-economic growth framework are also full economy approaches derived from aggregated 23 growth models. They are based on a single macroeconomic production function combining capital, 24 labour and sometimes energy to produce a generic good for consumption and investment. They are used 25 as the macroeconomic component of cost benefit IAMs (Nordhaus 1993) and some detailed-process 26 IAMs. 27 The disaggregation of economic actors and sectors and the representation of their interaction 28 differ across full economy frameworks. A main distinction is between models based on full Social 29 Accounting Matrix (SAM) and aggregated growth approaches. On the one hand, SAM-based 30 frameworks – CGE and macro-econometric – follow a multi-sectoral approach distinguishing from 31 several to a hundred of different economic sectors or production goods and represent sector specific 32 value-added, final consumption and interindustry intermediary consumption (Robinson 1989). They 33 also represent economic agents (firms, households, public administration, etc.) with specific behaviours 34 and budget constraints. On the other hand, macro-economic growth frameworks are reduced to a single 35 macro-economic agent producing, consuming and investing a single macroeconomic good without 36 considering interindustry relationships. In some detailed process IAMs, the aggregated growth approach 37 is combined with a detailed representation of energy supply and demand systems that surmises different 38 economic actors and subsectors. However, the energy system is driven by an aggregated growth engine 39 (Bauer et al. 2008). 40 Partial equilibrium frameworks do not cover the full economy but only represent a subset of economic 41 sectors and markets disconnected from the rest of the economy. They basically represent market balance 42 and adjustments for a subset of sectors under ceteris paribus assumptions about other markets (labour, 43 capital, etc.), income, etc. ignoring possible feedbacks. Partial equilibrium frameworks are used in 44 sectoral models, as well as to model several sectors and markets at the same time – e.g. energy and 45 agriculture markets – in energy system models and some detailed process IAMs but still without 46 covering the full economy. Do Not Cite, Quote or Distribute I-8 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 In most models the treatment of economic growth follows Solow or Ramsey growth approach based 2 on the evolution through time of production factors endowment and productivity. Classically, labour 3 endowment and demography are exogenous, and capital accumulates through investment. Partial 4 equilibrium frameworks do not model economic growth but use exogenous growth assumptions derived 5 from growth models. Factors productivity evolution is assumed exogenous in most cases i.e. general 6 technical progress is assumed to be an autonomous process. A few models feature endogenous growth 7 aspects where factor productivity increases with cumulated macroeconomic investment. Models also 8 differ about the content of technical progress and alternatively consider un-biased total factor 9 productivity improvement or labour specific factor augmenting productivity. In multi-sectoral 10 macroeconomic models, economic growth comes with endogenous changes of the sectoral composition 11 of GDP known as structural change. Structural change results from the interplay between 12 differentiated changes of productivity between sectors and of the structure of final demand as income 13 grows (Herrendorf et al. 2014). If general technical progress is mostly assumed exogenous and 14 autonomous at an aggregated level, innovation in relation to energy demand and technical systems 15 follow more detailed specifications in models. Energy efficiency can be assumed an autonomous 16 process at different levels – macroeconomic, sector or technology – or energy technical change can be 17 endogenous and induced as a learning by doing process or as a result of R&D investments (learning- 18 by-searching) (Löschel 2002). 19 Multi-regional models consider interactions between regions through trade of energy goods, non- 20 energy goods and services – depending on model scope – and emission permits in the context of climate 21 policy. For each type of goods, trade is usually represented as a common pool where regions interact 22 with the pool through supply (exports) or demand (imports). A few models consider bilateral trade flows 23 between regions. Traded goods can be assumed as perfectly substitutable between regions of origin 24 (Heckscher-Ohlin assumption) such as is often the case for energy commodities or as imperfectly 25 substitutable (e.g. Armington goods) for non-energy goods. The representation of trade and capital 26 imbalances at the regional level and their evolution through time vary across model and imbalances are 27 either not considered (regional current accounts are balanced at each point in time), or a constraint for 28 intertemporal balance is included (an export surplus today will be balanced by an import surplus in the 29 future) or else trade imbalances follow other rules such as a convergence towards zero in the long run 30 (Foure et al. 2020). 31 Strategic interaction can also occur between regions especially in the presence of externalities such as 32 climate change, energy prices or technology spillovers. Intertemporal models can include several types 33 of strategic interaction: i) a cooperative Pareto optimal solution where all externalities are internalised 34 and based on the maximization of a global discounted welfare with weighted regional welfare (Negishi 35 weights), ii) a non-cooperative solution that is strategically optimal for each region (Nash equilibrium) 36 (Leimbach et al. 2017b), and iii) partially cooperative solutions (Eyckmans and Tulkens 2003; Yang 37 2008; Bréchet et al. 2011; Tulkens 2019), akin to climate clubs (Nordhaus 2015). 38 Models cover different investment flows depending on the economic framework used. Partial 39 equilibrium models compute energy system and/or sectoral (transport, building, industry, etc.) 40 technology specific investment flows associated with productive capacities and equipment. Full 41 economy models compute both energy system and macroeconomic investment, the second being used 42 to increase macroeconomic capital stock. Full economy multi-sectoral models compute sector specific 43 (energy and non-energy sectors) investment and capital flows with some details about the investments 44 goods involved. 45 Full economy models differ in the representation of macro-finance. In most CGE and macro-economic 46 growth frameworks financial mechanisms are only implicit and total financial capacity and investment 47 are constrained by savings. Consequently, investment in a given sector (e.g. low carbon energy) fully 48 crowds-out investment in other sectors. In macro-econometric frameworks, macro-finance is sometimes Do Not Cite, Quote or Distribute I-9 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 explicit, and investments can be financed by credit on top of savings, which implies more limited 2 crowding-out of investments (Mercure et al. 2019). Macro-financial constraints are usually not 3 accounted for in partial equilibrium models. 4 Models compare economic flows over time through discounting. Table I.5summarizes key 5 characteristics of different models assessed in AR6, including the uses of discounting. In cost-benefit 6 analysis (CBA), discounting enables to compare mitigation costs and climate change damage. In the 7 context of mitigation and in cost-effectiveness analysis (CEA), discounting allows comparing 8 mitigation costs over time. 9 In optimization models a social discount rate is used to compare costs and benefits over time. In the 10 case of partial equilibrium optimization models, the objective is typically to minimize total discounted 11 system cost. The social discount rate is then an exogenous parameter, which can be assumed constant 12 or changing (generally decreasing) over time (e.g. Gambhir et al. 2017 where a 5% discount rate is 13 used). In the case of intertemporal welfare optimization models, a Ramsey intertemporal optimization 14 framework is generally used, considering a representative agent who decides how to allocate her 15 consumption, and hence saving, over time subject to a resource constraint. Ramsey (1928) shows that 16 the solution must always satisfy the Ramsey Equation, which provides the determinants of the social 17 discount rate. The Ramsey Equation is given as follows: 18 𝜌 = 𝛿 + 𝜂𝑔𝑡 19 where 𝜌 is the consumption discount rate (aka social discount rate), 𝛿 is the utility discount rate (aka 20 pure time discount rate, or time preferences rate) which is a value judgement that determines the present 21 value of a change in the utility experienced in the future and hence it is an ethical parameter, gt is the 22 growth rate of consumption per capita overtime, and 𝜂 is the elasticity of marginal utility of 23 consumption, which is also a value judgement and hence an ethical parameter. The parameter 𝜂 is also 24 a measure of risk aversion and a measure of society’s aversion to inequality within and across 25 generations. The pure time preference rate is an exogenous parameter, but the social discount rate is 26 endogenously computed by the model itself and depends on the growth rate of consumption per capita 27 over time. Note that more complex frameworks disentangle inequality aversion from risk aversion, and 28 introduce uncertainty, leading to extensions of the social discount rate equation (see for instance Gollier 29 2013) 30 Discounting is also used for ex-post comparison of mitigation cost pathways across models and 31 scenarios. Values typically used for such ex-post comparison are 2%-5% (e.g. Admiraal et al. 2016). 32 Across this report, whenever discounting is used for ex-post comparisons, the discount rate applied is 33 stated explicitly. 34 The choice of the appropriate social discount rate (and the appropriate rate of pure time preference when 35 applicable) is highly debated (see e.g. Arrow et al. (2013), Gollier and Hammitt (2014), Polasky and 36 Dampha (2021)) and two general approaches are commonly used. Based on ethical principles, the 37 prescriptive approach states that the discount rate should reflect how costs and benefits supported by 38 different generations should be weighted. The descriptive approach identifies the social discount rate to 39 the risk free rate of return to capital as observed in the real economy, which generally yields higher 40 values. 41 In CBA the choice of discount rate is crucial for the balance of mitigation costs and avoided climate 42 damages in the long run and a lower discount rate yields more abatement effort and lower global 43 temperature increases (Stern 2006; Hänsel et al. 2020). In CEA, the choice of social discount rate 44 influences the timing of emission reductions to limit warming to a given temperature level. A lower 45 discount rate increases short-term emissions reductions, lowers temperature overshoot, favours 46 currently available mitigation options (energy efficiency, renewable energy, etc.) over future Do Not Cite, Quote or Distribute I-10 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 deployment of net negative emission and distributes mitigation effort more evenly between generations 2 (Emmerling et al. 2019; Strefler et al. 2021b). 3 Outside social discounting for intertemporal optimization, discounting is used in simulation models to 4 compute the lifecycle costs of investment decisions (e.g. energy efficiency choices, choices between 5 different types of technologies based on their levelized costs – LCOE). In this case, the discount rate 6 can be interpreted as the cost of capital faced by investors. The cost of capital influences the merit order 7 of technologies and lower capital cost favours capital intensive technologies over technologies with 8 higher variable costs. Models can reflect regional, sectoral or technology specific cost of capital - 9 through heterogeneous discount rates for lifecycle cost estimates in simulation models (Iyer et al. 2015) 10 or as hurdle rates in energy optimization models (Ameli et al. 2021). In some cases, simulation models 11 may also produce mitigation pathways following the Hotelling principle and assuming that the carbon 12 price rises at the social discount rate (e.g. GCAM scenarios in the SSP study with carbon prices 13 increasing at 5% yearly (Guivarch and Rogelj 2017)). 14 15 3. Energy system modelling 16 In the literature, the energy system models are categorized based on different criteria, such as (a) energy 17 sectors covered, (b) geographical coverage, (c) time resolution, (d) methodology, and (e) programming 18 techniques. In the following sections, examples on different types of energy system models applied in 19 Chapter 6 are presented. 20 3.1. Bottom-up models 21 3.1.1. Modelling electricity system operation and planning with large scale penetration of renewables 22 A number of advanced grid modelling approaches have been developed (Sani Hassan et al. 2018), such 23 as robust optimization (Jiang et al. 2012), interval optimization (Dvorkin et al. 2015), or stochastic 24 optimization (Meibom et al. 2011; Monforti et al. 2014) to optimally schedule the operation of the future 25 low carbon systems with high penetration of variable renewable energies (VRE). Advanced stochastic 26 models demonstrated that this would not only lead to significantly higher cost of system management 27 but may eventually limit the ability of the system to accommodate renewable generation (Badesa et al. 28 2020; Hansen et al. 2019; Perez et al. 2019; Bistline and Young 2019). Modelling tools such as 29 European Model for Power system Investment with Renewable Energy (EMPIRE) (Skar et al. 2016), 30 Renewable Energy Mix for Sustainable Electricity Supply (REMix) (Scholz et al. 2017), European Unit 31 Commitment And Dispatch model (EUCAD) (Després 2015), SWITCH (Fripp 2012), GenX (TNO 32 2021), and Python for Power System Analysis (PyPSA) (Brown et al. 2018) investigated these issues. 33 SWITCH is a stochastic model, in which investments in renewable and conventional power plants is 34 optimized over a multi-year period (Fripp 2012). In GenX the operational flexibility as well as capacity 35 planning is optimized from a system-wide perspective (TNO 2021). PyPSA is an optimization model 36 for modern electricity systems, including unit commitment of generation plants, renewable sources, 37 storage, and interaction with other energy vectors (Brown et al. 2018). 38 Furthermore, advanced modelling tools have been developed for the purpose of providing estimations 39 of system wide inertial frequency response that would assist system operators in maintaining adequate 40 system inertia (Sharma et al. 2011; Teng and Strbac 2017). These innovative models also provided 41 fundamental evidence regarding the role and value of advanced technologies and control systems in 42 supporting cost effective operation of future electricity systems with very high penetration of renewable 43 generation. In particular, the importance of enhancing the control capabilities of renewable generation 44 and applying flexible technologies, such as energy storage (Hall and Bain 2008; Obi et al. 2017; 45 Arbabzadeh et al. 2019), demand side response (DSR), interconnection (Aghajani et al. 2017) and Do Not Cite, Quote or Distribute I-11 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 transmission grid extensions (Schaber et al. 2012) for provide system stability control, is demonstrated 2 through novel system integration models (Sinsel et al. 2020; Lund et al. 2015). 3 A novel modelling framework is proposed to deliver inertia and support primary frequency control 4 through variable-speed wind turbines (Morren et al. 2006) and PVs (Waffenschmidt and Hui 2016; Liu 5 et al. 2017), including quantification of the value of this technology in future renewable generation 6 dominated power grids (Chu et al. 2020). Advanced models for controlling distributed energy storage 7 systems to provide an effective virtual inertia have been developed, demonstrating the provision of 8 virtual-synchronous-machine capabilities for storage devices with power electronic converters, which 9 can support system frequency management following disturbances (Hammad et al. 2019; Markovic et 10 al. 2019). Regarding the application of interconnection for exchange of balancing services between 11 neighbouring power grids, alternative control schemes for High Voltage Direct Current (HVDC) 12 converters have been proposed demonstrating that this would reduce the cost of balancing (Tosatto et 13 al. 2020). 14 3.1.2. Modelling the interaction between different energy sectors 15 Several integrated models have been developed in order to study the interaction between different 16 energy vectors and whole system approaches, such as Integrated Energy System Simulation model 17 (IESM) (NREL 2020), Integrated Whole-Energy System (IWES) (Strbac et al. 2018), UK TIMES (Daly 18 and Fais 2014), and Calliope (Pfenninger and Pickering 2018). 19 IESM is an approach in which the multi-system energy challenge is investigated holistically rather than 20 looking at each of the systems in isolation. IESM capabilities include co-optimization across multiple 21 energy systems, including electricity, natural gas, hydrogen, and water systems. These provide the 22 opportunity to perform hydro, thermal, and gas infrastructure investment and resource use coordination 23 for time horizons ranging from sub-hourly (markets and operations) to multi-years (planning) (NREL 24 2020). 25 IWES model incorporates detail modelling of electricity, gas, transport, hydrogen, and heat systems 26 and captures the complex interactions across those energy vectors. The IWES model also considers the 27 short-term operation and long-term investment timescales (from seconds to years) simultaneously, 28 while coordinating operation of and investment in local district and national/international level energy 29 infrastructures (Strbac et al. 2018). 30 The UK TIMES Model (‘The Integrated MARKAL-EFOM System’) uses linear-programming to 31 produce a least-cost energy system, optimized according to a number of user constraints, over medium 32 to long-term time horizons. It portrays the UK energy system, from fuel extraction and trading to fuel 33 processing and transport, electricity generation and all final energy demands (Taylor et al. 2014; Daly 34 and Fais 2014). The model generates scenarios for the evolution of the energy system based on different 35 assumptions around the evolution of demands, future technology costs, measuring energy system costs 36 and all greenhouse gases (GHGs) associated with the scenario. UKTM is built using the TIMES model 37 generator: as a partial equilibrium energy system and technologically detailed model, is well suited to 38 investigate the economic, social, and technological trade-offs between long-term divergent energy 39 scenarios. 40 Calliope is an open source Python-based toolchain for developing energy system models, focusing on 41 flexibility, high temporal and spatial granularities. This model has the ability to execute many runs on 42 the same base model, with clear separation of model (data) and framework (code) (Pfenninger and 43 Pickering 2018). 44 Do Not Cite, Quote or Distribute I-12 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 3.2. Modelling of energy systems in context of economy 2 To study the impact of low carbon energy systems on the economy, numerous integrated assessment 3 modelling tools (Top-down models) are applied, such as: General Equilibrium Model for Economy- 4 Energy-Environment (GEM-E3) (Capros et al. 2013), ENV-Linkages (Burniaux and Chateau 2010), and 5 Emissions Prediction and Policy Analysis (EPPA) (Chen et al. 2016). 6 GEM-E3 is a recursive dynamic computable general equilibrium model that covers the interactions 7 between the economy, the energy system and the environment. It is especially designed to evaluate 8 energy, climate, and environmental policies. GEM-E3 can evaluate consistently the distributional and 9 macro-economic effects of policies for the various economic sectors and agents across the 10 countries/regions (Capros et al. 2013). 11 The modelling work based on ENV-Linkages (as a successor to the OECD GREEN) provides insights 12 to policy makers in identifying least-cost policies by taking into account environmental issues, such as 13 phasing out fossil fuel subsidies, and climate change mitigation (Burniaux and Chateau 2010). 14 In the EPPA model different processes (e.g., economic and technological), which have impacts on the 15 environment from regional to global at multiple scales is simulated. The outputs of this modeling (e.g., 16 greenhouse gas emissions, air and water pollutants) are provided to the MIT Earth System (MESM), 17 which investigated the interaction between sub-models of physical, dynamical and chemical processes 18 in different systems (Chen et al. 2016). 19 20 3.3. Hybrid models 21 Hybrid models are a combination of macro-economic models (i.e., top-down) with at least one energy 22 sector model (i.e., bottom-up) that could benefit from the advantages of both mentioned approaches. In 23 this regard, linking these two models can be carried out either manually through transferring the data 24 from one model to the other (soft-linking), or automatically (hard-linking) (Prina et al. 2020). In this 25 section, some of these models are presented including World Energy Model (WEM) (IEA 2020a), the 26 National Energy Modelling System (NEMS) (Fattahi et al. 2020). 27 The WEM is a simulation model covering energy supply, energy transformation and energy demand. 28 The majority of the end-use sectors use stock models to characterize the energy infrastructure. In 29 addition, energy-related CO2 emissions and investments related to energy developments are specified. 30 The model is focused on determining the share of alternative technologies in satisfying energy service 31 demand. This includes investment costs, operating and maintenance costs, fuel costs and in some cases 32 costs for emitting CO2 (IEA 2020a). 33 The NEMS is an energy-economy modelling system applied for the U.S.A. through 2030. NEMS 34 projects considers the production, imports, conversion, consumption, and prices of energy, subject to 35 assumptions on macroeconomic and financial factors, world energy markets, resource availability and 36 costs, behavioural and technological choice criteria, cost and performance characteristics of energy 37 technologies, and demographics. NEMS was designed and implemented by the Energy Information 38 Administration (EIA) of the U.S. Department of Energy. NEMS is used by EIA to project the energy, 39 economic, environmental, and security impacts on the United States considering alternative energy 40 policies and assumptions related to energy markets (Fattahi et al. 2020). 41 Do Not Cite, Quote or Distribute I-13 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 4. Building sector models 2 4.1. Models purpose, scope and types 3 GHG emissions and mitigation potentials in the building sector are modelled using either a top-down, 4 a bottom-up or a hybrid approach, which combines both bottom-up and top-down (Figure I.1.). 5 1. The top-down models are used for assessing economic-wide responses of building policies. 6 These models are either economic or technological and have low granularity (Figure I.1.). 7 2. The bottom-up models are data intensive and based on microscopic data of individual end-uses 8 and the characteristics of each component of buildings. Bottom-up models can be either 9 physics-based, also known as engineering models; data-driven, also known as statistical 10 models; or a combination of both, also known as hybrid bottom-up models. Bottom-up models 11 are useful to assess the technico-economic potentials of the overall building stock by 12 extrapolating the estimated energy consumption of a representative set of invidual buildings 13 (Duerinck et al. 2008; Hall and Buckley 2016; Bourdeau et al. 2019) (Figure I.1.). 14 3. Hybrid models used for buildings can be either optimisation or simulation models (Duerinck et 15 al. 2008; Hall and Buckley 2016; Bourdeau et al. 2019) (Figure I.1.). The latter can also be 16 agent-based models and could be combined with building performance models to allow for an 17 assessment of occupants behaviour (Sachs et al. 2019a; Papadopoulos and Azar 2016; Niamir 18 et al. 2020). Hybrid models are used for exploring the impacts of resource constraints and for 19 investigating the role of specific technological choices as well as for analysing the impact of 20 specific building policies. 21 The use of Geographical Information Systems (GIS) layers (Reinhart and Cerezo Davila 2016) 22 combined to machine learning techniques (Bourdeau et al. 2019) allows creating detailed datasets of 23 building characteristics while optimising the computing time. Thus, leading to a better representation 24 of energy demand of buildings and a more accurate assessment of GHG mitigation potential. 25 GHG emissions Hybrid Top-down Bottom-up Optimisation Simulation Hybrid Economic Technological bottom-up Statistical Engineering 26 27 Figure I.1. Modelling approaches of GHG emissions used in the building sector 28 29 4.2. Representation of energy demand and GHG emissions 30 Comprehensive models represent energy demand per energy carrier and end-use for both residential 31 and non-residential buildings, for different countries or set of countries, further disaggregated across 32 urban/rural and income groups. Drivers of energy demand considered include population, the floor area Do Not Cite, Quote or Distribute I-14 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 per capita, appliances ownership and to some extent occupants’ behaviour in residential buildings. The 2 former being included in top-down, hybrid and bottom-up models while the latter is, usually included 3 in bottom-up and agent-based models (IEA 2021; Niamir et al. 2020). In non-residential buildings, 4 value-added is considered among the drivers. 5 GHG emissions from buildings are usually modelled on the basis of the estimated energy demand per 6 energy carrier and appropriate emissions factors. The purpose of most building models is to assess the 7 impact of mitigation measures on energy demand in the use phase of buildings and for a given 8 assumption on the per-capita floor area and technological improvement (Pauliuk et al. 2021b) and (IEA 9 2021). After decades of ignoring material cycles and embodied emissions (Pauliuk et al. 2017), few 10 IAMs are now including material stocks and flows (Zhong et al. 2021; Deetman et al. 2020; IEA 2021). 11 However, the top-down nature of these models and the modelling methodology of embodied emissions, 12 which are added onto the emissions estimated in the use phase, questions the policy relevance of these 13 estimates. As of today, the resource efficiency and climate change (RECC) scenario (Pauliuk and 14 Heeren 2021; Pauliuk et al. 2021b; Fishman et al. 2021; Hertwich et al. 2020) is the only global scenario 15 identified which includes measures to limit, at the first place, embodied emissions from buildings. The 16 scenario is modelled using the bottom-up ODYM-RECC model. 17 18 4.3. Representation of mitigation options 19 The assessment conducted in Chapter 9 was based on the SER (Sufficiency, Efficiency, Renewable) 20 framework with sufficiency being all the measures and daily practices which avoid, at the first place, 21 the demand for energy, materials, water, land and other natural resources over the life cycle of buildings 22 and appliances/equipment, while providing decent living standard for all within the planetary 23 boundaries. By contrast to efficiency, sufficiency measures do not consume energy in the use phase. 24 Efficiency improvement of the building envelope and appliances/equipment are the main mitigation 25 options considered in the existing models/scenarios. They are, usually, combined with market-based 26 and information instruments and to some extent with behaviour change. As of today, Grubler et al. 27 (2018), (Pauliuk et al. 2021b), Kuhnhenn et al. (2020), Millward-Hopkins et al. (2020), Kikstra et al. 28 (2021), van Vuuren et al. (2021) are the only six global models/scenarios to include sufficiency 29 measures, out of which detailed data were available only for two scenarios (Pauliuk et al. 2021b; van 30 Vuuren et al. 2021). 31 32 4.4. Representation of climate change impacts 33 In total, 931 scenarios were submitted to AR6 scenario database out of which only two scenarios 34 provided detailed data allowing for an assessment of climate change impacts based on the SER 35 framework considered in the building chapter. Additional 78 bottom-up models/scenarios were gathered 36 (Table I.1.). Mitigation potentials from these scenarios are assessed using either a decomposition 37 analysis (Chapter 9, Section 9.3.) or an aggregation of bottom-up potential estimates for different 38 countries into regional and then global figures (Chapter 9, Section 9.6.). 