Editing IPCC:AR6/WGII/Chapter-16 (section)

==== 16.6.3.5 Large-scale Singular Events (RFC5) ====

<div id="h3-46-siblings" class="h3-siblings"></div>

This RFC, large-scale singular events (sometimes called tipping points or critical thresholds), considers abrupt, drastic and sometimes irreversible changes in physical, ecological or social systems in response to smooth variations in driving forces (accompanied by natural variability) ( [[#Oppenheimer--2014|Oppenheimer et al., 2014]] ; [[#O’Neill--2017|O’Neill et al., 2017]] ). SR15 [[IPCC:Wg2:Chapter:Chapter-3#3.5.2|Section 3.5.2.5]] presented four examples, including the cryosphere (West Antarctic ice sheet, Greenland ice sheet), thermohaline circulation (slowdown of the Atlantic Meridional Overturning Circulation), the El Niño-Southern Oscillation (ENSO) as a global mode of climate variability, and the role of the Southern Ocean in the global carbon cycle ( [[#Hoegh-Guldberg--2018b|Hoegh-Guldberg et al., 2018b]] ). While most of the literature assessed here focuses on the resultant changes to climate-related hazards such as sea level rise, in this assessment, evidence about the implications of accelerated sea level rise for human and natural systems is also considered. If sea level rise is accelerated by ice sheet melt, the associated impacts are projected to occur decades earlier than otherwise, directly affecting coastal systems including cities and settlements by the sea (CCP2) and wetlands (Chapter 2). The associated disruption to ports is projected to severely compromise global supply chains and maritime trade with local–global geo-political and economic consequences. To compensate for this acceleration, adaptation would need to occur much faster and at a much greater scale than otherwise, or indeed than has previously been observed (CCP2). The costs of accommodating port growth and adapting to sea level rise amount to USD 22–768 billion before 2050 globally ( ''medium evidence'' , ''high agreement'' ) (see Sections 2.1, 2.2; Cross-Chapter Box SLR in Chapter 3).

In AR5 Section 19.6.3.6 ( [[#Oppenheimer--2014|Oppenheimer et al., 2014]] ), the boundary between undetectable and moderate risk is set at levels between 0.6°C and 1.6°C above pre-industrial levels (i.e., 0°C and 1°C above the 1986–2005 level) with ''high confidence'' , based on emerging early-warning signals of regime shifts in Arctic and warm water coral reef systems. The risk transition boundary between moderate and high risk was set between 1.6°C and 3.6°C above pre-industrial levels (i.e., 1°C and 3°C above the 1986–2005 level), with ''medium confidence'' based on projections of ice sheet loss, with faster increase between 1°C and 2°C than between 2°C and 3°C. The literature available at the time did not allow AR5 to assess the boundary between high and very high risk.

In SR15 [[IPCC:Wg2:Chapter:Chapter-3#3.5.2|Section 3.5.2.5]] ( [[#Hoegh-Guldberg--2018b|Hoegh-Guldberg et al., 2018b]] ), new assessments of the potential collapse of the West Antarctic ice sheet (WAIS) initiated by marine ice sheet instability (MISI) resulted in lowering the upper end of the transition from undetectable and moderate risk from 1.6°C to 1°C warming above pre-industrial levels, and lowering the upper end of the transition from moderate to high risk to 2.5°C. Although SR15 did not produce embers beyond 2.5°C, authors reported that the transition to very high risk was assessed as lying above 5°C in light of growing literature on ice sheet contributions to SLR.