39 Scenarios considered in the illustrative mitigation pathways included in Chapter 3 were assessed, 40 compared to current policy scenario. The assessment was possible for only the combined direct CO 2 41 emissions for both residential and non-residential buildings due to lack of data on other gases as well 42 as on indirect and embodied emissions. The assessment shows mitigation potentials, compared to 43 current policies scenarios, at a global level ranging from 9% to 13% by 2030 and from 58% to 89% in 44 2050 (Figure I.2-b). Do Not Cite, Quote or Distribute I-15 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 There are great discrepancies in the projected potentials by the IAMs across regions and scenarios. In 2 the deep electrification and high renewable scenario, emissions in Africa are projected to increase by 3 88% by 2030, followed by a decrease of 97% by 2050 compared to current policies scenario. Similarly, 4 in the sustainable development scenario, emissions in developing Asia are projected, compared to 5 current policies scenario, to increase by 56% by 2030, followed by a decrease of 75% by 2050. Such 6 variations in emissions over two decades in the developing world raise questions about the policy 7 relevance of these scenarios. In developed countries, emissions are projected to go down in all regions 8 across all scenarios, except in SSP2 scenario in Asia-Pacific, where emissions are projected to increase 9 by 18% by 2030 followed by a decrease of 25% by 2050, compared to current policies scenario. It is 10 worth noting that, across all scenarios, Eastern Asia is the region with the lowest estimated mitigation 11 potential compared to the current policies (Figure I.2-b). 12 13 4.5. Representation of sustainable development dimensions 14 Link to sustainable development goals is not always explicit in buildings models/scenarios. However, 15 some models include requirements to ensure the access to decent living standard for all Kikstra et al. 16 (2021), Millward-Hopkins et al. (2020), Grubler et al. (2018) or to specifically meet the 2030 SDG 7 17 goal (IEA 2020a, 2021). 18 Do Not Cite, Quote or Distribute I-16 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII World 4000 3500 CO2 emisisons (Direct ) MtCO2 3000 -9% -8% -13% 2500 2000 -58% 1500 1000 -82% -89% 500 0 2030 2050 2030 2050 2030 2050 2030 2050 NGFS2_Current Policies SSP2_openres_lc_50 SusDev_SDP-PkBudg1000 DeepElec_SSP2_ HighRE_Budg900 *The percentages correspond to the increase or decrease in relation to the same year with the Current Policies Scenario as a baseline. 1 2 a) Global 3 4 5 b) Regional 6 Figure I.2. GHG mitigation potentials is scenarios considered in the illustrative mitigation pathways 7 considered in Chapter 3. Do Not Cite, Quote or Distribute I-17 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2 4.6. Models underlying the assessment in Chapter 9 3 The AR6 scenario database received 101 models, with a building component, out of which 96 were 4 IAM models and five building specific models. This is equivalent to 931 scenarios. After an initial 5 screening, quality control and further vetted to assess if they sufficiently represented historical trends 6 and climate goals, 43 models (42 IAMs and 1 building specific model) were kept for the assessment. 7 Thus, reducing the number of scenarios to assess to 554. The unvetted scenarios are still available in 8 the database. After a final screening based on the SER (Sufficiency, Efficiency, Renewable) framework, 9 only two IAMs were kept. Given the top-down nature of IAMs and their weaknesses in assessing 10 mitigation measures, especially sufficiency measures, 78 bottom-up models with technological 11 representation have been included in the assessment (Table I.1.). These additional bottom-up models 12 were not submitted to AR6 scenario database. However, scenario owners supplied Chapter 9 with the 13 underlying assumptions and data. Do Not Cite, Quote or Distribute I-18 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Table I.1. Models underlying the assessment in Chapter 9. Model name/Institution Example of publications Model description Geographic scope Building type included Energy demand using the model World Energy Model A simulation model Global Residential and non- The building module includes a (IEA 2020a);(IEA 2021) (WEM)/International with detailed bottom-up residential stock model with detailed Energy Agency (IEA) building stock model technologies, end uses and energy carriers. Activity variables such as floor area and appliance ownership are projected by end-use. A cost- based approach, influenced by policy and other constrains, is used to allocate between almost 100 technologies. Energy demand projections are based on country-level historical data for both residential and non- residential buildings. The buildings module is integrated within the wider World Energy Model. IMAGE 3.2 model/ A modular Integrated Global Residential and non- Energy demand is calculated as (van Vuuren et al. 2021) Assessment Model residential buildings a function of household Netherlands using a simulation expenditures and population Environmental model for energy growth, disaggregating across Assessment Agency demand urban/rural and income groups. The model includes a building stock model (residential) with a detailed description of end-uses, energy carrier use and building technologies for both residential Do Not Cite, Quote or Distribute I-19 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII and non-residential buildings. A scenario analysis assessing assumptions on lifestyle changes has also been conducted. Resource Efficiency and Bottom-up building Global Residential buildings Energy demand is calculated by (Pauliuk et al. 2021b); (Hertwich et al. 2020); (Pauliuk Climate Change (RECC) stock-flow model the model BuildME, a physical et al. 2021a); (Fishman et al. 2021); (Pauliuk and model. Research estimating material and model using the EnergyPlus Heeren 2021) Institutions: Norwegian energy flows associated simulation engine, University of Science & with housing stock incorporating country/region- Technology, and growth, driven by input specific projections of envelope University of Freiburg. parameters of and equipment efficiency Funding Institutions: population and floor UNEP and International area per capita Resource Panel A total of 77 bottom-up Bottom-up technology- Three global (all Residential and/or non- In most cases, energy demand (Alaidroos and Krarti 2015; Bashmakov 2017; Brugger models out of which 67 rich models with sufficiency models), residential buildings was modelled by multiplying et al. 2021; Bürger et al. 2019; Butler et al. 2020; Calise were technology-rich detailed building and six regional (regions unit of energy consumption of et al. 2021; Chaichaloempreecha et al. 2017; and 10 sufficiency- other technology stock here refer to regions technologies/product/buildings Colenbrander et al. 2019; Csoknyai et al. 2016; de la focussed models including several with stocks of corresponding Rue du Can et al. 2019, 2018; de Melo and de Martino countries), two technologies/products and/or Jannuzzi 2015; Department of Environmental Affairs subnational, and the buildings at national level. The 2014; Dioha et al. 2019; Duscha et al. 2019; Energetics rest national stocks of buildings and/or 2016; Gagnon, Peter, Margolis, Robert, Melius, technologies/products rely on Jennifer, Phillips, Saleb, Elmore 2016; González- very detailed stock modelling in Mahecha et al. 2019; Grande-Acosta and Islas- the future relying on such Samperio 2020; Horváth et al. 2016; Iten R., Jakob M., statistics in the past. The Catenazzi G, Reiter U., Wunderlich A. 2017; Kamal et potential is demonstrated al. 2019; Khan et al. 2017; Krarti 2019; Krarti et al. replacing the business-as-usual 2017; Kusumadewi and Limmeechokchai 2015, 2017; technologies and practices with Kwag et al. 2019; Markewitz et al. 2015; Merini et al. demonstrated best available or 2020; Minami et al. 2019; Momonoki et al. 2017; Nadel commercially feasible 2016; Novikova et al. 2018a,b; Filippi Oberegger et al. technologies and practices. The 2020; Oluleye et al. 2018, 2016; Onyenokporo and studies rely on all, the Ochedi 2019; Ostermeyer, Y.; Camarasa, C.; Naegeli, Do Not Cite, Quote or Distribute I-20 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII combination, or either of the C.; Saraf, S.; Jakob, M.; Hamilton, I; Catenazzi 2018; following mitigation options: Ostermeyer et al. 2019a; Ostermeyer, Y.; Camarasa, C.; the construction of new high- Saraf, S.; Naegeli, C.; Jakob, M.; Palacios, A, performance buildings using Catenazzi 2018; Ostermeyer et al. 2018, 2019b; Ploss building design, forms, and et al. 2017; Prada-hernández et al. 2015; Radpour et al. passive construction methods; 2017; Rosas-Flores and Rosas-Flores 2020; Roscini et the thermal efficiency al. 2020; Sandberg et al. 2021; Streicher et al. 2017; improvement of building Subramanyam et al. 2017a,b; Sugiyama et al. 2020b; envelopes of the existing stock; Tan et al. 2018; Timilsina et al. 2016; Toleikyte et al. the installation of advanced 2018; Trottier 2016; Wakiyama and Kuramochi 2017; HVAC systems, equipment and Wilson et al. 2017b; Xing et al. 2021; Yeh et al. 2016; appliances; the exchange of Yu et al. 2018; Zhou et al. 2018; ADB 2017; Zhang et lights, appliances, and office al. 2020; Mirasgedis et al. 2017)(Grubler et al. equipment, including ICT, 2018)(Millward-Hopkins et al. 2020)(Levesque et al. water heating, and cooking; 2019)(Bierwirth and Thomas 2019)(Roscini et al. active and passive DSM 2020)(Cabrera Serrenho et al. 2019) (Roca-Puigròs et measures; as well as onsite al. 2020)(Negawatt 2017)(Virage-Energie Nord-Pas- production and use of de-Calais. 2016) renewable energy. Many bottom-up studies considered the measures as an integrated package due to their technological complementarity and interdependence, rather than the penetration of individual technologies applied in an incremental manner in or to these buildings. Do Not Cite, Quote or Distribute I-21 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 5. Transport models 2 5.1. Purpose and scope of models 3 GHG emissions from transport are largely a function of travel demand, transport mode, and 4 transport technology and fuel. The purpose of transportation system models is to describe how future 5 demand for transport can be fulfilled through different modes and technologies under different climate 6 change mitigation targets or policies. Within a given transport mode, technologies differ by efficiency 7 and fuel use. 8 Common components of transportation energy systems models mirror these main drivers of GHG 9 emissions. Most models will also quantify how much movement occurs, or the travel demand 10 associated with each mode. Models commonly quantify demand through transportation mode (e.g. 11 active transit, passenger vehicles, trucks, boats, planes, etc.) or how movement occurs (e.g. passenger 12 travel distance p.km and freight distance t.km). Higher fidelity models provide more nuanced 13 breakdowns of demand by trips of various lengths such as short-, medium-, and long-distance trips or 14 by region (e.g. kilometres or t.km per region). The scope of the model often determines how much 15 information it provides on where and when movement occurs. While larger scale models typically 16 provide aggregate travel demand, higher resolution travel demand models can be integrated into 17 transportation system models and provide much more information on origin and destination of trips, 18 when and where trips occur, and the route of travel taken. This level of detail is not often characterised 19 in the output of system models but can be employed as a “base” model to determine how travel occurs 20 before aggregation (Edelenbosch et al. 2017a; Yeh et al. 2017). 21 A key distinguishing feature between different model types is how they control the above components. 22 Our review of the transport energy system models can be broadly divided into three main categories: i) 23 optimisation models, ii) simulation models, and iii) accounting and exploratory models. 24 i) Optimization models: Identify least cost pathways to meet policy targets (such as CO2 25 emission targets of transport modes or economy-wide) given constraints (such as rate of 26 adoption of vehicle technologies or vehicle efficiency standards). For example MessageIX- 27 TransportV5 (Krey et al. 2016) and TIMES (Daly et al. 2014). 28 29 ii) Simulation models: Simulate behaviour of consumers and producers given prices, policies, 30 and other factors by using parameters calibrated to historically observed behaviours such 31 as demand price elasticity and consumer preferences. For example models by Barter et al. 32 (2015), Brooker et al. (2015) and Schäfer (2017). 33 34 iii) Accounting and exploratory models: Track the outcomes (such as resources use and 35 emissions) of key decisions (such as the adoption of advanced fuels or vehicle technologies) 36 that are based on what-if scenarios. The major difference between accounting models 37 versus optimisation and simulation models are that key decision variables such as new 38 technologies adoptions typically follow modeler’s assumptions as opposed to being 39 determined by mathematical formulations as in optimisation and simulation models. See 40 models in Fulton et al. (2009), IEA (2020a), Gota et al. (2019) and Khalili et al. (2019). 41 42 Due to the model types’ relative strengths and weaknesses, they are commonly applied to certain 43 problem types (Table I.2.). Models can do forecasting, which makes projections of how futures 44 may evolve, or backcasting, which makes projections of a future that meets a predefined goal such 45 as a policy target of 80% reduction in GHG emissions from a historical level by a certain year. 46 Models often are also used to explore what-if questions, to confirm the feasibility of certain Do Not Cite, Quote or Distribute I-22 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 assumptions/outcomes, and to quantify the impacts of a change such as a policy under different 2 conditions. Enhancing fuel efficiency standards, banning internal combustion engines, setting fuel 3 quality standards, and the impacts of new technologies are the typical examples of problem types 4 analysed in energy system models. 5 6 Table I.2. Taxonomy of transport models by method (modelling type) and application (problem type). Optimization Simulation Accounting Heuristic Problem Type model model model model Backcasting x x Forecasting x x x Exploring feasibility space x x x Impact analysis x x x 7 8 While these four model types drive the component dynamics in different ways, they commonly include 9 modules that include: learning and diffusion (via exogenous, e.g. autonomous learning, or endogenous 10 learning regarding costs and efficiency: i.e. cost decreases and/or efficiency increases as a function of 11 adoption, and increased diffusion due to lower costs) (Jochem et al. 2018), stock turnover (the 12 performance and characteristics of vehicle fleets including survival ages, mileages, fuel economies and 13 loads/occupancy rates are tracked for each new sales/vehicle stocks), consumer choice (theories of how 14 people invest in new technology and utilize different mode of transport based on their individual 15 preferences given the characteristics of mode or technology) (Daly et al. 2014; Schäfer 2017), or other 16 feedback loops (Linton et al. 2015). 17 IAMs (Krey et al. 2016; Edelenbosch et al. 2017a) are typically global in scope and seek to solve for 18 feasible pathways meeting a global temperature target (Annex III.I.9). This implies solving for 19 mitigation options within and across sectors. In contrast, global/national transport energy system 20 models (GTEM/NTEMs) typically only solve for feasible pathways within the transport sector (Yeh et 21 al. 2017). The range of feasible pathways can be determined through optimisation, simulation, 22 accounting and exploratory methods as we explained in Table I.2. Some GTEMs are linked to IAMs 23 model (Krey et al. 2016; Edelenbosch et al. 2017a; Roelfsema et al. 2020). The key difference between 24 IAMs and GTEM or NTEMs is whether the transportation systems is integrated with the rest of the 25 energy systems specifically regarding energy and fuel productions and use, fuel prices, economic 26 drivers such as GDP, and mitigation options given a policy goal. IAMs can endogenously determine 27 these factors because the transport sector is just one of many sectors captured by the IAM. While this 28 gives IAMs certain advantages, IAMs sacrifice resolution and complexity for this broader scope. For 29 example, most IAMs lack a sophisticated travel demand model that reflects the heterogeneity of 30 demands and consumer preferences, whereas GTEM/NTEMs can incorporate greater levels of details 31 regarding travel demands, consumer choices, and the details of transport policies. Consequently, what 32 GTEM/NTEMs lack in integration with other sectors they make up through more detailed analyses of 33 travel patterns, policies, and impacts (Yeh et al. 2017). 34 Several noteworthy recent active research areas in long-term transportation energy systems modelling 35 involves the consideration of infrastructure investment and consumer acceptance for non-fossil fuel 36 vehicles including charging for electric vehicles (Statharas et al. 2021; Jochem et al. 2019) and 37 refuelling stations for hydrogen vehicles (Rose and Neumann 2020); and the greater integration of the 38 electric, transport, residential, and the industrial sectors in fuel production, storage, and utilization 39 (Rottoli et al. 2021; Lester et al. 2020; Bellocchi et al. 2020; Olovsson et al. 2021). While 40 national/regional transport energy models have the advantage of exploring these relationships in greater 41 spatial, temporal, and policy details for specific country/regions (Jochem et al. 2019; Rottoli et al. 2021; 42 Statharas et al. 2021; Lester et al. 2020; Bellocchi et al. 2020), the IAMs have the advantage of Do Not Cite, Quote or Distribute I-23 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 examining these interactions across the entire economy at the global level (Brear et al. 2020; Rottoli et 2 al. 2021). 3 4 5.2. Inventory of transportation models included in AR6 5 The global/national transport energy system models included in the transportation chapter (Chapter 10) 6 are listed below in Table I.3. 7 8 Table I.3. GTEM/NTEMs models evaluated in Chapter 10. Economy- Model name Organisation Scope Resolution Period Method wide Mobility International Global Country 2050 Soft-link Accounting model Energy groups model (MoMo) Agency (IEA) Global International Global Country 2050 No Accounting Transportation Council on groups model Roadmap Clean Transportation (ICCT) MESSAGE- International Global Country 2100 Yes Optimization Transport V.5 Institute for groups model Applied Systems Analysis (IIASA) GCAM Pacific Global Country 2100 Yes Partial Northwest groups equilibrium National model Laboratory (PNNL) 9 10 11 6. Industry sector models 12 6.1. Types of industry sector models 13 Industry sector modelling approaches can vary considerably from one another. As other types of 14 models, industry sector models a key characteristic related to their geographical scope. While IAMs are 15 often global in scope, many bottom-up sector models are limited to individual countries or regions. The 16 models’ system boundaries also differ, with some models fully considering the use of energy for 17 feedstock purposes and other models focussing only on the use of energy for energetic purposes. 18 Differences between models also exist in regard to the differentiation between the industry sector and 19 the energy transformation sector, concerning e.g. the refineries and industrial power plants. 20 21 6.2. Representation of demand for industrial products 22 Industry sector models vary in regard to their representation of demand for industrial goods or products. 23 A more detailed representation of demand in a model allows for a more explicit discussion of different Do Not Cite, Quote or Distribute I-24 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 types of drivers of industrial demand and therefore a more detailed representation of demand side 2 strategies such as material recycling, longer use of products or sharing of products. 3 Particularly, in bottom-up models of the industry sector, demand for industrial products is often 4 considered in more detail than in top-down models by taking more drivers into account. These drivers 5 can be inter alia population, gross value added, construction activity, transport activity, but also changes 6 in material efficiency, recycling rates and scrap rates as well as product use efficiency (e.g. through 7 longer use of products or sharing of products) (Fleiter et al. 2018; Material Economics 2019; IEA 8 2020b). 9 10 6.3. Representation of mitigation options - mitigation options, how their uptake is 11 represented, how potentials and costs are represented 12 In most top-down IAMs, some energy-intensive sectors such as iron and steel or cement are included 13 separately at least in a generalised manner, but typically few if any sector-specific technologies are 14 explicitly represented. Instead, energy efficiency improvements in the industry sector and its subsectors 15 are often either determined by exogenous assumptions or are a function of energy prices. Likewise, fuel 16 switching occurs primarily as a result of changes in relative fuel prices, which in turn are influenced by 17 CO2 price developments. In IAMs that include specific technologies, fuel switching can be constrained 18 based on the characteristics of those technologies, while in IAMs with no technological detail more 19 generic constraints on fuel switching in the industry sector are embedded (Edelenbosch et al. 2017b). 20 In bottom-up models, individual technological mitigation options are represented in detail, especially 21 for energy-intensive sectors such as iron and steel, cement and chemicals. Typically, for each considered 22 technology not only specific energy demand but also investment and operating costs are included in 23 these models. Investment costs can change over time, either based on an exogenous assumption or on 24 an endogenized process such as a learning rate. While bottom-up models often consider technology- 25 specific learning, IAMs cover technological progress in a more general way associated to industry 26 branches. The uptake of new technologies is typically restricted in bottom-up models, for example by 27 assuming a minimum lifetime for existing stock or by assuming S-shaped diffusion curves (Fleiter et 28 al. 2018). The industrial sector models included in the industry chapter (Chapter 11) are listed in Table 29 I.4. 30 31 Do Not Cite, Quote or Distribute I-25 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Table I.4 Models underlying specific assessments in Industry Sector (Chapter 11). Model Model description Geo- Industrial Demand for industrial Examples of name and graphic sectors products publications institution scope included/ using the distinguished model Industry The bottom-up industry Global Aluminium, Demand for industrial (IEA 2020b, sector sector model is one of iron and steel, products is derived based 2021) model of four soft-linked models chemical and on country-level the ETP making up the ETP petrochemical, historical data on per model model: The four models cement, pulp capita consumption. This are an energy supply and paper and per capita consumption is optimization model and other industry. projected forward by three end-use sector using population models (transport, projections and industry industry, buildings). value-added projections. Technologies and fuels Demand for materials is in the industry sector derived by also taking the model are chosen based build-up of material on cost optimization. stocks into account. World Simulation model Global See ETP See ETP model (IEA 2020a, Energy consisting inter alia of model 2021) Model technologically (IEA) detailed bottom-up representations of several industry sectors. Material Modelling tool European Steel, Demand for industrial (Material Economics consisting of several Union chemicals products is derived based Economics modelling separate bottom-up (plastics & on scenarios of future 2019) framework models. ammonia), activity levels in key cement segments such as construction, mobility and food production. Separate models additionally explore opportunities for improving materials efficiency and increasing materials circulation. 2 3 4 6.4. Limitations and critical analysis 5 Aggregated, top-down models of the industry sector, as used in most IAMs, are typically calibrated 6 based on long-term historical data, for example on the diffusion of new technologies or on new fuels. 7 These models are therefore able to implicitly consider real-life restrictions of the whole sector that 8 bottom-up models (with their focus on individual technologies) may not fully take into account. These 9 restrictions may arise from inter alia delays in the construction of infrastructure or market actors 10 possessing incomplete information about new technologies. Furthermore, as IAMs also model the Do Not Cite, Quote or Distribute I-26 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 climate system, these models can principally take into account potential repercussions of climate change 2 impacts on the growth rate and structure of economies. 3 However, a downside of top-down models is that they are typically limited in their representation of 4 individual technologies and processes in the industry sector and particularly of technology-driven 5 structural change. This lack of technological detail limits the usefulness of these models to analyse 6 technology-specific and sector-specific mitigation measures and related policies. Top-down models 7 also tend to have a relatively aggregated representation of industrial energy demand, meaning demand- 8 side mitigations strategies such as recycling, product-service efficiency and demand reduction options 9 are difficult to assess with these models (Pauliuk et al. 2017). 10 In contrast, technology-rich bottom-up models allow detailed analysis of the potential of new 11 technologies, processes and fuels in individual industrial sectors to reduce GHG emission. Their often- 12 detailed analysis of the demand side allows demand-side mitigation strategies to be evaluated. 13 Furthermore, radical future changes in technology, climate policy or social norms can more easily be 14 reflected in bottom-up models than in top-down models which are calibrated on past observations. Both 15 types of models are typically not able to account for product substitution (e.g. steel vs. plastics) arising 16 from changing production cost differentials or changing product quality due to new production 17 processes. In principle, technology rich input-output models could fill this gap. 18 19 7. Land use modelling 20 Land use related IAM modelling results as presented in Chapter 7 are based on comprehensive land- 21 use models (LUMs) that are either integrated directly, or through emulators into the integrated 22 assessment framework. Given the increasing awareness of the importance of the land use sector to 23 achieve ambitious climate mitigation targets, LUMs and their integration into IAMs systems was one 24 of the key innovations to the integrated assessment over the past decade to allow for an economy wide 25 quantification of climate stabilization pathways. 26 LUMs allow to project developments in the land use sector over time and assess impacts of mitigation 27 policies on different economic (markets, trade, prices, demand, supply etc.) and environmental (land 28 use, emissions, fertiliser, irrigation water use, etc.) indicators. The following models submitted 29 scenarios to the AR6 database: AIM (Fujimori et al. 2014, 2017; Hasegawa et al. 2017), EPPA (Chen 30 et al. 2016), GCAM (Calvin et al. 2019), IMAGE (Stehfest et al. 2014), MERGE, MESSAGE- 31 GLOBIOM (Fricko et al. 2017; Havlík et al. 2014; Huppmann et al. 2019), POLES (Keramidas et al. 32 2017), REMIND-MAgPIE (Dietrich et al. 2019; Kriegler et al. 2017), WITCH (Emmerling et al. 2016). 33 34 7.1. Modelling of land use and land use change 35 LUMs represent different land use activities for managed land (agriculture including cropland and 36 pastures, managed forests, and dedicated energy crops) while natural lands (primary forests, natural 37 grasslands, shrubland, savannahs etc.) act as land reserve that can be converted to management 38 depending on other constraints (Popp et al. 2014a; Schmitz et al. 2014). Typically, the agricultural 39 sector has the greatest level of detail across land use sectors. LUMs include different crop- and livestock 40 production activities, some even at the spatially explicit level and differentiated by production system 41 (Havlík et al. 2014; Weindl et al. 2015). Forestry is covered with varying degree of complexity across 42 LUMs. While some models represent only afforestation/deforestation activities dynamically, others 43 have detailed representation of forest management activities and/or forest industries (Lauri et al. 2017). 44 The models endogenously determine the land allocation of different land use activities as well as land Do Not Cite, Quote or Distribute I-27 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 use changes according to different economic principles (land rent, substitution elasticities etc.) and/or 2 considering biophysical characteristics such as land suitability (Weindl et al. 2017; Schmitz et al. 2014). 3 4 7.2. Demand for food, feed, fibre and agricultural trade 5 LUMs project demand for food, feed, other industrial or energy uses for different agriculture and 6 forestry commodities over time. While partial equilibrium models typically use reduced-form demand 7 functions with greater level of detail at the commodity level, however limited agriculture and forestry, 8 CGE models represent demand starting from utility functions from which it is possible to derive demand 9 functions, and functional forms for income and price elasticities however for a more limited set of 10 agricultural and forestry commodities but with full coverage of all economic sectors (Valin et al. 2014; 11 von Lampe et al. 2014). Over time, demand for food, feed, and other industrial uses is projected 12 conditional on population and income growth while bioenergy demand is typically informed in PE 13 models by linking with IAMs/energy systems models, and is usually endogenous in CGE/IAMs 14 (Hasegawa et al. 2020). Depending on the model, demand projections are sensitive to price changes 15 (Valin et al. 2014). International trade is often represented in LUMs using either Armington or spatial 16 equilibrium approaches (von Lampe et al. 2014). 17 18 7.3. Treatment of land-based mitigation options 19 Two broad categories of land-based mitigation options are represented in LUMs: i) reduction of GHG 20 (CO2, CH4 and N2O) emissions from land use, ii) carbon sink enhancement options including biomass 21 supply for bioenergy. Each of these categories is underpinned by a portfolio of mitigation options with 22 varying degree of complexity and parameterisation across LUMs. The representation of mitigation 23 measures is influenced on the one hand, by the availability of data for its techno-economic 24 characteristics and future prospects as well as the computational challenge, e.g. in terms of spatial and 25 process detail, to represent the measure, and on the other hand, by structural differences and general 26 focus of the different LUMs, and prioritization of different mitigation options by the modelling teams. 27 While GHG emission reduction and CO2 sequestration options such as afforestation, are typically 28 covered directly in LUMs (Hasegawa et al. 2021), carbon sequestration from biomass supplied for 29 bioenergy coupled with carbon sequestration (BECCs) is usually not accounted for in LUMs but in the 30 energy sector and hence is taken care of directly in the IAMs. Yet, LUMs provide estimates of available 31 biomass for energy production and the impacts of its production. 32 33 7.3.1. Treatment of GHG emissions reduction 34 Agricultural non-CO2 emissions covered in LUMs include CH4 from enteric fermentation, manure 35 management and cultivation of rice paddies, and N2O emissions from soils (fertilizer and manure 36 application, crop residues) and manure management and are based on IPCC accounting guidelines 37 (IPCC 2019a). For each of those sources, LUMs typically represent a (sub)set of technical, structural 38 and demand side mitigation options. Technical options refer to technologies such as anaerobic digesters, 39 feed supplements or nitrogen inhibitors that are either explicitly represented (Frank et al. 2018) or 40 implicitly via the use of MACCs (Beach et al. 2015; Harmsen et al. 2019; Lucas et al. 2007). Emission 41 savings from structural changes refer to more fundamental changes in the agricultural sector for 42 example through international trade, production system changes or reallocation and substitution effects 43 (Havlík et al. 2014). Demand side options include dietary changes and reduction of food waste (Mbow 44 et al. 2019; Rosenzweig et al. 2020; Springmann et al. 2016; Ivanova et al. 2020; Ritchie et al. 2018; 45 Creutzig et al. 2018; Clark et al. 2020; Popp et al. 2010; Frank et al. 2019). For the forest sector, 46 emission reduction options are mainly targeting CO2 from deforestation (Rochedo et al. 2018; Eriksson Do Not Cite, Quote or Distribute I-28 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2020; Overmars et al. 2014; Bos et al. 2020; Hasegawa et al. 2017; Doelman et al. 2020). 2 Mitigation/restoration options for wetlands to reduce emissions from drained organic soils are typically 3 not represented in LUMs (Humpenöder et al. 2020). 4 There are significant differences between UNFCCC nationally reported GHG inventories and analytical 5 global land use models. According to Grassi et al. (2017), this discrepancy results in a 3GtCO2e 6 difference in estimates between country reports and global models. The difference relies on different 7 methods to classify and assess managed forests and its forest management fluxes (Houghton et al. 2012; 8 Pongratz et al. 2014; Tubiello et al. 2015; Smith et al. 2014; Grassi et al. 2017, 2021). While global 9 models account for GHG emissions from indirect human induced effects and natural effects in 10 unmanaged land, country only consider fluxes of land use and land use change in managed land. In 11 order to produce policy relevant land use model exercises, reconciling these differences is needed by 12 harmonising definitions and approaches of anthropogenic land and the treatment of indirect 13 environmental change (Grassi et al. 2017). 14 15 7.3.2. Treatment of terrestrial carbon dioxide removal options including biomass supply for 16 bioenergy 17 Terrestrial Carbon Dioxide Removal (tCDR) options are only partially included in LUMs and mostly 18 rely on afforestation and bioenergy with CCS (BECCS) (Smith et al. 2019; Fuss et al. 2014, 2018; Minx 19 et al. 2018; Butnar et al. 2020). Especially some nature-based solutions (Griscom et al. 2017) such as 20 soil carbon management (Paustian et al. 2016) which have the potential to alter the contribution of land- 21 based mitigation in terms of timing, potential and sustainability consequences are only recently 22 becoming implemented in LUMs (Frank et al. 2017; Humpenöder et al. 2020). The representation of 23 bioenergy feedstocks varies across models but typically LUMs have comprehensive representation of a 24 series of crops (starch, sugar, oil, wood/lignocellulosic feedstocks) or residues/byproducts that can be 25 used for liquid and solid bioenergy production (Hanssen et al. 2019). 26 27 7.4. Treatment of environmental and socio-economic impacts of land use 28 Aside reporting the implications on AFOLU GHG emissions, LUMs can provide a set of environmental 29 and socioeconomic impact indicators to assess the quantified climate stabilisation pathways in a broader 30 sustainable development agenda (Frank et al. 2021; Obersteiner et al. 2016; Soergel et al. 2021; van 31 Vuuren et al. 2019, 2015). These indicators typically span from land use area developments (Popp et 32 al. 2017; Stehfest et al. 2019), fertilizer use, irrigation water use and environmental flows (Bonsch et 33 al. 2015; Pastor et al. 2019; Chang et al. 2021; de Vos et al. 2021), and on biodiversity (Leclère et al. 34 2020; Marquardt et al. 2021), to market impacts on commodity prices and food consumption, or impact 35 on undernourishment (Fujimori et al. 2019a; Hasegawa et al. 2018; Doelman et al. 2019; Hasegawa et 36 al. 2020; Soergel et al. 2021). 37 38 8. Reduced complexity climate modelling 39 Climate model emulators (often referred to as reduced complexity or simple climate models) are used 40 to integrate the WG I knowledge of physical climate science in WG III assessment. Hence, emulators 41 are used to assess the climate implications of the GHG and other emissions trajectories that IAMs 42 produce (van Vuuren et al. 2008; Rogelj et al. 2018a; Clarke et al. 2014; Rogelj et al. 2011; Schaeffer 43 et al. 2015). The IAM literature typically uses one of two approaches: comprehensive emulators such 44 as MAGICC (Meinshausen et al. 2011) or Hector (Hartin et al. 2015) or minimal complexity 45 representations such as the representation used in DICE (Nordhaus 2018), PAGE (Yumashev et al. Do Not Cite, Quote or Distribute I-29 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2019; Kikstra et al. 2021c) and Fund (Waldhoff et al. 2014). In physical science research, a wider range 2 of different emulators are used (Nicholls et al. 2020b, 2021a). 3 A key application of emulators within IPCC WG III is the classification of emission scenarios with 4 respect to their global mean temperature outcomes (Clarke et al. 2014; Rogelj et al. 2018a). WG III 5 relies on emulators to assess the full range of carbon-cycle, and climate response uncertainty of 6 thousands of scenarios, as assessed by AR6 WG I. An exercise of such amplitude is currently infeasible 7 with more computationally demanding state-of-the-art Earth system models. Cross-chapter Box 7.1 of 8 WG I documents how emulators used in AR6 WG3 are consistent with the physical science assessment 9 of WG I (Forster et al. 2021). 10 Previous IPCC Assessment Reports relied either on the climate output from each individual IAM (IPCC 11 2000) or a more streamlined approach, where one consistent emulator setup was used to assess all 12 scenarios. For instance, in AR5 and SR1.5, MAGICC was used for scenario classification (Clarke et al. 13 2014; Rogelj et al. 2018a). In recent years, numerous other emulators have been developed and 14 increased confidence and understanding can thus be gained by combining insights from more than one 15 emulator. For example, SR1.5 used MAGICC for its scenario classification, with additional insights 16 provided by the FaIR model (Smith et al. 2018) The SR1.5 experience highlighted that the veracity of 17 emulators “is a substantial knowledge gap in the overall assessment of pathways and their temperature 18 thresholds” (Rogelj et al. 2018a). Since SR1.5, international research efforts have demonstrated 19 tractable ways to compare emulator performance (Nicholls et al. 2020b) as well as their ability to 20 accurately represent a set of uncertainty ranges in physical parameters (Nicholls et al. 2021b), such as 21 those reported by the AR6 WG I assessment (Forster et al. 2021). 22 Finally, the recently developed OpenSCM-Runner package (Nicholls et al. 2020a) provides users with 23 the ability to run multiple emulators from a single interface. OpenSCM-Runner has been built in 24 collaboration with the WG III research community and forms part of the WG III climate assessment 25 (Annex III.II.2.4.1). 26 27 9. Integrated assessment modelling 28 Process-based Integrated assessment models (IAMs) describe the coupled energy-land-economy- 29 climate system (Weyant 2009, 2017; Krey 2014). They typically capture all greenhouse gas (GHG) 30 emissions induced by human activities and, in many cases, other emissions of climate forcers like 31 sulphate aerosols. Process-based IAMs represent most GHG and climate pollutant emissions by 32 modelling the underlying processes in energy and land use. Those models are able to endogenously 33 describe the change in emissions due to changes in energy and land use activities, particularly in 34 response to climate action. But IAMs differ in the extent to which all emissions and the corresponding 35 sources, processes and activities are represented endogenously and, thus, can be subjected to policy 36 analysis.1 IAMs also differ regarding the scope of representing carbon removal options and their 37 interlinkage with other vital systems such as the energy and the land-use sectors. 38 Typically, IAMs consider multi-level systems of global, regional, national and local constraints and 39 balance equations for different categories such as emissions, material and energy flows, financial flows, 40 land availability that are solved simultaneously. Intertemporal IAMs can fully incorporate not only flow 41 constraints that are satisfied in each period, but also stock constraints that are aggregated over time and 42 require to balance activities over time. Changes of activities, e.g. induced by policies to reduce 43 emissions are connected to a variety of balance equations and constraints and therefore such policies FOOTNOTE1 See the common IAM documentation at www.iamcdocumentation.eu. Do Not Cite, Quote or Distribute I-30 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 lead to system wide changes that can be analysed with IAMs. Many IAMs also contain gridded 2 components to capture, e.g., land use and climate change processes where the spatial distribution 3 matters greatly for the dynamics of the system. Processes that operate on smaller spatial and temporal 4 scales than resolved by IAMs, such as temporal variability of renewables, are included by 5 parameterisation and statistical modelling approaches that capture the impact of these subscale 6 processes on the system dynamics at the macro level (Pietzcker et al. 2017). 7 Global IAMs are used to analyse global emissions scenarios extrapolating current trends under a variety 8 of assumptions and climate change action pathways under a variety of global goals. In recent years, a 9 class of national and regional IAMs have emerged that describe the coupled energy-land-economy 10 system in a given geography. They typically have higher sectorial, policy and technology resolution 11 than global models and make assumptions about boundary conditions set by global markets and 12 international policy regimes. These IAMs are used to study trends and transformation pathways for a 13 given region (Shukla and Chaturvedi 2011; Capros et al. 2014; Lucena et al. 2016). 14 15 9.1. Types of Integrated Assessment Models 16 IAMs include a variety of model types that can be distinguished into two broad classes (Weyant 2017). 17 The first class comprises cost-benefit IAMs that fully integrate a stylized socioeconomic model with a 18 reduced form climate model to simultaneously account for the costs of mitigation and the damages of 19 global warming using highly aggregate cost functions derived from more detailed models. In the model 20 context these functions do not explicitly represent the underlying processes, but map mitigation efforts 21 and temperature to costs. This closed-loop approach between climate and socioeconomic systems 22 enables cost-benefit analysis by balancing the cost of mitigation and the benefits of avoided climate 23 damages. This can be done in a globally cooperative setting to derive the globally optimal climate policy 24 where no region can further improve its welfare without reducing the welfare of another region (Pareto 25 optimum). Alternatively, it can be assumed that nations do not engage in emission mitigation at all or 26 mitigate in a non-cooperative way only considering the marginal benefit of their own action (Nash 27 equilibrium). Also, differing degrees of partial cooperation are possible. 28 The second class of IAMs, called process-based IAMs, focuses on the analysis of transformation 29 processes depending on a broad set of activities that induce emissions as side effects. They describe the 30 interlinkages between economic activity, energy use, land use, and emissions with emission reductions 31 and removals as well as broader sustainable development targets. GHGs and other climate pollutants 32 are caused by a broad range of activities that are driven by socioeconomic developments (Riahi et al. 33 2017) and also induce broader environmental consequences such as land-use change (Popp et al. 2017) 34 and air pollution (Rao et al. 2017b). With few exceptions, these models typically do not close the loop 35 with climate change and damages that affect the economy, but focus on emission scenarios and climate 36 change mitigation pathways. Due to the process based representations of emission sources and 37 alternatives it is not only possible to investigate the implications of policies on GHG emissions, but also 38 the trade-offs and synergies with social and environmental sustainability criteria (von Stechow et al. 39 2015) (Annex III.I.9.3). The analysis of different cross-sectorial synergies and trade-offs is frequently 40 termed a nexus analysis, such as the energy-water-land nexus. The analysis can also address 41 socioeconomic sustainability criteria such as energy access and human health. Process-based IAMs are 42 also used to explore the synergies and trade-offs of ‘common, but differentiated responsibilities’ by 43 analysing issues of burden sharing, equity, international cooperation, policy differentiation and transfer 44 measures (Tavoni et al. 2015; Leimbach and Giannousakis 2019; Bauer et al. 2020b; Fujimori et al. 45 2016). Do Not Cite, Quote or Distribute I-31 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 There exists a broad range of detailed process IAMs that differ regarding the economic modelling 2 approaches (Annex III.I.2) as well as the methodology and detail of sector representation (Annex III.I.3- 3 7) and how they are interlinked with each other. 4 This leads to differences in model results regarding global aggregates as well as sectorial and regional 5 outputs. Several approaches have been used to evaluate the performance of IAMs and understand 6 differences in IAM behaviour (Wilson et al. 2017a; Schwanitz 2013), including sensitivity analysis 7 (McJeon et al. 2011; Luderer et al. 2013; Rogelj et al. 2013a; Bosetti et al. 2015; Marangoni et al. 2017; 8 Giannousakis et al. 2021), model comparisons (Kriegler et al. 2014a, 2016; Tavoni et al. 2015; Kriegler 9 et al. 2015a; Riahi et al. 2015; Clarke et al. 2009; Riahi et al. 2017; Luderer et al. 2018; Roelfsema et 10 al. 2020; van Soest et al. 2021; Riahi et al. 2021), model diagnostics (Kriegler et al. 2015a; Wilkerson 11 et al. 2015; Harmsen et al. 2021), and comparison with historical patterns (Wilson et al. 2013; van 12 Sluisveld et al. 2015; Napp et al. 2017). 13 14 9.2. Components of integrated assessment models 15 9.2.1. Energy-economy component 16 Typically, IAMs comprise a model of energy flows, emissions and the associated costs (Krey 2014). 17 The demand for exploring the Paris Agreement climate goals led to model developments to make the 18 challenges and opportunities of the associated transformation pathways more transparent. Since AR5 19 much progress has been achieved to improve the representation of mitigation options in the energy 20 supply sector (e.g. renewable energy integration (Pietzcker et al. 2017), energy trade (Bauer et al. 2017, 21 2016; Jewell et al. 2018; McCollum et al. 2016), capacity inertia, carbon removals, decarbonisation 22 bottlenecks (Luderer et al. 2018) and technological and behavioural change measures in energy demand 23 sectors such as transport (Edelenbosch et al. 2017a; van Sluisveld et al. 2016; McCollum et al. 2017). 24 An energy sector model can be run as a partial equilibrium model using exogenous demand drivers for 25 final energy and energy services. These models derive mitigation policy costs in terms of additional 26 energy sector costs and area under the MAC curve. 27 Energy models can be also embedded into a broader, long-term macroeconomic context in a general 28 equilibrium model (Messner and Schrattenholzer 2000; Bauer et al. 2008). The demands for final energy 29 and energy services are endogenously driven by an economic growth model that also endogenizes the 30 economic allocation problem of macroeconomic resources for the energy sector that crowd out with 31 alternatives. This allows impact analysis of climate policies on economic growth and structural change, 32 investment financing and crowding-out as well as income distribution and tax revenue recycling 33 (Guivarch et al. 2011). Moreover, general equilibrium models also derive mitigation costs in terms of 34 GDP losses and Consumption losses, which comprise the full macroeconomic impacts rather than only 35 the narrow energy related costs (Paltsev and Capros 2013). 36 9.2.2. Land system component 37 In recent years substantial efforts have been devoted to improve and integrate land-use sector models 38 in IAMs (Popp et al. 2014b, 2017). This acknowledges the importance of land-use GHG emissions of 39 the agricultural and forestry sectors as well as the role of bioenergy, afforestation and other land-based 40 mitigation measures. The integration is particularly important in light of the long-term climate goals of 41 the Paris Agreement for four reasons (IPCC 2019b). First, the GHG emissions from the land use sector 42 accounts for LUC emissions account for more than 10% of global GHG emissions (Kuramochi et al. 43 2020) and some sources of CH4 and N2O constitute serious mitigation bottlenecks. Second, bioenergy 44 is identified as crucial primary energy source for low-emission energy supply and carbon removal 45 (Bauer et al. 2020a; Butnar et al. 2020; Calvin et al. 2021). Third, land use-based mitigation measures 46 such as afforestation and reduced deforestation have substantial mitigation potentials. Finally, land- Do Not Cite, Quote or Distribute I-32 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 cover changes alter the earth surface albedo, which has implications for regional and global climate. 2 Pursuing the Paris Agreement climate goals requires the inclusion of a broad set of options regarding 3 GHG emissions and removals, which will intensify the interaction between the energy, the economy 4 and the land use sector. Consequently, intersectoral policy coordination becomes more important and 5 the land-related synergies and trade-offs with sustainable development targets will intensify (Calvin et 6 al. 2014b; Humpenöder et al. 2018; Frank et al. 2017; Kreidenweis et al. 2016; van Vuuren et al. 2017a; 7 Bauer et al. 2020d). IAMs used by the IPCC in the AR6 have continuously improved the integration of 8 land-use models with energy models to explore climate mitigation scenarios under varying policy and 9 technology conditions (Rogelj et al. 2018a; Smith et al. 2019). However, feedbacks from changes in 10 climate variables are not or only to a limited degree included in the land use sector models. 11 .9.2.3. Climate system component 12 Reduced complexity climate models (often called simple climate models or emulators) are used for 13 communicating WG I physical climate science knowledge to the research communities associated with 14 other IPCC working groups (Annex III.I.8). They are used by IAMs to model the climate outcome of 15 the multi-gas emissions trajectories that IAMs produce (van Vuuren et al. 2011a). A main application 16 of such models is related to scenario classifications in WG III of the IPCC (Clarke et al. 2014; Rogelj 17 et al. 2018a). Since WG III assesses a large number of scenarios, it must rely on the use of these simple 18 climate models; more computationally demanding models (as used by WG I) will not be feasible to 19 apply. For consistency across the AR6 reports, it is important that these reduced-complexity models are 20 up to date with the latest assessments from IPCC WG I. This relies on calibrating these models so that 21 they match, as closely as possible, the assessments made by WG I (Annex III.II.2.4). The calibrated 22 models can then be used by WG III in various parts of its assessment. 23 24 9.3. Representation of nexus issues and sustainable development impacts in IAMs 25 An energy-water-land nexus approach integrates the analysis of linked resources and infrastructure 26 systems to provide a consistent platform for multi-sector decision-making (Howells et al. 2013). Many 27 of the IAMs that contributed to the assessment incorporate a nexus approach that considers 28 simultaneous constraints on land, water and energy, as well as important mutual dependencies (Calvin 29 et al. 2019; Fricko et al. 2017; Dietrich et al. 2019; Fujimori et al. 2017; van Vuuren et al. 2019). 30 Recently IAMs have also been integrated with life cycle assessment tools in assessing climate 31 mitigation policies to better understand the relevance of life cycle GHG emissions in cost-optimal 32 mitigation scenarios (Tokimatsu et al. 2020; Portugal-Pereira et al. 2016; Pehl et al. 2017; Arvesen et 33 al. 2018). This holistic perspective ensures mitigation pathways do not exacerbate challenges for other 34 sectors or environmental indicators. At the same time, pathways are leveraging potential synergies 35 along the way towards achieving multiple goals. 36 IAMs rely on biophysical models with a relatively high-degree of spatial and temporal resolution to 37 inform coarser scale economic models of the potentials and costs for land, water and energy systems 38 (Johnson et al. 2019). IAMs leverage population, GDP and urbanization projections to generate 39 consistent water, energy and crop demand projections across multiple sectors (e.g., agriculture, 40 livestock, domestic, manufacturing and electricity generation) (Mouratiadou et al., 2016). The highly- 41 distributed nature of decisions and impacts across sectors, particularly for land and water, has been 42 addressed using multi-scale frameworks that embed regional and sub-regional models within global 43 IAMs (Mosnier et al. 2014; Hejazi et al. 2015; Bijl et al. 2018; Portugal-Pereira et al. 2018). These 44 analyses have demonstrated how local constraints and policies interact with national and international 45 strategies aimed at reducing emissions. 46 Sustainable development impacts extending beyond climate outcomes have been assessed by the IAMs 47 that contributed to the assessment, particularly in the context of the targets and indicators consistent Do Not Cite, Quote or Distribute I-33 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 with the Sustainable Development Goals (SDGs). The representation of individual SDGs is diverse 2 (Figure I.3.), and recent model development has focused mainly on improving capabilities to assess 3 climate change mitigation policy combined with indicators for economic growth, resource access, air 4 pollution and land use (van Soest et al. 2019). Synergies and trade-offs across sustainable development 5 objectives can be quantified by analysing multi-sector impacts across ensembles of IAM scenarios 6 generated from single or multiple models (McCollum et al. 2013; Mouratiadou et al. 2016). Modules 7 have also been developed for IAMs with the specific purpose of incorporating policies that address non- 8 climatic sustainability outcomes (Fujimori et al. 2018; Parkinson et al. 2019; Cameron et al. 2016). 9 Similar features have been utilized to incorporate explicit adaptation measures and targeted policies that 10 balance mitigation goals with other sustainability criteria (Bertram et al. 2018; McCollum et al. 2018). 11 Do Not Cite, Quote or Distribute I-34 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2 Figure I.3. The representation of SDGs by IAMs. a) Individual target coverage from a multi-model 3 survey; and b) SDG interactions and coverage by IAM models according to a combination of expert and 4 model surveys. The strength dimension of SDG interactions is indicated by grey shading: darker shades 5 represent strong interactions while white represents no interactions. Orange cells indicate where there is 6 the highest agreement between the importance of interactions and model representation, while blue 7 coloured cells show the most important interactions without model representation. Source: van Soest et 8 al. (2019). 9 9.4. Policy analysis with IAMs 10 A key purpose of IAMs is to provide orientation knowledge for the deliberation of future climate action 11 strategies by policy makers, civil society and the private sector. This is done by presenting different 12 courses of actions (climate change and climate action pathways) towards a variety of long-term climate 13 outcomes under a broad range of assumptions about future socio-economic, institutional and 14 technological developments. The resulting climate change and climate action pathways can be analysed 15 in terms of their outcomes towards a set of societal goals (such as the SDGs) and the resulting trade- 16 offs between different pathways. Key trade-offs that have been investigated in the IAM literature are 17 between (1) no, moderate, and ambitious mitigation pathways (Riahi et al. 2017), (2) early vs. delayed 18 mitigation action (Riahi et al. 2015; Luderer et al. 2018), (3) global action with a focus on economic 19 efficiency equalizing marginal abatement costs across countries and sectors vs. regionally and 20 sectorially fragmented action (Kriegler et al. 2015b; Bertram et al. 2015; Kriegler et al. 2018b; 21 Roelfsema et al. 2020; Bauer et al. 2020b; Blanford et al. 2014a), (4) pathways with different emphasis 22 on supply side vs. demand side mitigation measures (van Vuuren et al. 2018; Grubler et al. 2018) or Do Not Cite, Quote or Distribute I-35 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 more broadly different sustainable development strategies (Riahi et al. 2012; van Vuuren et al. 2015; 2 Soergel et al. 2021), and (5) pathways with different preferences about technology deployment, in 3 particular with regard to CCS and carbon dioxide removals (Kriegler et al. 2014a; Krey 2014; Riahi et 4 al. 2015; Strefler et al. 2018, 2021b; Rose et al. 2020; Luderer et al. 2021). Key uncertainties that were 5 explored in the IAM literature are between (1) different socio-economic futures as, e.g., represented by 6 the Shared Socioeconomic Pathways (SSPs) (Riahi et al. 2017; Bauer et al. 2017; Popp et al. 2017), (2) 7 different technological developments (Bosetti et al. 2015) and (3) different resource potentials (Kriegler 8 et al. 2016). 