AR6 provides new evidence that relates to the location of the transition from undetectable to moderate risk. At the time of SR15, observations were suggesting that MISI might already be taking place in some parts of the WAIS, while AR5 supported assessment of an additional MISI contribution to SLR of several additional tenths of a metre over the next two centuries. Since SR15, new observations (WGI AR6 [[IPCC:Wg2:Chapter:Chapter-9#9.4.2.1|Section 9.4.2.1]] , [[#Fox-Kemper--2021|Fox-Kemper et al., 2021]] ) support the assessment of enhanced grounding line retreat and subsequent mass loss through basal melt in various parts of Antarctica, and year 2100 sea level projections for the RCP8.5 scenario have increased by 10–12 cm owing to ice dynamics. However, the onset of MISI is driven by ocean warming in specific locations (ice cavities beneath floating ice shelves), and the relation between these ocean temperatures and global mean temperature is indirect and ambiguous. In addition, MISI implies a self-sustaining instability in the absence of further forcing. Because forcing is still increasing, it cannot be unambiguously assessed whether MISI is driving the observed retreat of grounding lines in the WAIS, or whether this retreat is a purely forced response (and would stop if the warming stops) or is just a manifestation of natural variability in upwelling of warmer waters on the Antarctic continental shelves and, as a result, is just a temporary effect. Consistent with SROCC, AR6 states with ''medium confidence'' that sustained mass losses of several major glaciers in the Amundsen Sea Embayment (ASE) are compatible with the onset of MISI, but that whether unstable WAIS retreat has already begun or is imminent remains a critical uncertainty.

Whether associated with MISI or not, WGI AR6 ( [[#Fox-Kemper--2021|Fox-Kemper et al., 2021]] ) now assesses with ''very high confidence'' that mass loss from both the Antarctic (whether associated with MISI or not) and Greenland Ice Sheets, is more than seven times higher over the period 2010–2016 than over the period 1992–1999 for Greenland and four times higher for the same time intervals for Antarctica. Given their multi-century commitments to global SLR, this reinforces the assessment of estimating the boundary between undetectable and moderate risks for ice sheets to lie between 0.7°C (the level of global warming in the 1990s when melting began to accelerate) and 1°C (as in SR15), with a median of 0.9°C.

In the Amazon Forest, increases in tree mortality and a decline in the carbon sink are already reported ( [[#Brienen--2015|Brienen et al., 2015]] ; [[#Hubau--2020|Hubau et al., 2020]] ), and old-growth Amazon Rainforest may have become a net carbon source for the period 2010–2019 ( [[#Qin--2021|Qin et al., 2021]] ). Estimates which include land use emissions indicate the region may have become a net carbon source ( [[#Gatti--2021|Gatti et al., 2021]] ). Fire activity is an important driver, and both bigger fires ( [[#Lizundia-Loiola--2020|Lizundia-Loiola et al., 2020]] ) and longer fire season ( [[#Jolly--2015|Jolly et al., 2015]] ) have been reported in South America, although this is strongly linked to land use and land use change as well as climate ( [[#Kelley--2021|Kelley et al., 2021]] ), and indeed land use change may be a stronger driver of potential loss of the Amazon Forest than climate change. The risk of climate-change-related loss of the Amazon Forest is assessed already above ‘undetectable’, but has only emerged over the last few years, when global warming had reached 1°C, and is linked to land use as well as GSAT levels. [[IPCC:Wg2:Chapter:Chapter-2|Chapter 2]] has assessed ecosystem carbon loss from tipping points in tropical forest and loss of Arctic permafrost, and finds a transition to moderate risk over the range 0.6–0.9°C ( ''medium confidence'' ). Specifically, WGII AR6 Table SM2.5 finds that ‘Primary tropical forest comprised a net source of carbon to the atmosphere, 2001–2019 (emissions 0.6 Gt y −1 , net 0.1 Gt y −1 ) ( [[#Harris--2021|Harris et al., 2021]] ). Anthropogenic climate change has thawed Arctic permafrost ( [[#Guo--2020|Guo et al., 2020]] ), carbon emissions 1.7 ± 0.8 Gt y −1 , 2003–2017 ( [[#Natali--2019|Natali et al., 2019]] )’. This also supports the upper limit for this transition lying at 1°C.