9 Policy analysis with IAMs follows the approach that a baseline scenario is augmented by some kind of 10 policy intervention. To address the uncertainties in baseline projections, the scientific community has 11 developed the Shared Socioeconomic Pathways (SSPs) that provide a set of vastly different future 12 developments as reference cases (Annex III.II.1.2.2). Most scenarios used in AR6 are based on the 13 middle-of-the-road reference system (SSP2). Depending on the research interest the baseline can be 14 defined as a no-policy baseline or it can include policies that either address GHG emissions like the 15 NDCs or other pre-existing policies such as energy subsidies and taxes. There is no standard definition 16 for baseline scenarios regarding the inclusion of policies. The baseline scenario is augmented by 17 additional policies like a carbon tax aiming towards a long-term climate goal. Hence, the IAM based 18 policy analysis assumes a reference system like SSP2 within which policy scenarios are compared with 19 a baseline scenario. 20 Most policy analysis with process-based IAMs apply a mix of short-term policy evaluation and long- 21 term policy optimization. Policy evaluation applies an exogenous set of policies such as the stated NDCs 22 and evaluates the emission outcomes. Policy optimization is mostly implemented as a cost-effectiveness 23 analysis: a long-term climate stabilisation target is set to derive the optimal mitigation strategy that 24 equalizes marginal abatement cost across sectors, GHGs and countries. This optimal mitigation strategy 25 can be implemented by a broad set of well-coordinated sector specific policies or by comprehensive 26 carbon pricing policies. 27 Most commonly the baseline scenario is either a no-policy baseline or based on the NDCs applying an 28 extrapolation beyond 2030 (Roelfsema et al. 2020; Grant et al. 2020). The climate policy regimes most 29 commonly applied include a long-term target to be reached. The optimal climate strategy can be phased 30 in gradually or applied immediately after 2020. It can focus on a global carbon price equalizing marginal 31 abatement costs across countries or policy intensities can vary across countries and sectors in the near- 32 to medium-term. The climate policy regime can or cannot include effort sharing mechanisms and 33 transfers between regions. Also, it can be extended to include additional sector policies such as 34 improved forest protection or fossil fuel subsidy removal. If certain technologies or activities are related 35 to spill-overs such as technology learning carbon-pricing might be complemented by technology 36 support (Schultes et al. 2018). If carbon pricing policies are fragmented or delayed additional and early 37 sector policies can help reduce distortions and carbon leakage effects (Bauer et al. 2020b). All these 38 variations to the policy regime can lead to very different transformation pathways and policy costs, 39 which is a core result of the IAM analysis. 40 By applying sensitivity analysis IAMs can be used to assess the importance to strategically develop new 41 technologies and options for mitigation and identify sticking points in climate policy frameworks. The 42 sensitivity analysis evaluates differences in outcomes subject to changes in assumptions. For instance, 43 the assumption about the timing and costs of CCS and CDR availability can be varied (Bauer et al. 44 2020a). The differences in mitigation costs and the transformation pathway support the assessment of 45 policy prioritization by identifying and quantifying crucial levers for achieving long-term climate 46 mitigation targets such as R&D efforts and timing of policies. 47 Do Not Cite, Quote or Distribute I-36 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 9.5. Limitations of IAMs 2 The application of IAMs and its results for providing orientation knowledge on climate change response 3 strategies has been criticised based on four arguments (Keppo et al. 2021; Gambhir et al. 2019). First, 4 there are concerns that IAMs are missing important dynamics, e.g. with regard to climate damages and 5 economic co-benefits of mitigation (Stern 2016), demand side responses (Wilson et al. 2012), 6 bioenergy, land degradation and management (Creutzig et al. 2014; IPCC 2019b), carbon dioxide 7 removal (Smith et al. 2016), rapid technological progress in the renewable energy sector (Creutzig et 8 al. 2017), actor heterogeneity, and distributional impacts of climate change and climate policy. This has 9 given rise to criticism that IAMs lack credibility in set of crucial assumptions, among which stands out 10 the critique on the availability of carbon dioxide removal technologies (Bednar et al. 2019; Anderson 11 and Peters 2016). 12 These concerns spur continuous model development and improvements in scenario design (Keppo et 13 al. 2021), particularly with regard to improved representations of energy demand, renewable energy, 14 carbon dioxide removal technologies, and land management. IAMs are aiming to keep pace with the 15 development of sector-specific models, including latest advances in estimating and modelling climate 16 damages (Piontek et al. 2018). In places, where dynamic modelling approaches are lacking, scenarios 17 are being used to explore relevant futures (Grubler et al. 2018). Moreover, sector-specific model 18 comparison studies have brought together domain experts and modellers to improve model 19 representations in these areas (Pietzcker et al. 2017; Edelenbosch et al. 2017a; Harmsen et al. 2020; 20 Rose et al. 2020; Bauer et al. 2020a). Although most models are still relying on the concept of a single 21 representative household representing entire regions, efforts are under way to better represent agent 22 heterogeneity and distributional impacts of climate change and climate mitigation policies (Rao et al. 23 2017a; Peng et al. 2021). 24 Second, concerns have been raised that IAMs are non-transparent and thus make it difficult to grasp 25 context and meaning of their results (Skea et al. 2021). These concerns have facilitated a substantially 26 increase in model documentation (see the common IAM documentation at www.iamcdocumentation.eu 27 as entry point) and open-source models. Nonetheless, more communication tools and co-production of 28 knowledge formats will be needed to contextualize IAM results for users (Auer et al. 2021). When 29 projecting over a century, uncertainties are large and cannot be ignored. Efforts have been undertaken 30 (Marangoni et al. 2017; Gillingham et al. 2018; Harmsen et al. 2021; Wilson et al. 2021) to diagnose 31 key similarities and differences between models and better gauge robust findings from these models 32 and how much they depend on key assumptions (as for example long term growth of the economy, the 33 monetary implication of climate damages or the diffusion and cost of key mitigation technologies). 34 Third, there are concerns that IAMs are describing transformative change on the level of energy and 35 land use, but are largely silent about the underlying socio-cultural transitions that could imply 36 restructuring of society and institutions. Weyant (2017) notes the inability of IAMs to mimic extreme 37 and discontinuous outcomes related to these underlying drivers as one of their major limitations. This 38 is relevant when modelling extreme climate damages as well as when modelling disruptive changes. 39 Dialogues and collaborative work between IAM researchers and social scientists have explored ways 40 to bridge insights from the various communities to provide a more complete picture of high impact 41 climate change scenarios and, on the other end, deep transformation pathways (Turnheim et al. 2015; 42 Geels et al. 2016; Trutnevyte et al. 2019). The extension of IAM research to sustainable development 43 pathways is giving rise to further inter-disciplinary research on underlying transformations towards the 44 Paris climate goals and other sustainable development goals (Kriegler et al. 2018c; Sachs et al. 2019b). 45 Finally, there are concerns that IAM analysis could focus on only a subset of relevant futures and thus 46 push society in certain direction without sufficient scrutiny (Beck and Mahony 2017). IAMs aim to 47 explore a wide range of socio-economic, technology and policy assumptions (Riahi et al. 2017), but it 48 remains a constant challenge to capture all relevant perspectives (O’Neill et al. 2020). These concerns Do Not Cite, Quote or Distribute I-37 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 can be addressed by adopting an iterative approach between researchers and societal actors in shaping 2 research questions and IAM applications (Edenhofer and Kowarsch 2015). IAM research is constantly 3 taking up concerns about research gaps and fills it with new pathway research, as e.g. occurred for low 4 energy demand and limited bioenergy with CCS scenarios (Grubler et al. 2018; van Vuuren et al. 2018). 5 Do Not Cite, Quote or Distribute I-38 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 10. Key characteristics of models that contributed mitigation scenarios to the assessment2 2 Table I.5: Comparison of modelling characteristics as stated by contributing modelling teams to the AR6 database. Attributes include regional scope, sectoral 3 coverage, type of baseline or benchmark setup as a basis for mitigation policies comparison, technology diffusion, capital vintaging and "sunsetting" of technologies 4 and variety of discount rates approaches. Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) GMM (Global MARKAL Model) WEM (World Energy Model) MESSAGEix-GLOBIOM 1.1 REMIND 2.1 - MAgPIE 4.2 E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS China DREAM JRC-EU-TIMES TIAM-ECN 1.1 TIMES-France McKinsey 1.0 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 BLUES 2.0 TIMES_PT MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM Global Regional scope National Non-global multi-region Full system (covering all GHGs from all sectors) Sectoral coverage Energy Buildings Transport 5 Industry 6 FOOTNOTE2 The tables are limited to the integrated models that have provided the information to a survey circulated in 2021, and therefore do not have a comprehensive coverage of all models that submitted scenarios to the AR6 scenario database. Do Not Cite, Quote or Distribute I-39 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) GMM (Global MARKAL Model) WEM (World Energy Model) MESSAGEix-GLOBIOM 1.1 REMIND 2.1 - MAgPIE 4.2 E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS China DREAM JRC-EU-TIMES TIAM-ECN 1.1 TIMES-France McKinsey 1.0 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 BLUES 2.0 TIMES_PT MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM and "sunsetting" of Technology diffusion baseline/benchma Well-functioning markets in equilibrium Characteristics of Regulatory and/or pricing policies rk setup Socioeconomic costs & benefits of climate change impacts Physical impacts of climate change on key processes Logit substitution Constant elasticity of substitution Lowest marginal cost w/ expansion constraints Technology choice depends on agents' preferences Technologies w/o constraints or marginal cost w/ expansion constraints Single capital stock with fixed lifetime and load Capital vintaging factor, early retirement via reduction in load technologies Capital vintaging with fixed lifetime and load factors, early retirement of vintages or reduction Single capital stock with fixed lifetime and load factor, without early retirement Mix of the above for different technologies As a property of an intertemporal welfare Discount rates function (social discount rate) In an objective function of an intertemporal optimization, to sum values at different times To compute lifecycle costs of investment decisions or return on investments, in functions 1 representing agents investment choices 2 Do Not Cite, Quote or Distribute I-40 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2 Table I.6: Overview of evaluated GHG emissions as stated by contributing modelling teams to the AR6 database: carbon dioxide (CO 2) from energy, industrial 3 processes and land use change, methane (CH4) from fossil fuel combustion, from fugitive and process activities, and agricultural biogenic fluxes, nitrous oxide 4 (N2O), hydrofluorocarbons (HFCs), perfluorocarbons (PFCs), sulphur hexafluoride (SF6), sulphur dioxide (SO2), black and organic carbon, and non-methane 5 volatile organic compounds (NMVOC). Levels of emission factor (EF) evaluation were classified in four categories: linked to explicit technology but for average 6 fuel, linked to the evolution of other emissions, dependent on average technology classes, and based on an average activity sector. Type of GHG emissions evaluation Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) EF linked to explicit technology w/ or a w/o fuel representation GMM (Global MARKAL Model) EF linked to evolution of other WEM (World Energy Model) b MESSAGEix-GLOBIOM 1.1 emissions REMIND 2.1 - MAgPIE 4.2 Average EF for technology class c E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 EF for sector d REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS China DREAM JRC-EU-TIMES TIAM-ECN 1.1 TIMES-France McKinsey 1.0 Not represented e 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 BLUES 2.0 TIMES_PT MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM CO2 energy a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a CO2 industrial processes a d a a a b a e d a a a a a c a a d a b a a a a a a c a e e a a CO2 land-use change a d a a a b a e e c d a e d e a d a e c e e a e d e e e e e e e CH4 fossil (combustion) a a a a a b a e e c a a a a c a e a a a e a a a a e e e e e a e CH4 fossil (fugitive and process) a d a a a b a e e a a a e a c a e c e d e d a a c e e e e e a e CH4 biogenic a e a a a b a e e a d a e d b a e d e c e a a e d e e e e e a e N2 O a d a a a b a e e a d a a d c a e a e c e a a a a e e e e e a e HFCs d e e a a e a e e e d d e c d e e e e d e c e a b e e e e e e e PFCs d e e a a e a e e e d e e c d e e e e d e c e a b e e e e e e e SF6 d e e a a e a e e e d d e c d e e e e d e c e a b e e e e e e e SO2 a a e d a e a e e e d a e a e e a a e a e e a e b e e e e e e a Black carbon a d e d a e a e e e e a e a e e e a e a e e e a e e e e e e e a Organic carbon a d e d a e a e e e e a e a e e e a e a e e e e e e e e e e e a 7 Non-methane volatile organic compounds (NMVOC) a a e d a e a e e e d a e a e e a a e a e e a e e e e e e e e a 8 Do Not Cite, Quote or Distribute I-41 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 11. Comparison of mitigation and removal measures represented by models that contributed mitigation scenarios to the 2 assessment3 3 Table I.7: Overview of demand- and supply-side mitigation and removal measures in the energy, transport, building, industry and AFOLU sectors, as stated by 4 contributing modelling teams to the AR6 database. Levels of inclusion were classified in two dimensions of explicit versus implicit and endogenous or exogenous. An 5 explicit level suggests that the measure is directly represented in the model, while an implicit level refers to measures that are estimated indirectly by a proxy. An 6 endogenous level reflects measures that are included in the dynamics of the model framework, whereas an exogenous level refers to measures that are not part of the 7 model dynamics. 8 Level of inclusion Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) GMM (Global MARKAL Model) WEM (World Energy Model) MESSAGEix-GLOBIOM 1.1 REMIND 2.1 - MAgPIE 4.2 E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS JRC-EU-TIMES China DREAM TIAM-ECN 1.1 TIMES-France McKinsey 1.0 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 TIMES_PT BLUES 2.0 MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM Demand side measures Energy efficiency improvements in energy end uses A B A C A A C B A A C C A C A A A A A C B A A B A A A A A A A A Electrification of transport demand A C A A A A A A A A A A A A A A A A A A B A A B B A A A A A A A Electrification of energy demand for buildings A C A C A C A A A A A A A A A A A A A C B A A B B A A A A A A A Electrification of industrial energy demand A C A C A C C A A A A A A C A A A A A C B A A B B A A A A A A A CCS in industrial process applications A A A A A E A E A A A A A A A A A A A C B A A E B A A A A A A A FOOTNOTE3 The tables are limited to the integrated models that have provided the information to a survey circulated in 2021, and therefore do not have a comprehensive coverage of all models that submitted scenarios to the AR6 scenario database. Do Not Cite, Quote or Distribute I-42 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Level of inclusion Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) GMM (Global MARKAL Model) WEM (World Energy Model) MESSAGEix-GLOBIOM 1.1 REMIND 2.1 - MAgPIE 4.2 E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS JRC-EU-TIMES China DREAM TIAM-ECN 1.1 TIMES-France McKinsey 1.0 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 TIMES_PT BLUES 2.0 MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM Higher share of useful energy in final energy C B A C A D A D A A C C A A C A C C A C B A A B B A A A A A A A Reduced energy and service demand in industry C C A C A C C D C B C C D C C C B C B C B B A A C C A C C C B A Reduced energy and service demand in buildings C C D C A D C D C B C C D A A C B C B C B B A B D C A C C C B B Reduced energy and service demand in transport C C A C A A A A D B D C D A C C B C B D B B A B D C A C C C B A Reduced energy and service demand in international C E A C C C C D D B D C D A C C B C B D B E A B E C E C C C E B transport Reduced material demand C B B C C D E E E A E E E E E E B E B E D B B B D E E C C B B B Urban form E E B E C D E D E E E E E E E E E E C E D B B B E E A E E E E D Switch from traditional biomass and modern fuels B A A B A E A C E B A D A A A A A B A D B E A B B A A A A E A E Dietary changes (e.g., reducing meat consumption) B E B A B B A E E A E A E E E E E B E E E E B E E E E E E E E E Food processing A E A C B B E E E E A E E E E E E E E E D E A E E E E E E E E E Reduction of food waste B E E E B E C E E B E B E E E E E B E E E E D E E E E E E E E E Substitution of livestock-based products with plant-based A E B A B D E E E B E E E E E E E B E E E E B E E E E E E E E E products Supply side measures Decarbonisation of electricity: Solar PV A A A A A A A A A A A A A A A A A A A A B A A A A A A A A A A A Solar CSP E E A E A A A A A A A A A A A A A A A A B E A A E A A A A A A E Hydropower A A A A A A B A A A A A A A A A A A A D B A A A A A A A A A A A Nuclear energy A A A A A A A A A A A A A A A A A A A A B A A A A A A A A A A A Advanced, small modular nuclear reactor designs (SMR) E E E C C E E E A E A E E E C A D E C E B E E E E A A E E E E E Do Not Cite, Quote or Distribute I-43 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Level of inclusion Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) GMM (Global MARKAL Model) WEM (World Energy Model) MESSAGEix-GLOBIOM 1.1 REMIND 2.1 - MAgPIE 4.2 E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS JRC-EU-TIMES China DREAM TIAM-ECN 1.1 TIMES-France McKinsey 1.0 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 TIMES_PT BLUES 2.0 MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM Fuel cells (hydrogen) E A A A A E A A A A A A A A A A B A A A B A A B E A A A A A A A CCS at coal and gas-fired power plants A A A A A A A A A A A A A A A A A A A A B A A E B A A A A A A A Ocean energy (incl. tidal and current energy) E E E E C E E A A D A E E A E A A E A E B A E E E A A A A A A E High-temperature geothermal heat A E A E C E A A A D A A A A C A A A A E B A E E B A A A A A A E Wind (on-shore and off-shore lumped together) A A E A A A E A E E A A E E A A A A A A B E E A A A A E E E E E Wind (on-shore and off-shore represented individually) E E A E A A A A A A A A A A A A A E A A B A A E A A E A A A A A Bio-electricity, including biomass co-firing, without CCS A A A A A A A A A A A A A A A A A A A A B A A A B A A A A A A A Bio-electricity, including biomass co-firing, with CCS A A A A A A A A A A A A A A A A A A A A B A A E E A A A A A A A Decarbonisation of non-electric fuels: 1st generation biofuels A E A A A A A E A C A B A A A A A B A A B A A B E A A A A A A A 2nd generation biofuels (grassy/woody biomass A E A A A A A A A C A A A A C A A A A A B A A E E A A A A A A A to liquids) without CCS 2nd generation biofuels (grassy/woody biomass A E A A A A A A A C A A A A C A E A A A B A A E E A A A A A A A to liquids) with CCS Solar and geothermal heating A E A E C E E A A C A A E A C A A A A E B A A B B A A A A A A A Nuclear process heat E E E E C E E E A E A E E E C E E E A E B E E E B A A A E E E E Hydrogen from fossil fuels with CCS E E A A A E C A A A A A A A A A A A A A B A A E E A A A A A A A Hydrogen from electrolysis E E A A A E A A A A A A A A A A A A A A B A A B E A A A A A A A Hydrogen from biomass without CCS E E A A A A A E A D A A A A A A A A A E B A A E E A A A E A A A Hydrogen from biomass with CCS E E A A A E A E A D A A A A A A A A A E B A A E E A A A E A A A Algae biofuels without CCS E E E E E E E E E E E E A E E E E E C E B E E E E E E E E E E E Do Not Cite, Quote or Distribute I-44 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Level of inclusion Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) GMM (Global MARKAL Model) WEM (World Energy Model) MESSAGEix-GLOBIOM 1.1 REMIND 2.1 - MAgPIE 4.2 E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS JRC-EU-TIMES China DREAM TIAM-ECN 1.1 TIMES-France McKinsey 1.0 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 TIMES_PT BLUES 2.0 MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM Algae biofuels with CCS E E E E E E E E E E E E E E E E E E C E B E E E E E E E E E E E Power-to-gas, methanisation, synthetic fuels, fed with E E A A C E E A A E A E E A A A A E A E B A A E E A A A E A A A fossil CO2 Power-to-gas, methanisation, syn-fuels, fed with biogenic E E A E C E E A A E A E E A A A A E A E B A A E E A A A E A A A or atmospheric CO2 Fuel switching and replacing fossil fuels by electricity in C A A A A A A A A A A A A A A A A A A A B A A B B A A A A A A A end-use sectors Other processes: Substitution of halocarbons for refrigerants and insulation E E E E A E C E E E E D E C C E E E E C E A E B E E E E E E E E Reduced gas flaring and leakage in extractive industries C E A B C E C E E A E D E A E B E C A C E E A B B E E E E E E E Electrical transmission efficiency improvements, including E E A C C E E D C E D E E E C B A E A C D E A B B B C E A E B E smart grids Grid integration of intermittent renewables C E A C A C A C C E C A A A C A A A A A D A A A E A A C E E A C Electricity storage C D A A A E A A C A C A A A A A A A A A B A A A D A A C A A A A AFOLU Measures: Reduced deforestation, forest protection, avoided forest A D A A A B A E E A E A E C E B D A E C E E A E B E E E E E E E conversion Methane reductions in rice paddies A E A C A C C E E A E A E C E B E C E C E A A B E E E E E E E E Livestock and grazing management A E A C A A C E E A E A E C E B E C E C E A A B D E E E E E E E Increasing agricultural productivity A C A C A A A E E A E A A C E D D C E E E E A E D E E E E E E E Nitrogen pollution reductions A E B C A A A E E A E A E C E D E C E E E A B B B E E E E E E E Changing agricultural practices enhancing soil carbon E E E C A E E E E A E E A C E B E E E E E E A E D E E E E E E E Do Not Cite, Quote or Distribute I-45 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Level of inclusion Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) GMM (Global MARKAL Model) WEM (World Energy Model) MESSAGEix-GLOBIOM 1.1 REMIND 2.1 - MAgPIE 4.2 E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS JRC-EU-TIMES China DREAM TIAM-ECN 1.1 TIMES-France McKinsey 1.0 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 TIMES_PT BLUES 2.0 MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM Agroforestry and silviculture E C A E D E E E E B E E E E E B E E E E E E A E D E E E E E E E Land-use planning E D A E B E E E E E E E E E E E E E E E E E A E B E E E E E E E Urban and peri-urban agriculture and forestry E E E E D E E E E E E E E E E E E E E E E E E E E E E E E E E E Fire management and (ecological) pest control C E E E D E D E E E E E E E E E E E E E E E E E D E E E E E E E Conservation agriculture E E A E D E E E E E E A E E E D E E E E E E A E E E E E E E E E Influence on land albedo of land use change E E E E A E E E E E E E E E E E E D E E E E E E E E E E E E E E Manure management A E E E A C C E E A E A E E E B E C E C E A A B E E E E E E E B Reduce food post-harvest losses B D E E D E D E E E E B E E E E E E E E E E E E E E E E E E E E Recovery of forestry and agricultural residues E E A E A B A E E E E A E C E E D E E E E E A E B E E E E E E E Forest Management – increasing forest productivity C E E C C B D E E E E A E C E E D E E C E E E E D E E E E E E A Forest Management – increasing timber/biomass C E E E C B D E E E E A E C E E D E E C E E E E D E E E E E E A extraction Forest Management – remediating natural disturbances E E E E B B E E E E E E E E E E E E E C E E E E D E E E E E E E Forest Management – conservation for carbon E D E E B B D E E A E A E E E E D E E C E E E E E E E E E E E E sequestration Carbon dioxide removal: Bioenergy production with carbon capture and A A A A A A A A A A A A A A A A A A A A B A A E E A A A A A A A sequestration (BECCS) Direct air capture and storage (DACS) E E A A A E E A A A A E E A A A A A A A B A A E E E A A A E A E Mineralization of atmospheric CO2 through enhanced E E E E E E E E E C E E E E E E E A E E E E E E E E E E E E E E weathering of rocks Afforestation / Reforestation A A A A A B A E E C E A E C C B C A E A E E A E B E E E E E E E Do Not Cite, Quote or Distribute I-46 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Level of inclusion Global integrated and energy models National integrated models STEM (Swiss TIMES Energy Systems Model) GMM (Global MARKAL Model) WEM (World Energy Model) MESSAGEix-GLOBIOM 1.1 REMIND 2.1 - MAgPIE 4.2 E4SMA-EU-TIMES 1.0 AIM/Enduse-Japan TIMES-Sweden 2.0 REmap GRO2020 IMAGE 3.0 & 3.2 TIMES-China 2.0 CONTO-RUS 1.0 GENeSYS-MOD PROMETHEUS JRC-EU-TIMES China DREAM TIAM-ECN 1.1 TIMES-France McKinsey 1.0 7see6-20_GB MERGE-ETL COFFEE 1.1 C3IAM 2.0 TIMES_PT BLUES 2.0 MUSE 1.0 IMACLIM EPPA 6 WITCH GCAM POLES AIM Restoration of wetlands E E E E C E E E E E E E E E E E E E E E E E E E E E E E E E E E Biochar E E E E D E E E E E A E E E E E E E E E E E E E E E A E E E E E Soil carbon enhancement, enhancing carbon E E A C D D E E E E A A E C E E E C E E E E A E E E E E E E E E sequestration in biota and soils Material substitution of fossil CO2 with bio-CO2 in E E A C A E E E E E A E E E E D E E A E D E A E E A E E E E E E industrial application Ocean iron fertilization E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E Ocean alkalinisation E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E Carbon capture and usage (CCU): Bioplastics, carbon fibre and other construction materials E E A A E E C E A D A E E E A E A E A E E E A E B A E A E E A E 1 2 3 Do Not Cite, Quote or Distribute I-47 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Part II. Scenarios 2 3 1. Overview on climate change scenarios 4 Scenarios are descriptions of alternative future developments. They are used to explore the potential 5 implications of possible future developments and how they might depend on alternative courses of 6 action. They are particularly useful in the context of deep uncertainty. Scenarios are conditional on the 7 realization of external assumptions and can be used to explore possible outcomes under a variety of 8 assumptions. 9 Future climate change is a prime example for the application of scenarios. It is driven by human 10 activities across the world and thus can be altered by human agency. It affects all regions over many 11 centuries to come. Humankind’s response to climate change touches not only on the way we use energy 12 and land, but also on socio-economic and institutional layers of societal development. Climate change 13 scenarios provide a central tool to analyse this wicked problem. 14 1.1. Purposes of climate change scenarios 15 Climate change scenarios are developed for a number of purposes (O’Neill et al. 2020). First, they are 16 constructed to explore possible climate change futures covering the causal chain from (i) socio- 17 economic developments to (ii) energy and land use to (iii) greenhouse gas emissions to (iv) changes in 18 the atmospheric composition of greenhouse gases and short-lived climate forcers and the associated 19 radiative forcing to (v) changes in temperature and precipitation patterns to (vi) bio-physical impacts of 20 climate change and finally to (vii) impacts on socio-economic developments, thus closing the loop. 21 Quantitative scenarios exploring possible climate change futures are often called climate change 22 projections and climate change impact projections 23 Second, climate change scenarios are developed to explore pathways towards long-term climate goals. 24 Goal-oriented scenarios often carry the word pathway in their name, such as climate change mitigation 25 pathway, climate change adaptation pathway, or more generally climate change transition / 26 transformation pathway. They are sometimes called backcasting4 scenarios, or short backcasts, in the 27 literature, particularly when contrasted with forecasts (Robinson 1982). Goal-oriented / backcasting 28 scenarios are inherently normative and intricately linked to human intervention. They can be used to 29 compare and contrast different courses of actions. For example, they are applied in climate change 30 mitigation analysis by comparing reference scenarios without or with only moderate climate policy 31 intervention, sometimes called baseline scenarios, with mitigations pathways that achieve certain 32 climate goals (Grant et al. 2020). Transformation pathways to climate goals are examples of backcasting 33 scenarios. Among other things, they can be used to learn about the multi-dimensional trade-offs between 34 raising or lowering ambition (Clarke et al. 2014; Schleussner et al. 2016). In addition, different 35 transformation pathways to the same goal are often used to analyse trade-offs between different routes 36 towards this goal (Rogelj et al. 2018a). These scenarios need to be looked at as a set to understand 37 attainable outcomes and the trade-offs between them. With scenarios, context matters. 38 Third, climate change scenarios are used to integrate knowledge and analysis between the three different 39 climate change research communities working on the climate system and its response to human 40 interference (linked to WG I of the IPCC), climate change impacts, adaptation and vulnerability (linked FOOTNOTE4 Backcasting is different from Hindcasting. Hindcasting refers to testing the ability of a mathematical model to reproduce past events. In contrast Backcasting begins with a desired future outcome and calculates a pathway from the present to that outcome consistent with constraints. Do Not Cite, Quote or Distribute II-48 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 to WG II) and climate change mitigation (linked to WG III) (IPCC 2000; van Vuuren et al. 2011b; 2 O’Neill et al. 2016) (Annex III.II.1.3). This involves the adoption of common scenario frameworks that 3 allow the consistent use of, e.g., shared emissions scenarios, socio-economic development scenarios 4 and climate change projections (Moss et al. 2010; Kriegler et al. 2012; van Vuuren et al. 2012, 2014; 5 O’Neill et al. 2014). The integrative power of scenarios extends beyond the climate change research 6 community into neighbouring fields such as the social sciences and ecology (Rosa et al. 2020; Pereira 7 et al. 2020). To foster such integration, underlying scenario narratives have proven extremely useful as 8 they allow to develop and link quantitative scenario expressions in very different domains of knowledge 9 (O’Neill et al. 2020). 10 Fourth, climate change scenarios and their assessment aim to inform society (Kowarsch et al. 2017; 11 Weber et al. 2018; Auer et al. 2021). To achieve this, it is important to connect climate change scenarios 12 to broader societal development goals (Riahi et al. 2012; van Vuuren et al. 2015; Kriegler et al. 2018c; 13 Soergel et al. 2021) and relate them to social, sectoral and regional contexts (Absar and Preston 2015; 14 Frame et al. 2018; Kok et al. 2019; Aguiar et al. 2020). To this end, scenarios can be seen as tools for 15 societal discourse and decision making to coordinate perceptions about possible and desirable futures 16 between societal actors (Edenhofer and Kowarsch 2015; Beck and Mahony 2017). 17 18 1.2. Types of climate change mitigation scenarios 19 Different types of climate change scenarios are linked to different purposes and knowledge domains 20 and different models used to construct them (Annex III, Part I). Global reference and mitigation 21 scenarios and their associated emissions projections, which are often called emission scenarios, and 22 national, sector and service transition scenarios are key types of scenarios assessed in the Working 23 Group III report. They are briefly summarized below5. 24 A brief description of the common climate change scenario framework with relevance for all three IPCC 25 Working Groups is provided in Annex III.II.1.3, and a discussion how the WG I and WG II assessments 26 relate to the WG III scenario assessment is given in Annex III.II.2.4. 27 1.2.1. Global mitigation scenarios 28 Global mitigation scenarios are mostly derived from global integrated assessment models (Annex III.9. 29 Integrated assessment modelling) and have been developed in single model studies as well as multi- 30 model comparison studies. The research questions of these studies have evolved together with the 31 climate policy debate and the knowledge about climate change, drivers, and response measures. The 32 assessment of global mitigation pathways in the 5th Assessment Report (AR5) (Clarke et al. 2014) was 33 informed, inter alia, by a number of large-scale multi model studies comparing overshoot and not-to- 34 exceed scenarios for a range of concentration stabilization targets (Energy Modelling Forum (EMF) 35 study 22: EMF22) (Clarke et al. 2009), exploring the economics of different decarbonisation strategies 36 and robust characteristics of the energy transition in global mitigation pathways (EMF27, RECIPE) 37 (Luderer et al. 2012; Krey and Riahi 2013; Kriegler et al. 2014a), and analysing co-benefits and trade- 38 offs of mitigation strategies with energy security, energy access, and air quality objectives (Global 39 Energy Assessment: GEA) (Riahi et al. 2012; McCollum et al. 2011, 2013; Rogelj et al. 2013b; Rao et 40 al. 2013). They also investigated the importance of international cooperation for reaching ambitious 41 climate goals (EMF22, EMF27, AMPERE) (Clarke et al. 2009; Blanford et al. 2014b; Kriegler et al. 42 2015b), the implications of collective action towards the 2°C goal from 2020 onwards vs. delayed 43 mitigation action (AMPERE, LIMITS) (Riahi et al. 2015; Kriegler et al. 2014b), and the distribution of FOOTNOTE5 The terms mitigation / transition / transformation scenarios and mitigation / transition / transformation pathways are used interchangeably, as they refer to goal-oriented scenarios. Do Not Cite, Quote or Distribute II-49 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 mitigation costs and burden sharing schemes in global mitigation pathways (LIMITS) (Tavoni et al. 2 2014, 2015). Scenarios from these and other studies were collected in a scenario database supporting 3 the AR5 assessment (Krey et al. 2014). With a shelf life of 8 to 14 years, they are now outdated and no 4 longer part of this assessment. 5 Since AR5, many new studies published global mitigation pathways and associated emissions 6 projections. After the adoption of the Paris Agreement, several large-scale multi-model studies newly 7 investigated pathway limiting warming to 1.5°C (ADVANCE: Luderer et al. (2018); CD-LINKS: 8 McCollum et al. (2018a); ENGAGE: Riahi et al. (2021); SSPs: Rogelj et al. (2018b)), allowing this 9 report to conduct a robust assessment of 1.5°C pathways. Most scenario studies took the hybrid climate 10 policy architecture of the Paris Agreement with global goals, nationally determined contributions 11 (NDCs) and an increasing number of implemented national climate policies as a starting point, 12 including hybrid studies with participation of global and national modelling teams to inform the global 13 stocktake (ENGAGE: Fujimori et al. (2021); COMMIT: van Soest et al. (2021); CD-LINKS: Schaeffer 14 et al. (2020), Roelfsema et al. (2020)). Multi-model studies covered a range of scenarios from 15 extrapolating current policy trends and the implementation of NDCs, respectively, to limiting warming 16 to 1.5°-2°C with immediate global action and after passing through the NDCs in 2030, respectively. 17 These scenarios are used to investigate, among others, the end of century warming implications of 18 extrapolating current policy trends and NDCs (Perdana et al. 2020); the ability of the NDCs to keep 19 limiting warming to 1.5-2°C in reach (Luderer et al. 2018; Vrontisi et al. 2018; Roelfsema et al. 2020), 20 the scope for global accelerated action to go beyond the NDCs in 2030 (van Soest et al. 2021), and the 21 benefits of early action vs. the risk of overshoot and the use of net negative CO2 emissions in the long- 22 term (Riahi et al. 2021; Bertram et al. 2021; Hasegawa et al. 2021). Other large-scale multi-model 23 studies looked into specific topics: the international economic implications of the NDCs in 2030 24 (EMF36) (Böhringer et al. 2021), the impact of mitigating short-lived climate forcers on warming and 25 health co-benefits in mitigation pathways (EMF30) (Harmsen et al. 2020; Smith et al. 2020b) and the 26 role and implications of large-scale bioenergy deployment in global mitigation pathways (EMF33) 27 (Rose et al. 2020; Bauer et al. 2020a). 28 A large variety of recent modelling studies, mostly based on individual models, deepened research on 29 a diverse set of questions (Annex III.3.2. Global pathways). Selected examples are the impact of peak 30 vs. end of century targets on the timing of action in mitigation pathways (Rogelj et al. 2019a; Strefler 31 et al. 2021a); demand-side driven deep mitigation pathways with sustainable development co-benefits 32 (van Vuuren et al. 2018; Grubler et al. 2018; Bertram et al. 2018); synergies and trade-offs between 33 mitigation and sustainable development goals (Fujimori et al. 2020; Soergel et al. 2021); and the 34 integration of climate impacts into mitigation pathways (Schultes et al. 2021). There have also been a 35 number of recent sectoral studies with global integrated assessment models and other global models 36 across all sectors, e.g. the energy sector (IEA 2021; IRENA 2020; Kober et al. 2020) and transport 37 sector (Rottoli et al. 2021; Fisch-Romito and Guivarch 2019; Edelenbosch et al. 2017a; Zhang et al. 38 2018; Paltsev et al. 2022; Mercure et al. 2018; Lam and Mercure 2021). Very recent work investigated 39 the impact of COVID on mitigation pathways (Kikstra et al. 2021a) and co-designed global scenarios 40 for users in the financial sector (NGFS 2021). In addition to these policy-, technology- and sector- 41 oriented studies, a few diagnostic studies developed mitigation scenarios to diagnose model behaviour 42 (Harmsen et al. 2021) and explore model harmonization (Giarola et al. 2021). 43 The scenarios from most of these and many other studies were collected in the AR6 scenario database 44 (Annex III.II.3.2) and are primarily assessed in Chapter 3 of the report. However sectoral chapters have 45 also used the scenarios, including their climate mitigation categorizations to ensure consistent cross- 46 chapter treatment. Only a small fraction of these scenarios were already available to the assessment of 47 global mitigation pathways in the Special Report on 1.5°C Warming (SR15) (Rogelj et al. 2018a) and 48 were included in the supporting SR15 database (Huppmann et al. 2018). Do Not Cite, Quote or Distribute II-50 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 1.2.2. National transition scenarios 2 A large number of transition scenarios is developed on a national/regional level by national integrated 3 assessment, energy-economy or computable general equilibrium models, among others. These aim to 4 analyse the implications of current climate plans of countries and regions, as well as long-term strategies 5 until 2050 investigating different degrees of low carbon development. National/regional transition 6 scenarios are assessed in Chapter 4 of the Report. 7 Recent research has focused on several different types of national transition scenarios that focus on 8 accelerated climate mitigation pathways in the near-term to 2050. These include scenarios considered 9 by the authors as tied to meeting specific global climate goals6 and scenarios tied to specific policy 10 targets (e.g., carbon neutrality or 80-95% reduction from a certain baseline year). A majority of the 11 accelerated national transition modelling studies up to 2050 evaluate pathways that the authors consider 12 compatible with a 2˚C global warming limit, with fewer scenarios defined as compatible with 1.5˚C 13 global pathways. Regionally, national transition scenarios have centred on countries in Asia 14 (particularly in China, India, Japan), in the European Union, and in North America, with fewer and 15 more narrowly focused scenario studies in Latin America and Africa (Lepault and Lecocq 2021). 16 1.2.3. Sector transition scenarios 17 There are also a range of sector transition scenarios, both on the global and the country level. These 18 include scenarios for the transition of the electricity, buildings, industry, transport and AFOLU sectors 19 until 2050. Due to the accelerated electrification in mitigation pathways, sector coupling plays an 20 increasingly important role to overcome decarbonisation bottlenecks, complicating a separate sector- 21 by-sector scenario assessment. Likewise, the energy-water-land nexus limits the scope a separate 22 assessment of the energy and agricultural sectors. Nevertheless, sector transition scenarios play an 23 important role for this assessment as they can usually offer much more technology, policy and behaviour 24 detail than integrated assessment models. They are primarily assessed in the sector chapters of the 25 report. Their projections of emissions reductions in the sectors in the near- to medium-term is used to 26 check the sector dynamics of global models in Chapter 3 of the Report. 27 Recent transition scenarios considered overarching accelerated climate mitigation strategies across 28 multiple sectors, including demand reduction, energy efficiency improvement, electrification and 29 switching to low carbon fuels. The sectoral strategies considered are often specific to national resource 30 availability, political, economic, climate, and technological conditions. Many sectoral transition 31 strategies have focused on the energy supply sectors, particularly the power sector, and the role for 32 renewable and bio-based fuels in decarbonising energy supply and carbon capture and sequestration 33 (CCS). Some studies present comprehensive scenarios for both supply-side and demand-side sectors, 34 including sector-specific technologies, strategies, and policies. Nearly all demand sector scenarios have 35 emphasized the need for energy efficiency, conservation and reduction through technological changes, 36 with a limited number of models also exploring possible behavioural changes enabled by new 37 technological and societal innovations. 38 1.2.4. Service transition scenarios 39 A central feature of service transition pathways is a focus on the provision of adequate energy services 40 to provide decent standards of living for all as the main scenario objective. Energy services are proxies 41 for well-being, with common examples being provision of shelter (expressed as m 2/capita), mobility 42 (expressed as passenger-kilometres), nutrition (expressed as kCal/capita), and thermal comfort FOOTNOTE6 National emission pathways in the near- or mid-term cannot be linked to long-term mitigation goals without making additional assumptions about emissions by other countries up to the mid-term, and assumptions by all countries up to 2100 (see Chapter 4, Box 4.1). Do Not Cite, Quote or Distribute II-51 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 (expressed as degree-days) (Creutzig et al. 2018). (Creutzig et al. 2018). Service transition pathways 2 seek to meet adequate levels of such energy services with minimal carbon emissions, using 3 combinations of demand- and supply-side options. Ideally this is done by improving the efficiency of 4 service provision systems to minimize overall final energy and resource demand, thereby reducing 5 pressure on supply-side and carbon dioxide removal technologies (Grubler et al. 2018). Specifically, 6 this includes providing convenient access to end-use services (health care, education, communication, 7 etc.), while minimizing both primary and end-use energy required. Service transition pathways provide 8 a compelling scenario narrative focused on wellbeing, resulting in technology and policy pathways that 9 give explicit priority to decent living standards. Furthermore, more efficient service provision often 10 involves combinations of behavioural, infrastructural and technological change, expanding the options 11 available to policymakers for achieving mitigation goals (van Sluisveld et al. 2016, 2018). These 12 dimensions are synergistic, in particular in that behavioural and lifestyle changes often require 13 infrastructures adequately matching lifestyles. Service transition scenarios are primarily assessed in 14 Chapter 5 of the report. 15 16 1.3. Scenario framework for climate change research 17 1.3.1. History of scenario frameworks used by the IPCC 18 For the first three assessment reports the IPCC directly commissioned emission scenarios, with social, 19 economic, energy and partially policy aspects as drivers of projected GHG emissions. The first set of 20 scenarios, the ‘SA90’ of the IPCC First Assessment Report (IPCC 1990), had four distinct scenarios, 21 ‘business-as-usual’ and three policy scenarios of increasing ambition. The set of ‘IS92’ scenarios used 22 in the Second Assessment Report (SAR) investigated variations of business-as-usual scenarios with 23 respect to uncertainties about the key drivers of economic growth, technology and population (Leggett 24 et al. 1992). The SRES scenarios from the IPCC Special Report on Emission Scenarios (SRES) (IPCC 25 2000) were produced by multiple modelling organizations and were used in the Third and Fourth 26 Assessment reports (TAR and AR4). Four distinct scenario families were characterized by narratives 27 and projections of key drivers like population development and economic growth (but no policy 28 measures) to examine their influence on a range of GHG and air pollutant emissions. Until the 4 th 29 Assessment Report, the IPCC organized the scenario development process centrally. Since then, 30 scenarios are developed by the research community and the IPCC limited its role to catalysing and 31 assessing scenarios. To shorten development times, a parallel approach was chosen (Moss et al. 2010) 32 and representative concentration pathways (RCPs) were developed (van Vuuren et al. 2011b) to inform 33 the next generation of climate modelling for the 5th Assessment Report (AR5). RCPs explored four 34 different emissions and atmospheric composition pathways structured to result in different levels of 35 radiative forcing in 2100: 2.6, 4.5, 6.0 and 8.5 W/m2. They were used as an input to the Climate Model 36 Intercomparison Project Phase 5 (CMIP5) (Taylor et al. 2011) and its results were assessed in AR5 37 (Collins et al. 2013). 38 1.3.2. Current scenario framework and SSP-based emission scenarios 39 The current scenario framework for climate change research (van Vuuren et al. 2014; O’Neill et al. 40 2014; Kriegler et al. 2014c) is based on the concept of Shared Socio-Economic Pathways (SSPs) 41 (Kriegler et al. 2012; O’Neill et al. 2014). Unlike their predecessor scenarios from the SRES (IPCC 42 2000), their underlying narratives are motivated by the purpose of using the framework for mitigation 43 and adaptation policy analysis. Hence the narratives are structured to cover the space of socio-economic 44 challenges to both adaptation and mitigation. They tell five stories of sustainability (SSP1), middle of 45 the road development (SSP2), regional rivalry (SSP3), inequality (SSP4) and fossil-fuelled 46 development (SSP5) (O’Neill et al. 2017). SSP1, SSP2, and SSP3 were structured to explore futures 47 with socio-economic challenges to adaptation and mitigation increasing from low to high with Do Not Cite, Quote or Distribute II-52 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 increasing number of SSP. SSP4 was structured to explore a world with high socio-economic challenges 2 to adaptation but low socio-economic challenges to mitigation, while SSP5 explored a world with low 3 challenges to adaptation but high challenges to mitigation. The five narratives have been translated into 4 population and education (Kc and Lutz 2017), economic growth (Dellink et al. 2017; Crespo Cuaresma 5 2017; Leimbach et al. 2017a), and urbanization projections (Jiang and O’Neill 2017) for each of the 6 SSPs. 7 The SSP narratives and associated projections of socio-economic drivers provide the core components 8 for building SSP-based scenario families. These basic SSPs are not scenarios or goal-oriented pathways 9 themselves (despite carrying “pathway” in the name), but building blocks from which to develop full- 10 fledged scenarios. In particular, their basic elements do not make quantitative assumptions about energy 11 and land use, emissions, climate change, climate impacts and climate policy. Even though including 12 these aspects in the scenario building process may alter some of the basic elements, e.g. projections of 13 economic growth, the resulting scenario remains associated with its underlying SSP. To improve the 14 ability of SSPs to capture socio-economic environments, basic SSPs have been extended in various 15 ways, including the addition of quantitative projections on further key socio-economic dimensions like 16 inequality (Rao et al. 2019), governance (Andrijevic et al. 2019), and gender equality (Andrijevic et al. 17 2020a). Extensions also included spatially downscaled projections of, e.g., population developments 18 (Jones and O’Neill 2016). By now, the SSPs have been widely used in climate change research ranging 19 from projections of future climate change to mitigation, impact, adaptation and vulnerability analysis 20 (O’Neill et al. 2020). 21 The integrated assessment modelling community has used the SSPs to provide a set of global integrated 22 energy-land use-emissions scenarios (Riahi et al. 2017; Rogelj et al. 2018b; Bauer et al. 2017; Popp et 23 al. 2017; Rao et al. 2017b; van Vuuren et al. 2017b; Fricko et al. 2017; Fujimori et al. 2017; Calvin et 24 al. 2017; Kriegler et al. 2017) in line with the matrix architecture of the scenario framework (van Vuuren 25 et al. 2014) (Figure II.1.). It is structured along two dimensions: socio-economic assumptions varied 26 along the SSPs, and climate (forcing) outcomes varied along the Representative Concentration 27 Pathways (RCPs) (van Vuuren et al. 2011b). To distinguish resulting emission scenarios from the 28 original four RCPs (RCP2.6, RCP4.5, RCP6.0, and RCP8.5), they are typically named SSPx-y with x 29 = 1,…,5 the SSP label and y = {1.9, 2.6, 3.4, 4.5, 6.0, 7.0, 8.5} W/m2 the nominal forcing level in 2100. 30 The four forcing levels that were already covered by the original RCPs are bolded here. 31 The new SSP-based emissions and concentrations pathways provided the input for CMIP6 (Eyring et 32 al. 2015; O’Neill et al. 2016) and its climate change projections are assessed in AR6 (WG1 AR6 Cross- 33 chapter Box 1.2, WGI AR6 Chapter 4). From the original set of more than 100 SSP-based energy-land 34 use-emissions scenarios produced by six IAMs (Figure II.1.), five Tier-1 scenarios (SSP1-1.9, SSP1- 35 2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5), and four Tier-2 scenarios (SSP4-3.4, SSP4-6.0, variants of SSP7- 36 3.0, SSP5-3.4) were selected7 (O’Neill et al. 2016), further processed and harmonized with historic 37 emissions and land use change estimates (Gidden et al. 2019; Hurtt et al. 2020), and then taken up by 38 CMIP6 models. WGI focuses its assessment of CMIP6 climate change projections on the five Tier-1 39 scenarios (WGI Chapter 4), but also uses the Tier 2 scenarios where they allow assessment of specific 40 aspects like air pollution. All SSP-based IAM scenarios from the original studies are included in the 41 AR6 emissions scenario database and are part of the assessment of global mitigation pathways in 42 Chapter 3. FOOTNOTE7 Each SSPx-y combination was calculated by multiple IAMs. The specific scenarios developed by the marker models for the associated SSPs (SSP1: IMAGE; SSP2: MESSAGE-GLOBIOM; SSP3: AIM; SSP4: GCAM; SSP5: REMIND-MAgPIE) were selected as Tier 1/Tier 2 scenario for use in CMIP6. Tier 2 variants include SSP7-3.0 with low emissions of short lived climate forcers and SSP5-3.4 with high overshoot from following SSP5-8.5 until 2040. Do Not Cite, Quote or Distribute II-53 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 IAMs could not identify SSP-based emissions scenarios for all combinations of SSPs and RCPs (Figure 2 II.1.) (Riahi et al. 2017; Rogelj et al. 2018b). The highest emission scenarios leading to forcing levels 3 similar to RCP8.5 could only be obtained in a baseline without climate policy in SSP5 (SSP5-8.5). 4 Since by now climate policies are implemented in many countries around the world, the likelihood of 5 future emission levels as high as in SSP5-8.5 has become small (Ho et al. 2019). Baselines without 6 climate policies for SSP1 and SSP4 reach up to 6.0-7.0 W/m2, with baselines for SSP2 and SSP3 coming 7 in higher at around 7.0 W/m2. On the lower end, no 1.5°C (RCP1.9) and likely 2°C scenarios (RCP2.6) 8 could be identified for SSP3 due to the lack of cooperative action in this world of regional rivalry. 1.5°C 9 scenarios (RCP1.9) could only be reached by all models under SSP1 assumptions. Models struggled to 10 limit warming to 1.5°C under SSP4 assumptions due to limited ability to sustainably manage land, and 11 under SSP5 assumptions due to its high dependence on ample fossil fuel resources in the baseline 12 (Rogelj et al. 2018b). 13 14 15 Figure II.1. The SSP/RCP matrix showing the SSPs on the horizontal axis and the forcing levels on the 16 vertical axis [Adapted from Rogelj et al. (2018b) Figure 5; A = AIM, G = GCAM; I = IMAGE, M = 17 MESSAGE-GLOBIOM, R = REMIND-MAgPIE, W = WITCH]. Not all SSP/RCP combinations are 18 feasible (red triangles), and not all combinations were tried (grey triangles). Corresponding scenarios 19 were published in Riahi et al. (2017) and Rogelj et al. (2018b) and included the AR6 scenario database. 20 21 1.4. Key design choices and assumptions in mitigation scenarios 22 The development of a scenario involves design choices, in addition to the selection of the model. This 23 section will focus on key choices related to design of the scenario, and the respective socioeconomic, 24 technical, and policy assumptions. Model selection cannot be separated from these choices, but the Do Not Cite, Quote or Distribute II-54 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 various advantages and disadvantages of models are described in Annex III, Part I (Modelling 2 Methods). 3 Target setting: Goal-oriented scenarios in the climate scenario literature initially focussed on 4 concentration stabilisation but have now shifted towards temperature limits and associated carbon 5 budgets. In early model intercomparisons, climate targets were often specified as a CO2 equivalent 6 concentration level that could not be crossed, for example, 450ppm CO2-eq or 550ppm CO2-eq (Clarke 7 et al. 2009). These targets were either applied as not-to-exceed or overshoot targets. In the latter case, 8 concentration levels could be returned to the target level by 2100. Overshoot was particularly allowed 9 for low concentration and temperature targets as many models could not find a solution otherwise 10 (Clarke et al. 2009; Kriegler et al. 2014a; Blanford et al. 2014b; Rogelj et al. 2018b). Bioenergy with 11 Carbon Capture and Storage (BECCS) was an important technology that facilitated aggressive targets 12 to be met in 2100. Due to its ability to remove CO2 from the atmosphere and produce net negative CO2 13 emissions, it enabled overshoot of the target leading to a distinctive peak-and-decline behaviour in 14 concentration, radiative forcing, and temperature (Clarke et al. 2014; Fuss et al. 2014). The mitigation 15 scenarios based on the SSP-RCP framework also applied radiative forcing levels in 2100 (Riahi et al. 16 2017). Temperature targets were often implemented by imposing end-of-century carbon budgets, i.e. 17 cumulative emissions up until 2100. In the case of 2°C pathways, those budgets were usually chosen 18 such that the 2°C limit was not overshoot with some pre-defined probability (Luderer et al. 2018). 19 Arguably, the availability of net negative CO2 emissions has led to high levels of carbon dioxide 20 removal (CDR) in the second half of the century, although CDR deployment is often already substantial 21 to compensate residual emissions (Rogelj et al. 2018a). 22 Recent literature has increasingly focused on alternative approaches such as peak warming or peak CO2 23 budget constraints to implement targets (Rogelj et al. 2019b; Johansson et al. 2020; Riahi et al. 2021). 24 Nevertheless, due to the availability of net negative CO2 emissions and the assumption of standard 25 (exponentially increasing) emissions pricing profiles from economic theory, peak and decline 26 temperature profiles still occurred in a large number of mitigation pathways in the literature even in the 27 presence of peak warming and carbon budget targets (Strefler et al. 2021b). This has led to proposals 28 to combine peak targets with additional assumptions affecting the timing of emissions reductions like a 29 constraint on net negative CO2 emissions (Obersteiner et al. 2018; Rogelj et al. 2019a; Riahi et al. 2021) 30 and different carbon pricing profiles (Strefler et al. 2021b). These proposals are aiming at a stabilization 31 rather than a peak and decline of warming under a given warming limit. However, arguments in support 32 of peak and decline warming profiles also exist: the goal of hedging against positive feedback loops in 33 the Earth system (Lenton et al. 2019) and the aim of increasing the likelihood of staying below a 34 temperature limit towards the end of the century (Schleussner et al. 2016). It is also noteworthy that 35 peak and decline temperature pathways are connected to achieving net zero GHG emissions (with CO2- 36 eq emissions calculated using GWP100) in the second half of the century (Rogelj et al. 2021). 