The potential global loss of an entire ecosystem type, coral reefs, is also considered a large-scale singular event. In the 1990s when global warming was around 0.7°C large-scale coral reef bleaching also became apparent ( [[#16.2.3.1|Section 16.2.3.1]] ), also supporting the lower boundary for this transition in respect of coral reefs.

Overall, given the above evidence on ice sheets, Amazon Forest and coral reefs, the transition from undetectable to moderate risk is assessed to occur between 0.7°C and 1°C warming with a median of 0.9°C with ''high confidence'' .

The transition from moderate to high risk is informed by an assessment of risks at higher levels of warming than present. Nearly all climate models show warmer temperatures around Antarctica in conjunction with rising global mean temperature, and all ice sheet models show sustained mass loss from the WAIS after temperature increase halts (thus implying MISI takes place) at various levels between 1.5°C and 5°C. An increasing fraction of ice sheet models shows additional sustained mass loss from the East Antarctic Ice Sheet (EAIS) for peak warming between 2°C and 4°C, and all ice sheet models show mass loss for peak warming higher than 4°C. Therefore, we assess an increasing link between MISI, WAIS collapse and Antarctic mass loss, for increasing temperature levels ( ''high confidence'' ).

There is ''high confidence'' in the existence of threshold behaviour of the Greenland Ice Sheet in a warmer climate (WGI AR6 Ch 9, [[#Fox-Kemper--2021|Fox-Kemper et al., 2021]] ); however, there is ''low agreement'' on the nature of the thresholds and the associated tipping points. Similarly, the likelihood for accelerated and irreversible mass loss from Antarctica increases with increasing temperatures, but thresholds cannot yet be unambiguously identified. By the year 2100, sea level projections (AR6 WGI Figure SPM.8 ( [[#IPCC--2021|IPCC, 2021]] )) now range from 0.57 m (0.37–0.85 m) for the SSP1–1.9 scenario to 1.35 m (1.02–1.89 m) for the SS5–8.5 scenario and become 1.99 m for the latter scenario (1.02–4.83 m) in the case of low-likelihood, high-impact outcomes resulting from ice sheet instability, for which there is ''limited evidence'' . It should be noted that inclusion of such low-likelihood, high-impact outcomes dominated by not-well-understood processes affecting ice dynamics on the large ice caps of Greenland, and in particular Antarctica, would also enhance the sea level projections for other scenarios, but to a lesser extent for increasingly weaker forcing. No quantitative assessment of their effect in other scenarios than SSP5–8.5 yet exists as such simulations with ice sheet models have not been carried out, or only in a very limited amount.

It should be noted that ice sheets may take many centuries to respond, implying that risk levels increase over time for the same warming level. Therefore, we base judgements about risk transitions related to ice sheets primarily on their implications for 2000-year commitments to SLR from sustained mass loss from both ice sheets as projected by various ice sheet models, reaching 2.3–3.1 m at 1.5°C peak warming and 2–6 m at 2.0°C peak warming (WGI AR6 TS, Box TS.4 Figure 1; [[#Arias--2021|Arias et al., 2021]] ). This is an important feature of the approach to this RFC (i.e., it is not primarily focused on implications for the next 100–200 years). In addition, since the AR5, there is new evidence about the Last Interglacial (LIG), when global mean temperature was about 0.5–1.5°C above the pre-industrial era. AR6 assesses that it is ''virtually certain'' that sea level was higher than today at that time, ''likely'' by 5–10 m ( ''medium confidence'' ) (B.5.4 WGI AR6 SPM,( [[#IPCC--2021|IPCC, 2021]] )). Mid-Pliocene temperatures of 2.5°C (about 3 million years ago when global temperatures were 2.5–4°C higher) also provide evidence as an upper limit for the transition to high risk associated with long-term equilibrium SLR of 5–25 m (WGI AR6 SPM B.5.4). Projected SLR for 2300 in an RCP8.5 or SSP5–8.5 scenario (consistent with a peak warming range of 4–6°C, varies between 1.7–6.8 m and 2.2–5.9 m, respectively (WGI AR6 TS Box TS.4, [[#Arias--2021|Arias et al., 2021]] ), and when accounting for marine ice cliff instability taking place on Antarctica, these numbers may increase to a range of 9.5–16.2 m (WGI AR6 TS Box TS.4, [[#Arias--2021|Arias et al., 2021]] ).