37 Efficiency considerations: Process-based IAMs typically calculate cost-effective mitigation pathways 38 towards a given target as benchmark case (Clarke et al. 2014). In these pathways, global mitigation 39 costs are minimized by exploiting the abatement options with the least marginal costs across all sectors 40 and regions at any time, implicitly assuming a globally integrated and harmonized mitigation regime. 41 This idealized benchmark is typically compared across different climate targets or with reference 42 scenarios extrapolating current emissions trends (UNEP 2019). It naturally evolves over time as the 43 onset of cost-effective action is being set to the immediate future of respective studies. This onset was 44 pushed back from 2010-2015 in studies assessed by AR5 (Clarke et al. 2014) to the first modelling time 45 step after 2020 in studies assessed by AR6. 46 The notion of cost-effectiveness is sensitive to economic assumptions in the underlying models, 47 particularly concerning the assumptions on pre-existing market distortions (Guivarch et al. 2011; Clarke 48 et al. 2014; Krey et al. 2014) and the discount rate on future values. Those assumptions are often not Do Not Cite, Quote or Distribute II-55 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 clearly expressed. Most models have a discount rate of 3-5%, though the range of alternatives is larger. 2 Cost-benefit IAMs have had a tradition of exploring the importance of discount rates, but process-based 3 IAMs have generally not. A lower discount rate brings mitigation forward in time and uses less net 4 negative CO2 emissions in cases where target overshoot is allowed (Realmonte et al. 2019; Emmerling 5 et al. 2019). While most models report discount rates in documentation, there is arguably too little 6 sensitivity analysis of how the discount rate affects modelled outcomes. 7 Cost-effective pathways typically do not account for climate impacts below the temperature limit, 8 although recent updates to climate damage estimates suggest a strengthening of near-term action in 9 cost-effective mitigation pathways (Schultes et al. 2021). Recently, the research community has begun 10 to combine mitigation pathway analysis with ex-post analysis of associated climate impacts and the 11 benefits of mitigation (Drouet et al. 2021). Cost-effective pathways that tap into least cost abatement 12 options globally without considering compensation schemes to equalize the mitigation burden between 13 countries are not compatible with equity considerations. There is a large body of literature exploring 14 international burden sharing regimes to accompany globally cost-effective mitigation pathways (Tavoni 15 et al. 2015; van den Berg et al. 2020; Pan et al. 2017). 16 Policy assumptions: Cost-effective mitigation scenarios assume that climate policies are globally 17 uniform. There is a substantial literature contrasting these benchmark cases with pathways derived 18 under the assumption of regionally fragmented and heterogeneous mitigation policy regimes(Blanford 19 et al. 2014b; Kriegler et al. 2015b, 2018b; Roelfsema et al. 2020; van Soest et al. 2021; Bauer et al. 20 2020b). For example, the Shared Policy Assumptions (Kriegler et al. 2014c) used in the SSP-RCP 21 framework allow for some fragmentation of policy implementation, and many scenarios follow current 22 policies or emission pledges until 2030 before implementing stringent policies (Vrontisi et al. 2018; 23 Roelfsema et al. 2020; Riahi et al. 2015). Other studies assume a gradual strengthening of emissions 24 pledges and regulatory measures converging to a globally harmonized mitigation regime slowly over 25 time (Kriegler et al. 2018b; van Soest et al. 2021). With increasing announcements of mid-century 26 strategies and the rise of net zero CO2 or GHG targets, global mitigation scenario analysis has begun to 27 build in nationally specific policy targets until mid-century (NGFS 2021). 28 Scenarios limiting warming to below 2°C phase in climate policies in all regions and sectors. Almost 29 all converge to a harmonized global mitigation regime before the end of century (with the exception of 30 Bauer et al. (2020b)). In practice, policies are often a mix of regulations, standards, or subsidies. 31 Implementing these real-world policies can give different outcomes to optimal uniform carbon pricing 32 (Mercure et al. 2019). Modelled carbon prices will generally be lower when other policies are 33 implemented (Calvin et al. 2014a; Bertram et al. 2015). As countries implement more and a diverse set 34 of policies, the need to further develop the policy assumptions in models is becoming apparent (O’Neill 35 et al. 2020; Grant et al. 2020; Keppo et al. 2021). 36 Socio-economic drivers: Key socio-economic drivers of emission scenarios are assumptions on 37 population and economic activity. There are other socio-economic assumptions, often included in 38 underlying narratives (O’Neill et al. 2017), that strongly affect energy demand per capita / unit of GDP 39 and dietary choices (Popp et al. 2017; Bauer et al. 2017; Grubler et al. 2018; van Vuuren et al. 2018). 40 The SSPs are often used to help harmonise socio-economic assumptions, and further explore the 41 scenario space. Many studies focus on the middle-of-the-road SSP2 as their default assumption, and 42 many use SSP variations to explore the sensitivity of their results to socio-economic drivers (Riahi et 43 al. 2017; Rogelj et al. 2017; Marangoni et al. 2017). While the SSPs help harmonisation, they are not 44 unique and do not fully explore the scenario space (O’Neill et al. 2020). A wider range of narratives 45 describing alternative worlds is also conceivable. The sustainability world (SSP1), for example, is a 46 world with strong economic growth, but sustainability worlds with low growth or even elements of 47 degrowth in developed countries could also be explored. Thus, standardisation of scenario narratives 48 and drivers has advantages, but can also risk narrowing the scenario space that is explored by the Do Not Cite, Quote or Distribute II-56 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 literature. Consequently, many studies in the literature have adopted other socio-economic assumptions, 2 for example with regard to population and GDP (Kriegler et al. 2016; Gillingham et al. 2018) and 3 sustainable development trends (Soergel et al. 2021). 4 Technology availability and costs: Technology assumptions are a key component of IAMs, with some 5 models representing hundreds or thousands of technologies. Despite the importance of technology costs 6 (Creutzig et al. 2017), there has been limited comparison of technology assumptions across models 7 (Krey et al. 2019; Kriegler et al. 2015b). There is, however, a substantial literature on the sensitivity of 8 mitigation scenarios to technology assumptions, including model comparisons (Kriegler et al. 2014a; 9 Riahi et al. 2015), single model sensitivity studies (McJeon et al. 2011; Krey and Riahi 2013; 10 Giannousakis et al. 2021) and multi-model sensitivity studies (Bosetti et al. 2015). Not only are the 11 initial technology costs important, but also how these costs evolve over time either exogenously or 12 endogenously. Since IAMs have so many interacting technologies, assumptions on one technology can 13 affect the deployment of another. For example, limits on solar energy expansion rates, or integration, 14 may lead to higher levels of deployment for alternative technologies. Because of these interactions, it 15 can be difficult to determine what factors affect deployment across a range of models. 16 Within these key scenario design choices, model choice cannot be ignored. Not all models can 17 implement aspects of a scenario or implement in the same way. Alternative target implementations are 18 difficult for some model frameworks, and implementation issues also arise around technological change 19 and policy implementation. Certain scenario designs may lock out certain modelling frameworks. These 20 issues indicate the need for a diversity of scenario designs (Johansson et al. 2020) to ensure that model 21 diversity can be fully exploited. 22 It is possible for many assumptions to be harmonised, depending on the research question. The SSPs 23 were one project aimed at increasing harmonisation and comparability. It is also possible to harmonise 24 emission data, technology assumptions, and policies (Giarola et al. 2021). While harmonisation 25 facilitates greater comparability between studies, it also limits scenario and model diversity. The 26 advantages and disadvantages of harmonisation need to be discussed for each model study. 27 28 2. Use of scenarios in the assessment 29 2.1. Use of scenario literature and database 30 The WGIII assessment draws on the full literature on mitigation scenarios. To support the assessment, 31 as many as possible mitigation scenarios in the literature were collected in a scenario database with 32 harmonized output reporting (Annex III.II.3). The collection of mitigation pathways in a common 33 database is motivated by a number of reasons: First, to establish comparability of quantitative scenario 34 information in the literature which is often only sporadically available from tables and figures in peer- 35 reviewed publications, reports and electronic supplementary information. Moreover, this information is 36 often reported using different output variables and definitions requiring harmonization. Second, to 37 increase latitude of the assessment by establishing direct access to quantitative information underlying 38 the scenario literature. Third, to improve transparency and reproducibility of the assessment by making 39 the quantitative information underlying the scenario figures and tables shown in the report available to 40 the readers of AR6. The use of such scenario databases in AR5 of WG III (Krey et al. 2014) and SR1.5 41 (Huppmann et al. 2018) proved its value for the assessment as well as for broad use of the scenario 42 information by researchers and stakeholders. This is now being continued for AR6. 43 Do Not Cite, Quote or Distribute II-57 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2.2. Treatment of scenario uncertainty 2 The calls for scenarios issued in preparation of this assessment report allowed to collect a large 3 ensemble of scenarios, coming from many modelling teams using various modelling frameworks in 4 many different studies. Although large ensembles of scenarios were gathered, it should be 5 acknowledged that only a portion of the full uncertainty space is investigated, and that scenarios 6 ensemble distribution of results are an “artefact” of the context of the studies the scenarios were 7 developed in. This introduces “biases” in the ensemble, e.g. (i) the topics of the scenario studies 8 collected in the database determine coverage of the scenario space, with large model-comparison studies 9 putting large weight on selected topics over lesser explored topics explored by individual models, (ii) 10 some models are more represented than others, (iii) only “optimistic” models (i.e. models finding lower 11 mitigation costs) reach the lowest mitigation targets (Tavoni and Tol 2010). Where appropriate, 12 sampling bias was recognized in the assessment, but formal methods to reduce bias were not employed 13 due to conceptual limitations. 14 Furthermore, although it has been attempted to elicit scenario likelihoods from expert knowledge 15 (Christensen et al. 2018), scenarios are difficult to associate with probabilities as they typically describe 16 a situation of deep uncertainty (Grübler and Nakicenovic 2001). This and the non-statistical nature of 17 the scenario ensemble collected in the database does not allow a probabilistic interpretation of the 18 distribution of output variables in the scenario database. Throughout the report, descriptive statistics are 19 used to describe the spread of scenario outcomes across the scenarios ensemble. The ranges of results 20 and the position of scenarios outcomes relative to some thresholds of interest are analysed. In some 21 figures, the median of the distribution of results is plotted together with the interquartile range and 22 possibly other percentiles (5th-10th-90th-95th) to facilitate the assessment of results, but these should not 23 be interpreted in terms of likelihood of outcomes. 24 25 2.3. Feasibility of mitigation scenarios 26 In order to develop feasibility metrics of mitigation scenarios (Chapter 3.8), the assessment relied on 27 the multidimensional feasibility framework developed in Brutschin et al. (2021), considering five 28 feasibility dimensions: (i) geophysical, (ii) technological, (iii) economic, (iv) institutional and (v) socio- 29 cultural. For each dimension, a set of indicators were developed, capturing not only the scale but also 30 the timing and the disruptiveness of transformative change (Kriegler et al. 2018b). All AR6 scenarios 31 (C1-C3 climate categories) were categorized through this framework to quantify feasibility challenges 32 by climate category, time, policy architecture and by feasibility dimension, summarized in Figure 3.43 33 (Chapter 3). 34 Scenarios were categorized into three levels of concerns: (i) low levels of concern where transformation 35 is similar to the past or identified in the literature as feasible/plausible, (ii) medium levels of concern 36 that might be challenging but within reach, given certain enablers, (iii) high levels of concern 37 representing unprecedented levels of transformation attainable only under consistent enabling 38 conditions. Indicators’ thresholds defining these three levels of concern were obtained from the 39 available literature and developed with additional empirical literature. Table II.1 summarizes the main 40 indicators used and the associated thresholds for medium and high levels of concern. Finally, we 41 aggregated feasibility concerns for each dimension and each decade employing the geometric mean, a 42 non-compensatory method which limits the degree of substitutability between indicators, and used for 43 example by the United Nations for the HDI. Alternative aggregation scores such as the counting of 44 scenarios exceeding the thresholds was also implemented. 45 Do Not Cite, Quote or Distribute II-58 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Table II.1. Feasibility dimensions, associated indicators and thresholds for the onset of medium and high 2 concerns about feasibility (Chapter 3.8). Indicators Computation Medium High Source (Frank et al. Total primary energy 2021; Biomass potential generation from biomass in 100 EJ/y 245 EJ/y Creutzig et a given year al. 2014) Geophysical Wind potential Total secondary energy (Deng et al. 830 2000 generation from wind in a 2015; Eurek EJ/year EJ/year given year et al. 2017) (Rogner et al. Total primary energy 1600 50 000 2012; Solar potential generation from solar in a EJ/year EJ/year Moomaw et given year al. 2011) Analogy to Decadal percentage current difference in GDP in COVID-19 GDP loss 5% 10 % mitigation vs baseline spending scenario (Andrijevic et al. 2020b) (Brutschin et Carbon price levels (NPV) 120$ and Carbon price 60$ al. 2021; and decadal increases 5× Economic OECD 2021) Ratio between investments Energy (McCollum in mitigation vs baseline in 1.2 1.5 Investments et al. 2018) a given decade (Brutschin et Share of prematurely al. 2021; Stranded coal retired coal power Global 20 % 50 % assets generation in a given Energy decade Monitor 2021) Decadal percentage point (Brutschin et Wind/Solar scale- increase in the wind/solar al. 2021; 10 pp 20 pp up share in electricity Wilson et al. generation 2020) Established (Brutschin et Decadal percentage point al. 2021; increase in the nuclear Markard et Technological Nuclear scale-up 5 pp 10 pp share in electricity al. 2020; generation Wilson et al. 2020) Amount of CO2 captured 3 (Warszawski New Technologies BECCS scale-up 7 GtCO2/y in a given year GtCO2/y et al. 2021) Fossil CCS scale- Amount of CO2 captured 3.8 8.8 (Budinis et up in a given year GtCO2/y GtCO2/y al. 2018) Biofuels in Decadal percentage point (Nogueira et 5 pp 10 pp transport scale-up increase in the share of al. 2020) Do Not Cite, Quote or Distribute II-59 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII biofuels in the final energy demand of the transport sector Decadal percentage point increase in the share of Electricity in (Muratori et electricity in the final 10 pp 15 pp transport scale-up al. 2021) energy demand of the transport sector Total/transport/ industry/residential Decadal percentage (Grubler et 10 % 20 % energy demand decrease in demand al. 2018) decline Socio-cultural (Grubler et Decline of livestock Decadal percentage al. 2018; share in food decrease in the livestock 0.5 pp 1 pp Bajželj et al. demand share in total food demand 2014) Forest cover Decadal percentage (Brutschin et 2% 5% increase increase in forest cover al. 2021) Pasture cover Decadal percentage (Brutschin et 5% 10 % decrease decrease in pasture cover al. 2021) Institutional Governance level (Brutschin et Governance levels and per and >0.6 and <0.6 and al. 2021; capita CO2 emission decarbonization <20% >20% Andrijevic et reductions over a decade rate al. 2019) 1 2 3 2.4. Illustrative mitigation pathways 4 In the IPCC Special Report on 1.5°C Warming (SR1.5), illustrative pathways (IPs) were used in 5 addition to descriptions of the key characteristics of the full set of scenarios in the database to assess 6 and communicate the results from the scenario literature. While the latter express the spread in scenario 7 outcomes highlighting uncertain vs. robust outcomes, IPs can be used to contrast different stories of 8 mitigating climate change (Rogelj et al. 2018a). 9 Following the example of the SR1.5, IPs have also been selected for the AR6 of WGIII. In contrast to 10 SR1.5, the selection needed to cover a larger range of climate outcomes while keeping the number of 11 IPs limited. The selection focused on a range of critical themes that emerged from the AR6 assessment: 12 1) the level of ambitious of climate policy, 2) the different mitigation strategies, 3) timing of mitigation 13 actions and 4) the combination of climate policy with sustainable development policies. The IPs consist 14 of narratives (Table II.2) as well as possible quantifications. The IPs are illustrative and denote 15 implications of different societal choices for the development of future emissions and associated 16 transformations of main GHG emitting sectors. For Chapter 3, for each of the IPs a quantitative scenario 17 was selected from the AR6 scenario database to have particular characteristics and from diverse 18 modelling frameworks (Table II.3). 19 In total two reference pathways with warming above 2°C and five Illustrative Mitigation Pathways 20 (IMPs) limiting warming in the 1.5-2°C range were selected. The first reference pathway follows 21 current policies as formulated around 2018 (Current Policies, Cur-Pol) through to 2030 and then 22 continues to follow a similar mitigation effort to 2100. The associated quantitative scenario (NGFS Do Not Cite, Quote or Distribute II-60 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2021) selected by Chapter 3 leads to about 3-4 degree C warming at the end of the century. The second 2 reference pathway follows emission pledges to 2030 (NDCs) and then continues with moderate climate 3 action over time (Moderate Action, Mod-Act). 4 The five IMPs are deep mitigation pathways with warming in the 1.5-2°C range. The first IMP pursues 5 gradual strengthening beyond NDC ambition levels until 2030 and then acts to likely limit warming to 6 2°C warming (Climate Category C3) (IMP-GS) (van Soest et al. 2021) (Chapter 3.5.3). Three others 7 follow different mitigation strategies focusing on low energy demand (IMP-LD) (Grubler et al. 2018), 8 renewable electricity (IMP-Ren) (Luderer et al. 2021) and large-scale deployment of carbon dioxide 9 removal measures resulting in net negative CO2 emissions in the second half of the century (IMP-Neg). 10 The fifth IMP explicitly pursues a broad sustainable development agenda and follows SSP1 socio- 11 economic assumptions (IMP-SP) (Soergel et al. 2021). IMP-LD, IMP-Ren and IMP-SP limit warming 12 to 1.5°C with no or low overshoot (C1), while IMP-Neg has a higher overshoot and only returns to 13 nearly 1.5°C (50% chance) by 2100 (close to C2). In addition, two sensitivity cases for IMP-Ren and 14 IMP-Neg are considered that likely limit warming to 2°C (C3) rather than pursuing to limit warming to 15 1.5°C. 16 The IMPs are used in different parts of the report. We just mention some examples here. In Chapter 3, 17 they are used to illustrate key differences between the mitigation strategies, for instance in terms of 18 timing and sectoral action. In Chapter 6, Box 6.9 discusses the consequences for energy systems. 19 Chapter 7 discusses some of the land-use consequences. In Chapter 8, the implications of the IMPs are 20 further explored for urban systems where the elements of energy, innovation, policy, land use and 21 lifestyle interact {8.3, 8.4}. In Chapter 10, the consequences of different mitigation strategies for 22 mobility are highlighted in different figures. The IMPs are discussed further in Chapter 1.3, Chapter 3.2 23 and the respective sector chapters. 24 25 Do Not Cite, Quote or Distribute II-61 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Table II.2. Storylines for the two reference pathways and five Illustrative Mitigation Pathways (IMPs) 2 limiting warming to 1.5-2°C considered in the Report. Land use, food General char. Policy Innovation Energy Lifestyle biodiversity Continuation of current Implementation of Business-as-usual; Fossil fuels remain Further expansion of Demand will continue policies and trends; current climate policies slow progress in low- important; lock-in western diets; further to grow; no significant Cur-Pol and neglect of stated carbon technologies slow expansion of changes in current goals and objectives; agriculture area habits Grey Covid recovery) NDCs in 2030; as Strengthening of policies Modest change Mostly moving away Afforestation/reforest Modest change announced in 2020, to implement NDCs; compared to CurPol from coal; growth of ation policies as in compared to CurPol fragmentated policy some further >2030 renewables; some NDcs Mod-Act landscape; post-2030 strengthening and lock-in in fossil action consistent with mixed Covid recovery investments modest action until 2030 Mitigation in all sectors Successful international Further development CDR, transport Afforestation/reforest Not critical – some also includes a heavy climate policy regime of CDR options; H2/Elec based on ation, BECCS, induced via price Neg reliance on net negative with a focus on a long- negative emissions increased increases emissions (supply-side) term temperature goal competition for land Rapid deployment and Successful international Rapid further Renewable energy, Service provisioning technology climate policy regime; development of electrification; sector and demand changes development of policies and financial innovative electricity coupling; storage or to better adapt to Ren renewables; incentives favouring technologies and power-to-X high RE supply electrification; renewable energy policy regimes technologies; better interconnections Reduced demand leads Social innovation; Demand reduction; Lower food and Service provisioning to early emission efficiency; across all modal shifts in agricultural waste; and demand changes; LD reductions sectors transport; rapid less meat-intensive behavioural changes IMP diffusion of BAT in lifestyles buildings and industry Mitigation action is Until 2030, primarily Similar to Sup, but Similar to Sup, but gradually strengthened current NDCs are with some delay. with some delay. GS until 2030 compared to implemented – but NDCs, move towards strong, universal regime > 2030 Shifting pathways. SDG policies in addition Demand reduction; Lower food and Service provisioning Major transformations to climate policy renewable energy agricultural waste; and demand changes shift development (poverty reduction; less meat-intensive SP towards sustainability environmental lifestyles; and reduced inequality, protection afforestation. including deep GHG emissions reduction 3 4 Table II.3. Quantitative scenario selection by Chapter 3 to represent the two reference pathways and five 5 Illustrative Mitigation Pathways (IMPs) limiting warming to 1.5-2°C for the assessment in Chapter 3. These 6 quantitative representations of the IMPs have also been taken up by a few other chapters where suitable. 7 The warming profile of IMP-Neg peaks around 2060 and declines to below 1.5 (50% likelihood) shortly 8 after 2100. Whilst technically classified as a C3, it exhibits the characteristics of C2 high overshoot 9 pathways. Climate Scenario name in the AR6 scenario Acronym Category Model Reference database (II.3) (II.3.2) Cur-Pol C7 GCAM 5.3 NGFS2_Current Policies (NGFS 2021) Mod-Act C6 IMAGE 3.0 EN_INDCi2030_3000f (Riahi et al. 2021) Illustrative Mitigation Pathways (IMPs) Do Not Cite, Quote or Distribute II-62 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Neg C2* COFFEE 1.1 EN_NPi2020_400f_lowBECCS (Riahi et al. 2021) Ren C1 REMIND-MAgPIE 2.1-4.3 DeepElec_SSP2_ HighRE_Budg900 (Luderer et al. 2021) LD C1 MESSAGEix-GLOBIOM 1.0 LowEnergyDemand_1.3_IPCC (Grubler et al. 2018) GS C3 WITCH 5.0 CO_Bridge (van Soest et al. 2021) SP C1 REMIND-MAgPIE 2.1-4.2 SusDev_SDP-PkBudg1000 (Soergel et al. 2021) Sensitivity cases Neg-2.0 C3 AIM/CGE 2.2 EN_NPi2020_900f (Riahi et al. 2021) Ren-2.0 C3 MESSAGEix-GLOBIOM_GEI 1.0 SSP2_openres_lc_50 (Guo et al. 2021) 1 2 2.5. Scenario approaches to connect WG III with the WG I and WG II assessments 3 2.5.1. Assessment of WG III scenarios building on WG I physical climate knowledge 4 A transparent assessment pipeline has been set up across WG I and WG III to ensure integration of the 5 WG I assessment in the climate assessment of emission scenarios in WG III. This pipeline consists of 6 a step where emissions scenarios are harmonised with historical emissions (harmonisation), a step in 7 which species not reported by an IAM are filled in (infilling), and a step in which the emission 8 evolutions are assessed with three climate model emulators (Annex III.I.8) calibrated to the WG I 9 assessment. These three steps ensure a consistent and comparable assessment of the climate response 10 across emission scenarios from the literature. 11 Harmonisation: IAMs may use different historical datasets, and emission scenarios submitted to the 12 AR6 WG III scenario database (Annex III.II.3) are therefore harmonised against a common source of 13 historical emissions. To be consistent with WG I, we use the same historical emissions that were used 14 for CMIP6 and RCMIP (Gidden et al. 2018; Nicholls et al. 2020b). This dataset comprises many 15 different emission harmonisation sources (Hoesly et al. 2018; van Marle et al. 2017; Velders et al. 2015; 16 Quéré et al. 2016; Gütschow et al. 2016; Meinshausen et al. 2017) including CO2 from agriculture, 17 forestry, and land use change (mainly CEDS, (Hoesly et al. 2018)) that is on the lower end of historical 18 observation uncertainty as assessed in Chapter 2. The harmonisation is performed so that different 19 climate futures resulting from two different scenarios are a result of different future emission evolutions 20 within the scenarios, not due to different historical definitions and starting points. Sectoral CO 2 21 emissions from energy and industrial processes and CO2 from agriculture, forestry, and land use change 22 were harmonised separately. All other emissions species are harmonised based on the total reported 23 emissions per species. For CO2 from energy and industrial processes we use a ratio-based method with 24 convergence in 2080, in line with CMIP6 (Gidden et al. 2018, 2019). For CO2 from agriculture, forestry, 25 and land-use change and other emissions species with high historical interannual variability, we use an 26 offset method with convergence target 2150, to avoid strong harmonization effects resulting from 27 uncertainties in historical observations. For all remaining F-Gases, constant ratio harmonisation is used. 28 For all other emissions species, we use the default settings of Gidden et al. (2018, 2019a). 29 Infilling missing species: Infilling ensures that scenarios include all relevant anthropogenic emissions. 30 This reduces the risk of a biased climate assessment and is important because not all IAMs report all 31 climatically active emission species. Infilling was only performed for scenarios where models provided 32 native reporting of CO2 energy and industrial process, CO2 land use, CH4, and N2O emissions to avoid 33 gases that have large individual radiative forcing contributions and cannot be infilled with high 34 confidence. Models that did not meet this minimum reporting requirement were not included in the 35 climate assessment. Infilling is performed following the methods and guidelines in Lamboll et al. 36 (2020). Missing species are infilled based on the relationship with CO2 from energy and industrial 37 processes as found in the harmonised set of all scenarios reported to the WG III scenario database that 38 pass the vetting requirements. To ensure high stability to small changes, we apply a Quantile Rolling Do Not Cite, Quote or Distribute II-63 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Window method (Lamboll et al. 2020) for aerosol precursor emissions, volatile organic compounds and 2 greenhouse gases other than F-Gases, based on the quantile of the reported CO2 from energy and 3 industrial processes in the database at each time point. F-Gases and other gases with small radiative 4 forcing are infilled based on a pathway with lowest root mean squared difference, allowing for 5 consistency in spite of limited independently modelled pathways in the database. 6 WG I-calibrated emulators: Using expert judgement, emulators that reproduce the best estimates and 7 uncertainties of the majority of WG I assessed metrics are recommended for scenario classification use 8 by WG III (see WG I Cross-Chapter Box 7.1). MAGICC (v7) was used for the main scenario 9 classification, with both FaIR (v1.6.2) and CICERO-SCM (v2019vCH4) being used to provide 10 additional uncertainty ranges on reported statistics to capture climate model uncertainty. The WG I 11 emulators’ probabilistic parameter ensembles are derived such that they match a range of key climate 12 metrics assessed by WG I and the extent to which agreement is achieved is evaluated (WG I Cross- 13 Chapter Box 7.1). Of particular importance to this evaluation is the verification against the WG I 14 temperature assessment of the five scenarios assessed in Chapter 4 of WG I (SSP1-1.9, SSP1-2.6, SSP2- 15 4.5, SSP3-7.0, and SSP5-8.5). The inclusion of the temperature assessment as a benchmark for the 16 emulators provides the strongest verification that WG III’s scenario classification reflects the WG I 17 assessment. The comprehensive nature of the evaluation is a clear improvement on previous reports and 18 ensures that multiple components of the emulators, from their climate response to effective radiative 19 forcing through to their carbon cycles, have been examined before they are deemed fit for use by WG 20 III. 21 Scenario climate assessment: For the WG III scenario climate assessment, emulators are run hundreds 22 to thousands of times per scenario, sampling from an emulator-specific probabilistic parameter set, 23 which incorporates carbon cycle and climate system uncertainty in line with the WG I assessment (WG 24 I Cross-Chapter Box 7.1). Percentiles for different output variables provide information about the 25 spread in individual variables for a given scenario, but the set of variables for a given percentile do not 26 form an internally consistent climate change projection. Instead, joint distributions of these parameter 27 sets are employed by the calibrated emulators. Consistent climate change projections are represented 28 by individual ensemble member runs and the whole ensemble of these individual member runs. To 29 facilitate analysis, multiple percentiles of these large (hundred to thousand member) ensemble 30 distributions of projected climate variables are provided in the AR6 scenario database. The emulators 31 provide an assessment of global surface air temperature (GSAT) response to emission scenarios and its 32 key characteristics like peak warming and year of peak warming, ocean heat uptake, atmospheric CO2, 33 CH4 and N2O concentrations and effective radiative forcing from a range of species including CO 2, 34 CH4, N2O and aerosols for each emissions scenario as well as an estimate of CO2 and non-CO2 35 contributions to the temperature increase. The climate emulator’s GSAT projections are normalized to 36 match the WG I Ch.2 assessed total warming between 1850-1900 and 1995-2014 of 0.85°C. 37 The GSAT projections from the emulator runs are used for classifying those emissions scenarios in the 38 AR6 database that passed the initial vetting and allowed a robust climate assessment. MAGICC (v7) 39 was selected as emulator for the climate classification of scenarios, as it happens to be slightly warmer 40 than the other two considered climate emulators, particularly for the higher and long-term warming 41 scenarios - reflecting long-term warming in line with ESMs (WG1 Cross-Chapter Box 7.1). This means 42 that scenarios identified to stay below a given warming limit with a given probability by MAGICC will 43 in general be identified to have this property by the other two emulators as well. There is the possibility 44 that the other two emulators would classify a scenario in a lower warming class based on their slightly 45 cooler emulation of the temperature response. Unlike during the assessment of the SR1.5 database in 46 the IPCC SR1.5 report, the updated versions of FaIR and MAGICC are however very close, providing 47 robustness to the climate assessment. The other two emulators (FaIR and CICERO-SCM) were still 48 used to assess the overall uncertainty in the warming response for a single scenario or a set of scenarios, Do Not Cite, Quote or Distribute II-64 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 including both parametric and model uncertainty. Specifically, the 5th to 95th percentile range across the 2 three emulators is calculated, characterizing the joint climate uncertainty range of the three models. 3 Carbon budgets in WG1 and WG3: The remaining carbon budget corresponding to a certain level of 4 future warming depends on non-CO2 emissions of modelled pathways. Cross-Working Group Box 1 5 highlighted this key uncertainty in estimating carbon budgets. In this section (Figure II.2.), we put this 6 into the context of the dependence of carbon budgets on two aspects of the non-CO2 warming 7 contribution: (i) assumptions on historical non-CO2 emissions and how they can impact future non-CO2 8 warming estimates relative to a recent reference period (2010-2019) (Panel A) and (ii) the scenario set 9 underlying estimates of non-CO2 warming at the time of reaching net zero CO2 (Panel B). Both aspects 10 affect the estimated remaining carbon budget by changing the non-CO2 warming contribution from the 11 base year to the time of reaching net zero CO2. MAGICC7 is used in WGI in conjunction with different 12 input files for the historical warming. For the reported remaining carbon budget estimates (WG1 CB) 13 WGI is using the non-CO2 warming contributions from MAGICC7 in line with Meinshausen et al. 14 (2020) and in line with the CMIP6 GHG concentration projections, while the WGI emulator setup in 15 line with WG1 Cross Chapter box7.1 was used for the WG3 climate assessment. The WGIII assessment 16 uses thus MAGICC7 in line with Nicholls et al. (2021) in line with the emission harmonisation process 17 employed in WG3 (see above). The difference in historical assumptions changes the estimated non-CO2 18 contribution by up to ~0.05°C for the lower temperature levels, or slightly more than 10% of the 19 warming until 1.5°C relative to 2010-2019. For peak warmings around 2°C relative to pre-industrial 20 levels (~0.97C warming relative to 2010-2019 in below plots), the difference is offset by the difference 21 arising from using either the SR1.5 or AR6 scenario databases (see panel B in below plot). 22 Estimates of the remaining carbon budget that take into account non-CO2 uncertainty are not only 23 dependent on historical assumptions, but also on future non-CO2 scenario characteristics, which are 24 different across the various scenarios in the AR6 database. In panel B of Figure II.2., we show how the 25 SR15 database of scenarios, which was used to inform the WG1 remaining carbon budget, differs from 26 the larger set considered in the WG3 report (both using MAGICC7 using input files in line with Nicholls 27 et al. (2021). Overall, there is limited difference in the covered range of non-CO2 warming at different 28 peak surface temperature levels, leading to no clear change in estimated carbon budgets compared to 29 SR1.5 based on the full scenario database. However, as discussed in Cross-Working Group Box 1 and 30 shown in panel C of Figure II.2., mitigation strategies expressed by both the IAM footprint and scenario 31 design (e.g. dietary change scenarios) can have strong effects on estimated carbon budgets for staying 32 below 1.5°C. 33 Do Not Cite, Quote or Distribute II-65 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2 Figure II.2. Comparison of non-CO2 warming relevant for the derivation of cumulative carbon budgets - 3 and its sensitivity to A) assumptions on historical emissions and B) the set of investigated scenarios 4 (right). Panel C) shows how the relationship across scenarios between peak surface temperature and non- 5 CO2 warming and peak cumulative CO2 is different for modelling frameworks. All dashed regression 6 lines are at the 5th and 95th percentile, solid lines are a regression at the median. 7 All panels depict non-CO2 warming in relation to 2010-2019 at the time of peak cumulative CO2, using 8 MAGICC7. Coloured are those scenarios that reach net-zero CO2 this century, with dots in grey indicating 9 scenarios that do not reach net-zero CO2 but still remain below 2°C median peak warming relative to 2010-2019 10 levels in this century. The scenario set “AR6 database” in B) includes only scenarios of those model frameworks 11 that are shown in panel C) which have a detailed land-use model and enough scenarios to imply a relationship. 12 Panel A) The WG1 remaining carbon budget takes into account the non-CO2 warming in dependence of peak 13 surface temperatures via a regression line approach (lighter blue-coloured solid line). For the same scenario set, 14 with historical emissions assumptions as used in CCB7.1 (darker blue-coloured solid line) a relationship is found 15 with a difference of approximately 0.05°C. 16 Panel B) The WG3 database of scenarios tends to imply very similar non-CO2 warming at peak cumulative CO2 17 as the SR15 scenario database, especially around 1.5°C above pre-industrial (0.43°C above 2010-2019 levels), 18 though with slightly lower non-CO2 warming for higher peak temperatures. 19 Panel C) Regressions at the 5th, 50th, and 95th percentile indicate a model framework footprint affecting the 20 relationship between peak warming and non-CO2 warming at peak cumulative CO2. 21 22 2.5.2. Relating the WG II and WG III assessments by use of warming levels 23 WG II sets out common climate dimensions to help contextualize and facilitate consistent 24 communication of impacts and synthesis across WGII, as well as to facilitate WG I and WG II 25 integration, with the dimensions adopted when helpful and possible across WGII (AR6 WGII Cross- 26 Chapter CLIMATE Box 1.1). “Common climate dimensions” are defined as common Global Warming 27 Levels (GWLs), time periods, and levels of other variables as needed by WGII authors (see below for 28 a list of variables associated with these dimensions). Projected ranges for associated climate variables 29 were derived from the AR6 WGI report and supporting resources and help contextualize and inform the 30 projection of potential future climate impacts and key risks. The information enables the mapping of Do Not Cite, Quote or Distribute II-66 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 climate variable levels to climate projections by WGI (WGI SPM Table SPM.1) and vice versa, with 2 ranges of results provided to characterize the physical uncertainties relevant to assessing climate 3 impacts risk. Common socioeconomic dimensions are not adopted in WG II due to a desire to draw on 4 the full literature, inform the broad ranges of relevant possibilities (climate, development, adaptation, 5 mitigation), and be flexible. The impacts literature is wide-ranging and diverse, with a fraction based 6 on global socioeconomic scenarios. WGII’s approach allows chapters and cross-chapter boxes to assess 7 how impacts and ranges depend on socioeconomic factors affecting exposure, vulnerability, and 8 adaptation independently as appropriate for their literature. For example, WG II Chapter 16 assesses 9 how Representative Key Risks vary under low vs. high exposure/vulnerability conditions by drawing 10 on impact literature based on Shared Socio-economic Pathways (SSPs). In general, WGII chapters, 11 when possible and conducive with their literature, used GWLs or climate projections based on 12 Representative Concentration Pathways (RCPs) or SSPs to communicate information and facilitate 13 integration and synthesis, with impacts results characterized according to other drivers when possible 14 and relevant, such as socioeconomic condition. 15 In the context of common climate dimensions, WGII considers common projected GWL ranges by time 16 period, the timing for when GWLs might be reached, and projected continental level result ranges for 17 select temperature and precipitation variables by GWL (average and extremes), as well sea surface 18 temperature changes by GWL and ocean biome. Where available, WGII considers the assessed WGI 19 ranges as well as the raw CMIP5 and CMIP6 climate change projections (ranges and individual 20 projections) from Earth system models (Hauser et al. 2019). With WGII’s climate impacts literature 21 based primarily on climate projections available at the time of AR5 (CMIP5) and earlier, or assumed 22 temperature levels, it was important to be able to map climate variable levels to climate projections of 23 different vintages and vice versa. WG II’s common GWLs are based on AR6 WGI’s proposed “Tier 1” 24 dimensions of integration range—1.5, 2.0, 3.0, and 4.0˚C (relative to the 1850 to 1900 period), which 25 are simply proposed common GWLs to facilitate integration across and within WGs (WGI Chapter 1). 26 Within WG II, GWLs facilitate comparison of climate states across climate change projections, 27 assessment of the full impacts literature, and cross-chapter comparison. Across AR6, GWLs facilitate 28 integration across WGs of climate change projections, climate change risks, adaptation opportunities, 29 and mitigation. 30 For facilitating integration with WG III, GWLs need to be related to WG III’s classification of 31 mitigation efforts by temperature outcome. WG III’s Chapter 3 groups full century emissions 32 projections resulting from a large set of assessed mitigation scenarios into temperature classes (Annex 33 III.II.2.4, II.3.2.1, Chapter 3.2, 3.3). Scenarios are classified by median peak global mean temperature 34 increase since 1850-1900 in the bands <2°C, 2-2.5°C, 2.5-3°C, 3-4°C, and >4°C, with the range below 35 2°C broken out in greater detail using estimates of warming levels at peak and in 2100 for which the 36 warming response is projected to be likely higher (33th percentile), as likely as not higher or lower 37 (median), and likely lower (67th percentile) (Chapter 3.2, Annex III.II.3.1). WG II’s common GWLs 38 and WG III’s global warming scenario classes are relatable but differ in several important ways. While 39 GWLs represent temperature change that occurs at some point in time, emissions scenarios in a 40 temperature class result in an evolving warming response over time. The emissions scenario warming 41 also has a likelihood attached to the warming level at any point in time, i.e. actual warming outcomes 42 can be lower or higher than median warming projections within the range of the estimated uncertainty. 43 Thus, multiple WGII results across GWLs will be relevant to any particular WGIII emissions pathway, 44 including at the peak temperature level. 45 However, socioeconomic conditions are an important factor defining both impacts exposure, 46 vulnerability, and adaptation, as well as mitigation opportunity and costs, that needs special 47 considerations. The WG III scenario assessment is using additional classifications relating to, inter alia, 48 near term policy developments, technology availability, energy demand, population and economic Do Not Cite, Quote or Distribute II-67 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 growth (Annex III.II.3.2.2, Chapter 3.3), and a set of illustrative mitigation pathways with varying 2 socio-techno-economic assumptions (Annex III.II.2.4, Chapter 3.2). Synthesizing WG II assessments 3 of climate change impacts and WG III assessments of climate change mitigation efforts for similar 4 GWLs / global warming scenario classes would have to address how socio-techno-economic conditions 5 affect impacts, adaptation, and mitigation outcomes. Furthermore, a synthesis of mitigation costs and 6 mitigation benefits in terms of avoided climate change impacts would require a framework that ensures 7 consistency in socioeconomic development assumptions and emissions and adaptation dynamics and 8 allows for consideration of benefits and costs along the entire pathway (O’Neill et al. 2020) (Cross 9 Working Group Box 1 “Economic benefits from avoided climate impacts along long-term mitigation 10 pathways”). 11 12 3. WG III AR6 scenario database 13 [Note: The scenario numbers documented in this section refer to all scenarios that were submitted and not 14 retracted by the literature acceptance deadline of October 11, 2021, and that fulfilled the requirement of being 15 supported by an eligible literature source. Not all those scenarios were used in the assessment, e.g. some did not 16 pass the vetting process as documented in II.3.1. 17 18 As for previous IPCC reports of Working Group III, including the Special Report on 1.5 degrees (SR15) 19 (Huppmann et al. 2018; Rogelj et al. 2018a) and the Fifth Assessment Report (AR5) (Clarke et al. 2014; 20 Krey et al. 2014), quantitative information on mitigation pathways is collected in a dedicated AR6 21 scenario database8 to underpin the assessment. 22 By the time of the AR6 Literature Acceptance deadline of IPCC WGIII (11th October 2021) the AR6 23 scenario database comprised 191 unique modelling frameworks (including different versions and 24 country setups) from 95+ model families –, of which 98 globally comprehensive, 71 national or multi- 25 regional, and 20 sectoral models – with in total 3,131 scenarios, summarized in Table II.4.-Table II.10. 26 (global mitigation pathways), Table II.11. (national and regional mitigation pathways) and Table II.12. 27 (sector transition pathways) below. 28 3.1. Process of scenario collection and vetting 29 To facilitate the AR6 assessment, modelling teams were invited to submit their available emissions 30 scenarios to a web-based database hosted by the International Institute for Applied Systems Analysis 31 (IIASA)9. The co-chairs of Working Group III as well as a range of scientific institutions, including the 32 Integrated Assessment Modelling Consortium (IAMC), University of Cape Town (UCT) and the Centre 33 International de Recherche sur l’Environnement (CIRED), support the open call for scenarios which is 34 subdivided into four dedicated calls, 35 1. a call for global long-term scenarios to underpin the assessment in Chapter 3 as well as 36 facilitating integration with sectoral chapters 6, 7, 8, 9, 10 and 11, 37 2. a call for short- to medium-term scenarios at the national and regional scale underpinning the 38 assessment in Chapter 4, and 39 3. a call for building-focused scenarios to inform the assessment in Chapter 9, and 40 4. a call for transport-focused scenarios to inform the assessment in Chapter 10. FOOTNOTE8 https://data.ene.iiasa.ac.at/ar6-scenario-submission/ FOOTNOTE9 https://data.ene.iiasa.ac.at/ar6-scenario-submission/#/about Do Not Cite, Quote or Distribute II-68 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 A common data reporting template with a defined variable structure was used and all teams were 2 required to register and submit detailed model and scenario metadata. Scenarios were required to come 3 from a formal quantitative model and the scenarios must be published in accordance with IPCC 4 literature requirements. The calls for scenarios were open for a period of 22 months (September 2019- 5 July 2021), with updates possible until October 2021 in line with the literature acceptance deadline. The 6 data submission process included various quality control procedures to increase accuracy and 7 consistency in reporting. Additional categorization and processing of metadata over the full database 8 provided a wide range of indicators and categories that were made centrally available to Lead Authors 9 of the Report to enhance consistency of the assessment, such as: climate, policy and technology 10 categories; characteristics about emissions, energy, socioeconomics and carbon sequestration; metadata 11 such as literature references, model documentation and related projects. 12 For all scenarios reporting global data, a vetting process is undertaken to ensure that key indicators were 13 within reasonable ranges for the baseline period – primarily for indicators relating to emissions and the 14 energy sector (Table II.4). As part of the submission process, model teams were contacted individually 15 with information on the vetting outcome with regard to their submitted scenarios giving them the 16 opportunity to verify the reporting of their data. Checks on technology-specific variables for nuclear, 17 solar & wind and CCS screen not only for accuracy with respect to recent developments, but also 18 indicate reporting errors relating to different Primary Energy accounting methods. Whilst the criteria 19 ranges appear to be large, the focus of these scenarios is the medium-long term and there is also 20 uncertainty in the historical values. For vetting of the Illustrative Mitigation Pathways, the same criteria 21 were used, albeit with narrower ranges (Table II.4). Future values were also assessed and reported to 22 Lead Authors, but not used as exclusion criteria. Where possible the latest values available were used, 23 generally 2019, and if necessary extrapolated to 2020 as most models report only at 5-10 year intervals. 24 2020 as reported in most scenarios collected in the database does not include the impact of the COVID- 25 19 pandemic. 26 Almost three-quarters of submitted global scenarios passed the vetting. The remaining quarter 27 comprised a fraction of scenarios that were rolled over from the SR1.5 database, and were no longer 28 up-to-date with recent developments (excluding the COVID shock). This included scenarios that started 29 stringent mitigation action already in 2015. Other scenarios were expected to deviate from historical 30 trends due to their diagnostic design. All historical criteria for reported variables needed to be met in 31 order to pass the vetting. 32 2266 global scenarios were submitted to the scenario database that fulfilled a minimum requirement of 33 reporting at least one global emission or energy variable covering multiple sectors. 1686 global 34 scenarios passed the vetting criteria described in Table II.4. These scenarios were subsequently flagged 35 of meeting minimum quality standards for use in long term scenarios assessment. Additional criteria 36 for inclusion in the Chapter 3 climate assessment are described in Section 3.2.1. Climate classification 37 of global pathways 38 39 Do Not Cite, Quote or Distribute II-69 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Table II.4. Summary of the vetting criteria and ranges applied to the global scenarios for the climate 2 assessment and preliminary screening for Illustrative Mitigation Pathways. Rows do not sum to the same 3 total of scenarios as not all scenarios reported all variables. EIP stands for energy and industrial process 4 emissions Not Reference value Range (IP range) Pass Fail reported Historical Emissions (sources: EDGAR v6 IPCC and CEDS, 2019 values) CO2 total (EIP + AFOLU) 44,251 MtCO2/yr ±40% (±20%) 1848 23 395 emissions CO2 EIP emissions 37,646 MtCO2/yr ±20% (±10%) 2162 55 49 CH4 emissions 379 MtCH4/yr ±20% (±20%) 1651 139 476 CO2 emissions EIP 2010-2020 % - +0 to +50% 1742 74 450 change 0-250 (100) Mt CCS from Energy 2020 - 1624 77 565 CO2/yr Historical energy production (sources: IEA 2019; IRENA; BP; EMBERS; trends extrapolated to 2020) Primary Energy (2020, IEA) 578 EJ ±20% (±10%) 1813 73 380 Electricity Nuclear (2020, IEA) 9.77 EJ ±30% (±20%) 1603 266 397 Electricity Solar & Wind (2020. 8.51 EJ ±50% (±25%) 1459 377 430 IEA, IRENA, BP, EMBERS). Overall 1686 580 - Future criteria (not used for exclusion in climate assessment but flagged to authors as potentially problematic) No net negative CO2 emissions CO2 total in 2030 >0 1867 4 395 before 2030 CCS from Energy in 2030 < 2000 Mt CO2/yr 1518 183 565 < 20 EJ/yr Electricity from Nuclear in 2030 1595 274 397 CH4 emissions in 2040 100-1000 MtCH4/yr 1775 15 476 5 6 3.2. Global pathways 7 Scenarios were submitted by both individual studies and model inter-comparisons (see factsheets in the 8 Supplementary Material to this Annex). The main model inter-comparisons submitting scenarios are 9 shown in Table II.5. Model inter-comparisons have a shared experimental design and assess research 10 questions across different modelling platforms to enable more structured and systematic assessments. 11 The model comparison projects thus help to understand the robustness of the insights. 12 The number of submitted scenarios varies considerably by study, e.g. from 10 to almost 600 scenarios 13 for the model inter-comparison studies (Table II.5). The numbers of scenarios also vary substantially 14 by model (Table II.8.), highlighting the fact that the global scenario set collected in the AR6 scenario 15 database is not a statistical sample (Section II.2.2.). 16 Do Not Cite, Quote or Distribute II-70 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Table II.5. Model inter-comparison studies that submitted global scenarios to the AR6 scenario database 2 and for which at least one scenario passed the vetting. Scenario counts refer to all scenarios submitted by 3 a study (in brackets), those that passed vetting (centre) and those that passed the vetting and received a 4 climate assessment (left). Publication Key Number of Project Description Website year references scenarios The SSPs are part of a new framework that the climate change research community has adopted (Riahi et al. SSP model- to facilitate the https://tntcat.iias 2018 2017; Rogelj et 70 / 77 (126) comparison integrated analysis of a.ac.at/SspDb al. 2018b) future climate impacts, vulnerabilities, adaptation, and mitigation (II.1.3). Developed a new generation of advanced IAMs and applied the (Luderer et al. improved models to http://www.fp7- 2018 2018; Vrontisi 37 / 40 (72) explore different advance.eu/ et al. 2018) climate mitigation policy options in the post-Paris framework. ADVANCE (Edelenbosch http://www.fp7- Industry sector study 2017 0 / 6 (6) et al. 2017b) advance.eu/ Exploring the complex interplay between climate action and development, while (McCollum et simultaneously taking al. 2018; https://www.cd- CD-LINKS both global and national 2018 41 / 52 (77) Roelfsema et links.org/ perspectives and al. 2020) thereby informing the design of complementary climate- development policies. Exploring new climate policy scenarios on the (van Soest et https://themasite COMMIT global level and in 2021 41 / 59 (68) al. 2021) s.pbl.nl/commit/ different parts of the world Exploring new climate policy scenarios on the (Riahi et al. http://www.enga ENGAGE 2021 591 / 591 (603) global level and in 2021) ge-climate.org/ different parts of the Do Not Cite, Quote or Distribute II-71 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII world https://emf.stanf Energy Modelling (Smith et al. ord.edu/projects/ Forum study into the 2020a; emf-30-short- EMF30 2020 61 / 69 (149) role of non-CO2 climate Harmsen et al. lived-climate- forcers 2020) forcers-air- quality https://emf.stanf Energy Modelling (Rose et al. ord.edu/projects/ EMF33 Forum study into the 2020 2020; Bauer et emf-33-bio- 67 / 68 (173) role of bioenergy al. 2020a) energy-and-land- use Energy Modelling https://emf.stanf Forum study into the ord.edu/projects/ (Böhringer et EMF36 role of carbon pricing 2021 emf-36-carbon- 0 / 305 (320) al. 2021) and economic pricing-after- implications of NDCs paris-carpri Study for scenario- based financial risk https://www.ngfs (NGFS 2021) 24 / 24 (24) NGFS assessment with details 2021 .net/ngfs- (NGFS 2020) 2 / 2 (2)10 on impacts, and sectoral scenarios-portal and regional granularity Study on the long-term PARIS (Perdana et al. https://paris- implications of current 2020 3 / 25 (39) REINFORCE 2020) reinforce.eu policies and NDCs Study with a focus on PARIS harmonizing socio- (Giarola et al. https://paris- REINFORCE 2021 0 / 8 (16) economics and techno- 2021) reinforce.eu economics in baselines Study on the role of CLIMACAP- climate change (van der Zwaan 2016 n.a. 0 / 10 (22) LAMP mitigation in Latin et al. 2016) America 937 / 1336 Total (1697) 1 2 FOOTNOTE10 The first NGFS scenario publication in 2020 comprised 15 scenarios from the literature and 2 newly developed scenarios. The 15 scenarios are also contained in the database under their original study name. Do Not Cite, Quote or Distribute II-72 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 Table II.6. Single model studies that submitted global scenarios to the AR6 scenario database and for 2 which at least one scenario passed the vetting. Scenario counts refer to all scenarios submitted by a study 3 (in brackets), those that passed vetting (center) and those that passed the vetting and received a climate 4 assessment (left). Literature Number of Title of study reference11 scenarios Quantification of an efficiency–sovereignty trade-off in (Bauer et al. 2020b) 4 / 4 (4) climate policy. Transformation and innovation dynamics of the energy- (Baumstark et al. 18 / 18 (18) economic system within climate and sustainability limits. 