CMIP6 climate models project drying in the Amazon—especially in June–July–August, irrespective of future forcing scenario, but which increases with GSAT/higher scenarios ( [[#Lee--2021|Lee et al., 2021]] ). For higher GSAT levels, [[#Burton--2021|Burton et al. (2021)]] explore different forcing scenarios and found, regardless of scenario, burned area increases markedly with GSAT. New understanding of the role of vegetation stomata will act to exacerbate this drying ( [[#Richardson--2018b|Richardson et al., 2018b]] ). A transition to high risk of savannisation for the Amazon alone was assessed to lie between 1.5°C and 3°C with a median value of 2.0°C. A mean temperature increase of 2°C could reduce Arctic permafrost area ~15% by 2100 (Comyn-Platt et al., 2018). [[IPCC:Wg2:Chapter:Chapter-2|Chapter 2]] has assessed ecosystem carbon loss from tipping points in tropical forest and loss of Arctic permafrost, and finds a transition from moderate to high risk over the range 1.5°C to 3°C with a median of 2°C ( ''medium confidence'' , Table SM2.5, Figure 2.11). Its assessment of the transition from high to very high risk is located over the range 3–5°C ( ''low confidence'' , Table SM2.5, Figure 2.11) based on the potential for Amazon Forest dieback between 4°C and 5°C temperature increase above the pre-industrial period ( [[#Salazar--2010|Salazar and Nobre, 2010]] ).

One of the criteria for locating a transition to very high risk is a limited ability to adapt. In natural systems, limiting warming to 1.5°C rather than 2°C would enhance the ability of coastal wetlands to adapt naturally to SLR, since natural sedimentation rates more likely keep up with SLR (SR15, Hoegh-Guldberg 2018). In human systems, there is ''medium confidence'' that technical limits will be reached for hard protection to SLR beyond 2100 under high-emissions scenarios, with limits associated with socioeconomic and governance issues reached before 2100 (CCP2).

We therefore estimate the boundary between moderate and high risk to lie between 1.5°C and 2.5°C, with a median at 2.0°C, with ''medium confidence'' based on projections for melting ice sheets and drying in the Amazon. We also estimate the boundary between high and very high risk to lie between 2.5°C and 4°C, but with ''low confidence'' due to uncertainties in the projections of SLR at higher levels of warming and differences between levels of warming at which very high risks were assessed in different systems.

<div id="cross-working-group-box-economic" class="h2-container box-container"></div>

'''Cross-Working Group Box ECONOMIC | Estimating Global Economic Impacts from Climate Change'''
<div id="h2-26-siblings" class="h2-siblings"></div>

Authors: Steven Rose (USA), Delavane Diaz (USA), Tamma Carleton (USA), Laurent Drouet (Italy), Celine Guivarch (France), Aurélie Méjean (France), Franziska Piontek (Germany)