2021) Tracing international migration in projections of income and (Benveniste et al. 0 / 10 (10) inequality across the Shared Socioeconomic Pathways. 2021) Targeted policies can compensate most of the increased (Bertram et al. 3 / 3 (12) sustainability risks in 1.5 °C mitigation scenarios. 2018) Long term, cross country effects of buildings insulation (Edelenbosch et al. 0 / 8 (8) policies 2021) The role of the discount rate for emission pathways and (Emmerling et al. 4 / 4 (28) negative emissions. 2019) (Reilly et al. 2018; Studies with the EPPA model on the costs of low-carbon Morris et al. 2019, power generation, the cost and deployment of CCS, the 2021; Smith et al. economics of BECCS, the global electrification of light duty 7 / 7 (10) 2021; Fajardy et al. vehicles, the 2018 food, water, energy and climate outlook 2021; Paltsev et al. and the 2021 global change outlook 2021, 2022) Transportation infrastructures in a low carbon world: An (Fisch-Romito and 0 / 24 (32) evaluation of investment needs and their determinants Guivarch 2019) Measuring the sustainable development implications of (Fujimori et al. 5 / 5 (5) climate change mitigation. 2020) How uncertainty in technology costs and carbon dioxide (Giannousakis et al. 9 / 9 (9) removal availability affect climate mitigation pathways. 2021) A low energy demand scenario for meeting the 1.5 °C target (Grubler et al. and sustainable development goals without negative 1 / 1 (1) 2018) emission technologies. Global Energy Interconnection: A scenario analysis based on (Guo et al. 2021) 20 / 20 (20) the MESSAGEix-GLOBIOM Model. (Holden et al. Climate–carbon cycle uncertainties and the Paris Agreement. 0 / 5 (5) 2018) FOOTNOTE11 Publication date of scenarios coincides with year of publication. Do Not Cite, Quote or Distribute II-73 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Ratcheting ambition to limit warming to 1.5 °C–trade-offs (Holz et al. 2018) 6 / 6 (6) between emission reductions and carbon dioxide removal. Peatland protection and restoration are key for climate (Humpenöder et al. 0 / 3 (3) change mitigation 2020) Energy Technology Perspectives 2020. (IEA 2020b) 0 / 1 (1) World Energy Outlook 2020 – Analysis - IEA (IEA 2020a) 0 / 1 (1) Net Zero by 2050 – A Roadmap for the Global Energy (IEA 2021) 0 / 1 (1) Sector. Global Renewables Outlook: Energy transformation 2050. (IRENA 2020) 0 / 2 (2) Climate mitigation scenarios with persistent COVID-19- (Kikstra et al. 19 / 19 (19) related energy demand changes. 2021a) Global anthropogenic emissions of particulate matter (Klimont et al. 0 / 2 (2) including black carbon. 2017) Global energy perspectives to 2060 – WEC’s World Energy (Kober et al. 2020) 0 / 4 (4) Scenarios 2019. Prospects for fuel efficiency, electrification and fleet (Kodjak and 0 / 4 (4) decarbonisation Meszler 2019) Short term policies to keep the door open for Paris climate (Kriegler et al. 18 / 18 (18) goals. 2018b) Deep decarbonisation of buildings energy services through (Levesque et al. 4 / 4 (4) demand and supply transformations in a 1.5°C scenario. 2021) Designing a model for the global energy system-GENeSYS- MOD: An application of the Open-Source Energy Modeling (Löffler et al. 2017) 0 / 1 (1) System (OSeMOSYS) Impact of declining renewable energy costs on electrification (Luderer et al. 8 / 8 (8) in low emission scenarios. 2021) The road to achieving the long-term Paris targets: energy (Marcucci et al. 1 / 1 (3) transition and the role of direct air capture. 2017) The transition in energy demand sectors to limit global (Méjean et al. 0 / 3 (27) warming to 1.5 °C. 2019) Deep mitigation of CO2 and non-CO2 greenhouse gases (Ou et al. 2021) 34 / 35 (36) toward 1.5 °C and 2 °C futures Alternative electrification pathways for light-duty vehicles in (Rottoli et al. 2021) 8 / 8 (8) the European transport sector. Economic damages from on-going climate change imply (Schultes et al. 24 / 24 (24) deeper near-term emission cuts. 2021) Do Not Cite, Quote or Distribute II-74 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII A sustainable development pathway for climate action within (Soergel et al. 8 / 8 (8) the UN 2030 Agenda. 2021) Delayed mitigation narrows the passage between large-scale (Strefler et al. 7 / 7 (7) CDR and high costs 2018) Alternative carbon price trajectories can avoid excessive (Strefler et al. 9 / 9 (9) carbon removal. 2021b) (Strefler et al. Carbon dioxide removal technologies are not born equal. 8 / 8 (8) 2021a) The Impact of U.S. Re-engagement in Climate on the Paris (van de Ven et al. 0 / 10 (10) Targets. 2021) (Müller-Casseres et al. 2021; van The 2021 SSP scenarios of the IMAGE 3.2 model. 40 / 40 (40) Vuuren et al. 2014, 2021) Pathway comparison of limiting global warming to 2°C. (Wei et al. 2021) 0 / 5 (5) 265 / 350 Total (421) 1 2 3.2.1. Climate classification of global pathways 3 The global scenarios underpinning the assessment in Chapter 3 have been classified, to the degree 4 possible, by their warming outcome. The definition of the climate categories and the distribution of 5 scenarios in the database across these categories is shown in Table II.7. (Chapter 3.2). The first four of 6 these categories correspond to the ones used in the IPCC SR1.5 (Rogelj et al. 2018a) while the latter 7 four have been added as part of the AR6 to capture a broader range of warming outcomes. 8 For inclusion in the climate assessment, in addition to passing the vetting (Section II.3.1.), scenarios 9 needed to run until the end of century and report as a minimum CO2 (total and for energy & industrial 10 processes (EIP)), CH4 and N2O emissions to 2100. Where CO2 for AFOLU was not reported, the 11 difference between total and EIP in 2020 must be greater than 500 Mt CO2. Of the total 2425 global 12 scenarios submitted, 1594 could be assessed in terms of their associated climate response, and 1202 of 13 those passed the vetting process. 14 15 Table II.7. Classification of global pathways into warming levels using MAGICC (Chapter 3.2) Description Definition Scenarios Passed All Vetting C1: Below 1.5°C with no or low <1.5°C peak warming with ≥33% chance and < 1.5°C 97 160 overshoot end of century warming with >50% chance C2: Below 1.5°C with high <1.5°C peak warming with <33% chance and < 1.5°C 133 170 overshoot end of century warming with >50% chance C3: Likely below 2°C <2°C peak warming with >67% chance 311 374 C4: Below 2oC <2°C peak warming with >50% chance 159 213 Do Not Cite, Quote or Distribute II-75 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII C5: Below 2.5°C <2.5°C peak warming with >50% chance 212 258 C6: Below 3°C <3°C peak warming with >50% chance 97 129 C7: Below 4°C <4°C peak warming with >50% chance 164 230 C8: Above 4°C >4°C peak warming with ≥50% chance 29 40 Scenario time horizon <2100; insufficient emissions No climate assessment 484 692 species reported Total: 1686 2266 1 2 Table II.8. Global scenarios by modelling framework and climate category. Table includes scenarios numbers 3 that passed all vetting and checks and received categorization (in brackets total number of scenarios categorized 4 but not passing vetting). Unique model versions have been grouped into modelling frameworks for presentation 5 in this table12. For a full list of unique model versions, please see the AR6 Scenario Database. C1: C2: C3: C4: C5: C6: C7: C8: No Grand Total Below Below Likely Below Below Below Below Above climate 1.5°C 1.5°C below 2°C 2.5°C 3.0°C 4.0°C 4.0°C assessm Model group with no with 2°C ent or low high OS OS AIM/ CGE+Hub 4 (18) 3 (7) 17 (37) 8 (23) 13 (23) 4 (7) 6 (32) - (8) 7 (7) 55 (162) C-ROADS 3 (3) 2 (2) 1 (1) 6 (6) COFFEE 1 (1) 4 (7) 14 (16) 15 (22) 21 (24) 9 (11) 1 (3) 65 (84) DNE21+ - (4) - (7) - (10) - (3) - (4) - (8) 9 (10) - (46) EPPA 1 (3) 3 (4) 1 (1) 2 (2) 7 (10) En-ROADS - (2) - (1) - (3) GCAM 6 (10) 6 (9) 13 (17) 9 (16) 6 (13) - (1) 4 (6) 1 (1) 18 (63) 45 (136) GCAM-PR - (1) 1 (3) 2 (3) 13 (14) 3 (21) GEM-E3 2 (2) 10 (10) 12 (12) 6 (6) 5 (5) 3 (3) 3 (3) 4 (11) 41 (52) GRAPE-15 - (1) - (7) - (8) - (2) - (18) IMAGE 7 (16) 9 (9) 34 (34) 18 (18) 22 (22) 16 (16) 34 (34) 2 (2) 2 (2) 142 (153) MERGE-ETL - (1) 1 (1) - (1) 1 (3) MESSAGE - (1) - (4) - (3) - (1) - (1) - (10) MESSAGE- GLOBIOM 20 (20) 43 (48) 59 (61) 39 (40) 57 (59) 20 (22) 28 (33) - (1) 266 (284) POLES 4 (14) 10 (15) 26 (26) 24 (26) 20 (21) 11 (12) 19 (23) 1 (1) 114 (138) REMIND 13 (15) 12 (19) 34 (39) 1 (1) 7 (8) 6 (6) 22 (24) 9 (9) 104 (121) FOOTNOTE12 Scenario numbers by modelling framework combine submissions from different model versions of the same model (indicated by version number or project name in the AR6 scenario database). For the AIM, MESSAGE and REMIND modelling frameworks, the grouping covers the following distinct models (including different versions): AIM/ CGE+Hub: AIM/CGE, AIM/Hub MESSAGE: MESSAGE, MESSAGE-Transport MESSAGE-GLOBIOM: MESSAGE-GLOBIOM, MESSAGEix-GLOBIOM. REMIND: REMIND, REMIND-H13, REMIND-Buildings, REMIND-Transport, REMIND_EU Do Not Cite, Quote or Distribute II-76 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII REMIND-MAgPIE 28 (36) 32 (33) 50 (50) 15 (15) 27 (27) 13 (13) 26 (26) 2 (2) 193 (202) TIAM-ECN 20 (20) 6 (6) 10 (10) 4 (4) 5 (5) - (13) 45 (58) TIAM-UCL - (4) - (1) - (2) - (7) TIAM-WORLD - (3) - (2) - (4) - (2) - (11) WITCH 5 (13) 1 (9) 29 (35) 14 (16) 24 (24) 9 (9) 4 (4) 4 (4) 90 (114) WITCH- GLOBIOM 4 (5) 1 (1) 2 (9) - (4) - (8) - (7) 8 (15) 10 (10) 25 (59) Total 133 311 159 212 97 (160) (170) (374) (213) (258) 97 (129) 164 (230) 29 (40) 54 (124) 1202 (1698) 1 2 Table II.9. Global scenarios by modelling framework that were not included in the climate assessment 3 due to a time horizon shorter than 2100 or a limited reporting of emissions species that did not include 4 CO2 (total emissions or emissions from energy and industry), CH 4 and N2O. Unique model versions have 5 been grouped into modelling frameworks for presentation in this table 13. For a full list of unique model 6 versions, please see the AR6 Scenario Database. Model framework Time horizon Passed vetting Total BET 2100 0 16 C-GEM 2030 32 32 C3IAM 2100 5 14 CGE-MOD 2030 32 32 DART 2030 17 32 E3ME 2050 10 10 EC-MSMR 2030 32 32 EDF-GEPA 2030 32 32 EDGE-Buildings 2100 8 8 ENV-Linkages 2060 7 15 ENVISAGE 2030 32 32 FARM 2100 0 13 GAINS 2050 2 2 GEMINI-E3 2050 6 6 GENeSYS-MOD 2050 1 1 Global TIMES 2050 0 14 GMM 2060 4 4 Global Transportation 2050 4 4 Roadmap ICES 2030/2050 32 43 IEA ETP 2070 1 1 IEA WEM 2050 2 2 IRENA REmap GRO2020 2050 2 2 IMACLIM 2050/2080 30 68 IMACLIM-NLU 2100 1 3 LUT-ESTM 2050 0 1 MAgPIE 2100 3 3 FOOTNOTE13 Scenario numbers by modelling framework combine submissions from different model versions of the same model (indicated by version number or project name in the AR6 scenario database). Do Not Cite, Quote or Distribute II-77 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII MIGRATION 2100 10 10 MUSE 2100 5 11 McKinsey 2050 0 3 PROMETHEUS 2050 7 7 SNOW 2030 32 32 TEA 2030 32 32 TIAM-Grantham 2100 17 19 WEGDYN 2030 32 32 Total 430 568 1 2 Changes in climate classification of scenarios since SR1.5: Since the definition of warming classes 3 was unchanged from SR1.5 for the lower range of scenarios limiting warming to 2°C and below, 4 changes in overall emissions characteristics of scenarios in these classes from SR1.5 to AR6 would 5 need to come from the substantially larger ensemble of deep mitigation scenarios collected in the AR6 6 database compared to the SR1.5 database and from updates in the methodology of the climate 7 assessment. Updates since SR1.5 include the methodology for infilling and harmonization and the use 8 of an updated climate emulator (MAGICC v7) to provide consistency with WGI AR6 assessment 9 (II.2.5.1). Out of the full set of SR1.5, 57% of the 411 scenarios that were represented with global 10 temperature assessments in SR1.5 also have been assessed in AR6. Some SR1.5 scenarios could not be 11 taken on board since they are outdated (too early emissions reductions) and failed the vetting or do not 12 provide sufficient information/data to be included in AR6. 13 Comparison between SR1.5 and AR6 scenarios and associated climate responses are shown in Figure 14 II.3, bottom panel. We show that changes in the climate assessment pipeline are minor compared to 15 climate model uncertainty ranges in WGI (in the order of 0.1°C), but show considerable variation due 16 to different scenario characteristics. The updated harmonization and infilling together have a small 17 cooling effect compared to raw modelled emissions for the subset of 95 scenarios in C1, C2, and C3 18 that also were assessed in SR15 (SR1.5 Chap. 2, Table 2.4). This is due to both applying more advanced 19 harmonization methods consistent with the CMIP6 harmonization used for WGI, and changing the 20 historical harmonization year from 2010 to 2015. Together with the update in the climate emulator, we 21 find that the total AR6 assessment is remarkably consistent with SR1.5, albeit slightly cooler (in the 22 order of 0.05°C at peak temperature, 0.1°C in 2100). 23 The lowest temperature category (C1, limiting warming to 1.5 with no or low overshoot) used for 24 classifying the most ambitious climate mitigation pathways in the literature, indicates that emissions 25 are on average higher in AR6 in the near term (e.g., 2030) and the time of net zero CO2 is later by about 26 5 years compared to SR1.5 (Figure II.3, middle panel). These differences can in part be ascribed to the 27 fact that historical emissions in scenarios, especially among those that passed the vetting, have risen 28 since SR1.5 in line with inventories. This increase has moved the attainable near-term emissions 29 reductions upwards. As a result, the scenarios in the lowest category have also a lower probability to 30 stay below 1.5°C peak warming. Using the WGI emulators, we find that the median probability to stay 31 below 1.5°C in the lowest category (C1) has dropped from about 46% in the SR1.5 scenarios to 38% 32 among the AR6 scenarios. Note that the likelihood of the SR1.5 scenarios limiting warming to 1.5C 33 with limited or no overshoot has changed from 41% in SR1.5 to 46% in AR6 due to the updated climate 34 assessment using the WGI AR6 climate emulator. Within C1, the vast majority of scenarios that are 35 submitted to AR6 but were not assessed in SR1.5 have a median peak temperatures close to 1.6°C. The 36 AR6 scenarios in the lowest category show higher emissions and have a lower chance to keep warming 37 below 1.5°C, as indicated by the panels showing the distribution of peak warming and exceedance 38 probability in AR6 vs SR1.5, with for instance C1 median peak temperature warming going from 39 1.55°C in SR1.5 (1.52°C if reassessed with AR6 assessment pipeline) to 1.58°C in AR6. Do Not Cite, Quote or Distribute II-78 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 2 3 Figure II.3. Comparing multiple characteristics of scenarios underlying SR1.5 Table 2.4 to the AR6 4 assessment. 5 Top row: The probability of exceeding 1.5C for scenarios using the AR6 climate assessment pipeline for 6 C1, C2, and C3. AR6 shows all scenarios in AR6 that pass vetting requirements and get climate classification 7 C1, C2, or C3, (‘AR6 (n=541)’). The scenarios that are both in the AR6 database (passing the vetting) and were 8 used for SR1.5 Table 2.4, and are classified as C1, C2 and C3 using the AR6 assessment, are labelled as ‘AR6 9 and SR1.5 overlap (n=95)’. ‘SR1.5 (n=127)’ shows all SR1.5 scenarios (except 5 that were not resubmitted for 10 the AR6 report), including those that fail AR6 vetting, that are classified C1, C2, C3 with the updated AR6 11 temperature assessment. Dashed lines indicate cut-off temperature exceedance probabilities that align with AR6 12 category definitions. The violin area is proportional to the number of scenarios. Coloured lines indicate the 25th 13 and 75th percentile, while the dashed black line indicates the median. The insets in each figure show how the 14 temperature category classification have changed from SR1.5 to AR6 for those scenarios that are in both 15 databases. 16 Middle row: Characteristics of CO2 emissions pathways and the distribution of median peak temperature 17 assessments for C1 and C3. From left to right: (i) change in CO2 emissions levels and reductions in 2030, 2040 18 and 2050 between the AR6 (n=408), AR6 and SR1.5 overlap (n=60) and SR15 sets (n=91). (ii) The distribution 19 of scenarios with different median peak temperature scenario outcomes for C1 and C3 for AR6 and SR1.5 (both 20 with AR6 temperature assessment as a solid line and with SR1.5 temperature assessment as a dashed line with 21 median in yellow). (iii) Year of net-zero CO2 for C1 and C3 for AR6 and SR1.5. Within C3, 27 AR6 scenarios 22 and 2 SR1.5 scenarios with no net-zero year before 2100 have not been visualised. The violin area is proportional 23 to the number of scenarios. Coloured lines indicate the 25th and 75th percentile, while the dashed black line 24 indicates the median. 25 Bottom-row: Change in median global-mean surface air temperature (GSAT) between the AR6 and SR1.5 26 climate assessments for both 2100 values and peak temperature values during the 21st century. Positive 27 values indicate that the temperature assessment is higher for the same scenario than the SR1.5 climate assessment. Do Not Cite, Quote or Distribute II-79 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII 1 From left to right, the effect of using MAGICCv7 calibrated to the WGI assessment compared with MAGICC6 2 as used in SR1.5. The effect of more advanced emissions harmonization and infilling methods. The total is the 3 sum of the three components. Boxplots show the median and interquartile range, with the whiskers indicating the 4 95% range. 5 6 3.2.2. Policy classification of global scenarios 7 Global scenarios were also classified based on their assumptions regarding climate policy. This 8 information can be deduced from study protocols or the description of scenario designs in the published 9 literature. It has also been elicited as meta-information for scenarios that were submitted to the AR6 10 database. There are multiple purposes for a policy classification, including controlling for the level of 11 near-term action (Chapter 3.5) and estimating costs and other differences between two policy classes 12 (Chapter 3.6). Policy classes can be combined with climate classes, e.g. to identify scenarios that follow 13 the NDC until 2030 and likely limit warming to 2°C. 14 Table II.8 presents the policy classification that was chosen for this assessment and the distribution of 15 scenarios across the policy classes. There is top level distinction between diagnostic scenarios, scenarios 16 from cost-benefit analyses, scenarios without globally coordinated action, scenarios with immediate 17 such action, and hybrid scenarios that move to globally coordinated action after a period of diverse and 18 uncoordinated nation. On the second hierarchy level, scenarios are classified along distinctive features 19 of scenarios in each class. Scenarios without globally coordinated action are often used as reference 20 scenarios and come as baselines without climate policy efforts, as an extrapolation of current policy 21 trends or as implementation and extrapolation of NDCs (Grant et al. 2020). Scenarios that act 22 immediately to limit warming to some level can be distinguished by whether or not they include 23 transfers to reflect equity considerations (Tavoni et al. 2015; van den Berg et al. 2020; Bauer et al. 24 2020c) or by whether or not they assume additional policies augmenting a global carbon price (Soergel 25 et al. 2021). Scenarios that delay globally coordinated action until 2030 can differ in their assumptions 26 about the level of near-term action (van Soest et al. 2021; Roelfsema et al. 2020). 27 To identify the policy classification of each global scenario in the AR6 database, classes are first 28 assigned via text pattern matching on all the metadata collected when submitting the scenarios to the 29 database. The algorithm first looks for keywords and text patterns to establish whether a scenario 30 represents a global, fragmented, diagnostic or CBA policy setup. Then it looks for evidence on the 31 presence of specific regional policies, delayed actions and transfers of permits. Eventually the different 32 pieces of evidence are harmonized into a single policy categorization decision. The process has been 33 calibrated on the best-known scenarios belonging to the larger model intercomparison projects, and 34 fine-tuned on the other scenarios via further validation against the related literature, consistency checks 35 on reported emission and carbon price trajectories, exchanges with modellers and supervision by the 36 involved IPCC authors. If the information available is enough to qualify a policy category number but 37 not sufficient for a subcategory, then only the number is retained (e.g., P2 instead of P2a/b/c). A suffix 38 added after P0 further qualifies a diagnostic scenario as one of the other policy categories. 39 40 Table II.10. Policy classification of global scenarios. If the total for a class exceeds the sum of the 41 subclasses, there are scenarios in the class that could not be assigned to a subclass. Class Definition Number of scenarios Passed All vetting P0 Diagnostic scenario 99 138 P1 No globally coordinated climate policy and either 500 632 Do Not Cite, Quote or Distribute II-80 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII P1a • no climate mitigation efforts 124 179 P1b • current national mitigation efforts 59 72 P1c • NDCs 160 184 P1d • other policy assumptions 153 189 P2 Globally coordinated climate policies with immediate (i.e. 634 992 before 2030) action and P2a • without any transfer of emission permits 435 610 P2b • with transfers 70 143 P2c • with additional policy assumptions 55 83 P3 Globally coordinated climate policies with delayed (i.e. from 451 502 2030 onwards or after 2030) action, preceded by P3a • no mitigation commitment or current national policies 7 9 P3b • NDCs 426 464 P3c • NDCs and additional policies 18 29 P4 Cost-benefits analysis 2 2 Total 1686 2266 1 2 3.3. National and regional pathways 3 National and regional pathways have been collected in the AR6 scenario database to support the Chapter 4 4 assessment. In total more than 500 pathways for 24 countries/regions have been submitted to the AR6 5 scenario database by integrated assessment, energy-economic and computable general equilibrium 6 modelling research teams. This represents a limited sample of the overall literature on mitigation 7 pathways at the national level. The majority of these pathways originate from a set of larger model 8 intercomparison projects, JMIP/EMF35 (Sugiyama et al. 2020a) focusing on Japan, CD-LINKS 9 (Schaeffer et al. 2020; Roelfsema et al. 2020), COMMIT (van Soest et al. 2021), ENGAGE (Fujimori 10 et al. 2021), Paris Reinforce (Perdana et al. 2020; Nikas et al. 2021) each covering several 11 countries/regions from the following set of countries: Australia, Brazil, China, EU, India, Indonesia, 12 Japan, Korea, Russia, Thailand, USA, Vietnam. The remaining pathways stem from individual 13 modelling studies that were submitted/collected (Table II.11.). 14 15 Table II.11. National and regional mitigation pathways by modelling framework, region and scenario 16 type. Region Model CP NDC Other Total ARG IMACLIM-ARG 1 2 3 AUS TIMES-Australia 1 7 8 BRA BLUES-Brazil 2 2 15 19 BRA COPPE_MSB-Brazil 8 8 BRA IMACLIM-BRA 5 5 CHE STEM-Switzerland 1 11 12 CHN AIM/Hub-China 1 1 7 9 CHN C3IAM 3 11 14 CHN DREAM-China 1 1 CHN GENeSYS-MOD-CHN 3 3 Do Not Cite, Quote or Distribute II-81 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII CHN IPAC-AIM/technology-China 1 1 11 13 CHN PECE-China 2 2 CHN TIMES-Australia 1 1 CHN TIMES-China 1 2 8 11 ECU ELENA-Ecuador 2 2 ETH TIAM-ECN ETH 1 1 2 EU E4SMA-EU-TIMES 1 1 EU eTIMES-EU 23 23 EU JRC-EU-TIMES 8 8 EU PRIMES 2 2 9 13 EU REMIND_EU 9 9 FRA TIMES-France 8 8 GBR 7see 11 11 IDN AIM/Hub-Indonesia 2 2 IDN DDPP Energy 4 4 IND AIM/Enduse India 1 1 5 7 IND AIM/Hub-India 1 1 7 9 IND MARKAL-INDIA 2 3 13 18 JPN AIM/CGE-Enduse-Japan 6 6 JPN AIM/Enduse-Japan 3 3 69 75 JPN AIM/Hub-Japan 1 2 42 45 JPN DNE21-Japan 1 30 31 JPN DNE21+ V.14 (national) 1 1 4 6 JPN IEEJ-Japan 1 34 35 KEN TIAM-ECN KEN 1 1 2 4 KOR AIM/CGE-Korea 1 1 6 8 KOR AIM/Hub-Korea 1 1 7 9 MDG TIAM-ECN MDG 1 2 3 MEX GENeSYS-MOD-MEX 4 4 PRT TIMES-Portugal 1 3 4 RUS RU-TIMES 1 1 4 6 SWE TIMES-Sweden 4 4 THA AIM/Hub-Thailand 1 2 19 22 USA GCAM-USA 2 2 9 13 USA RIO-USA 12 12 Do Not Cite, Quote or Distribute II-82 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII VNM AIM/Hub-Vietnam 1 2 14 17 ZAF TIAM-ECN AFR 4 4 Total 29 39 466 534 1 Notes: The following scenario categories are distinguished in this table, CP = current policies, NDC = 2 implementation of Nationally Determined Contributions (NDCs) by 2025/30, Other = all other 3 scenarios. 4 5 3.4. Sector transition pathways 6 Sectoral transition pathways based on the AR6 Scenario database are addressed in a number of 7 Chapters, primarily Chapter 6 (Energy systems), 7 (AFOLU), 9 (Buildings), 10 (Transport) and 11 8 (Industry). These analyses cover both contributions from global IAMs and from sector-specific models 9 with regional or global coverage. The assessments cover a variety of perspectives, including long-term 10 global and macro-region trends for the sectors, sectoral analysis of the Illustrative Pathways, and 11 comparison of the scenarios between full-economy IAMs and sector-specific models on shorter time 12 horizons. These perspectives have a bi-directional utility – to understand how well IAMs are 13 representing sectoral trends from more granular models, and position sectoral models in the context of 14 full economy transitions to verify consistency with different climate outcomes. 15 16 Table II.12. Overview of how models and scenarios were used in sectoral chapters. All scenario and 17 model counts listed in the table are contained in the AR6 scenario database, with one exception: Chapter 18 9 (Buildings), which supplemented its dataset with a large number of scenarios separately pulled from the 19 sectoral literature. Scenario counts represents unique model-scenario combinations in the database. Sector # models # scenarios Key Key perspectives sections Energy 12 476 6.6 Regional and global energy system characteristics systems along mitigation pathways and at net-zero (Ch6) 18 536 6.7 emissions specifically: CO2 and GHG emissions; energy resource shares; electricity and hydrogen 13 776 6.7.1 shares of final energy; energy intensity; per-capita energy use; peak emissions; energy investments AFOLU 11 384 7.5.1 Regional and global GHG emissions and land use (Ch7) dynamics; economic mitigation potential for 14 572 7.5.2, different GHGs; integrated mitigation pathways 7.5.4 13 559 7.5.5 3 4 Do Not Cite, Quote or Distribute II-83 Total pages: 119 Final Government Distribution Annex III IPCC AR6 WGIII Buildings 80 82 9.3, 9.6 A mixture of top-down and bottom-up models. (Ch9) The former were either national, regional or global (of which 2 (of which 4 while the latter were global only with a breakdown are in AR6 are in AR6 per end use, building type, technologies and scenario scenario energy carrier database) database) Transport 24 1210 10.7 Global and regional transport demand, activity, (Ch10) modes, vehicles, fuels, and mitigation options. Industry 14 508 11.4.2 Global final energy use, CO2 emissions, carbon sequestration, fuel shares (Ch11) 1 Note 1: The number of models and scenarios reported in the table cannot be summed across chapters, as there is 2 considerable overlap in selected model-scenario combinations across chapters, depending on the filtering 3 processes used for relevant analyses. Moreover, the numbers in the table - and certainly not their sum - are not 4 intended to match those reported by Chap. 3 in Section II.3.2. 5 Note 2: Numbers shown in the model-count column are arrived at through the authors’ best judgement. This has 6 to do with the overlapping nature of unique model versions (within a given model family) as models evolve over 7 time. In this case, model versions with substantial overlap were considered the same model, whereas model 8 versions that differ significantly were counted as unique. For example, ‘MESSAGEix-GLOBIOM 1.0’ and 9 ‘MESSAGEix-GLOBIOM_1.1’ are counted as the same model, while ‘MESSAGEix-GLOBIOM 1.0’ and 10 ‘MESSAGE’ are counted as different. If instead counting all model versions uniquely, then the following counts 11 would apply to each chapter: Energy systems (30/38/29), AFOLU (18/27/25/4), Buildings (80), Transport (50), 12 Industry (32). 13 Note 3: The Transport chapter figures of Section 10.7 are produced from the final AR6 scenario database by the 14 code accompanying this report. The set of model and scenario names appearing in each plot or figure of 10.7 15 varies, depending on whether particular submissions to the database included the specific variables appearing in 16 that plot. Authors advise inspecting the data files accompanying each figure for the set of models/scenarios 17 specific to that figure, or running the code against the final database snapshot to reproduce the figures in question. 18 Do Not Cite, Quote or Distribute II-84 Total pages: 119