This Cross-Working Group Box assesses literature estimating the potential global aggregate economic costs of climate change and the social cost of carbon (SCC), where the former are sometimes referred to as estimates of global ‘climate damages’ and the latter are estimates of the potential monetised impacts to society of an additional metric ton of carbon dioxide emitted to the atmosphere. These measures include the economic costs of climate change that could be felt in market sectors such as agriculture, energy services, labour productivity and coastal resources, as well as non-market impacts such as other types of human health risks (including mortality effects) and ecosystems. Global economic impacts estimates can inform decisions about global climate management strategy, while SCC estimates can inform globally incremental emissions decisions. In practice, economic damage estimates have been used to explore economically efficient (‘economically optimal’) global emissions pathways (e.g., [[#Nordhaus--2017|Nordhaus and Moffat, 2017]] ), while SCCs have been used to inform federal and state-level policy assessment in some countries ( [[#Greenstone--2013|Greenstone et al., 2013]] ; [[#Rose--2016|Rose and Bistline, 2016]] ), but the type of SCC and application matters ( [[#Rose--2017|Rose, 2017]] ). This literature has been assessed in previous WGII reports (e.g., [[#Arent--2014|Arent et al., 2014]] ), and this box serves this need for this report. The assessment in this box was performed jointly across WGII and WGIII, building on the foundation of WGII AR6 Chapter 16’s ‘Risk to living standards’ assessment ( [[#16.5.2.3|Section 16.5.2.3.4]] ), which includes consideration of severe risks to global aggregate economic output, and WGIII AR6 Chapter 3’s assessment of the benefits of mitigation. It also informs Chapter 16’s global aggregate impacts Reasons for Concern and supports Chapter 18’s assessment of global emissions transitions, risk management and climate resilient development. In keeping with the broad risk framing presented in [[IPCC:Wg2:Chapter:Chapter-1|Chapter 1]] of this report, other lines of evidence regarding climate risks, beyond monetary estimates, should be considered in decision making, including key risks and Reasons for Concern.

'''Methods for estimating global economic costs of climate impacts'''
There are several broad approaches to estimating climate damages, including biophysical process models, structural economic models, statistical methods (also called empirical or econometric) and hybrid approaches, with each methodology having strengths and weaknesses. Process models simulate physical, natural science and/or engineering processes and their response to climate variables, which are then monetised (e.g., [[#Anthoff--2014|Anthoff and Tol, 2014]] ; [[#Sieg--2019|Sieg et al., 2019]] ; [[#Narita--2020|Narita et al., 2020]] ). Process approaches have the advantage of being explicit and interpretable, though they can be computationally intensive; may omit relevant impact channels, interactions and market dynamics affecting valuation; and often lack a rigorous empirical basis for calibration (Fisher-Vanden et al.). Structural economic modelling represents climate impacts on inputs, production, household consumption, aggregate investment, and markets for economic sectors and regional economies (e.g., [[#Reilly--2007|Reilly et al., 2007]] ; [[#Roson--2012|Roson and Van der Mensbrugghe, 2012]] ; [[#Anthoff--2014|Anthoff and Tol, 2014]] ; [[#Dellink--2019|Dellink et al., 2019]] ; Takakura et al., 2019), often using computable general equilibrium (CGE) frameworks. Structural models can evaluate how market and non-market impacts might enter and transmit through economies, and adaptation responses within input and output markets, consumer and investment choices, and inter-regional trade (e.g., [[#Darwin--2001|Darwin and Tol, 2001]] ; [[#Dellink--2019|Dellink et al., 2019]] ; Takakura et al., 2019). Statistical methods estimate economic impacts in a given sector (e.g., [[#Auffhammer--2018|Auffhammer, 2018]] ) or in aggregate (e.g., [[#Dell--2014|Dell et al., 2014]] ; [[#Burke--2015|Burke et al., 2015]] ; [[#Hsiang--2017|Hsiang et al., 2017]] ; [[#Pretis--2018|Pretis et al., 2018]] ; [[#Kahn--2019|Kahn et al., 2019]] ), inferred from observed changes in economic factors, weather and climate, with responses and net results constrained by available data. Since AR5, hybrid approaches have taken different forms to integrate process, statistical and/or structural methods, and represent a potentially promising means of leveraging the strengths of different approaches (e.g., [[#Moore--2015|Moore and Diaz, 2015]] ; and [[#Hsiang--2017|Hsiang et al., 2017]] ; [[#Moore--2017a|Moore et al., 2017a]] ; [[#Ricke--2018|Ricke et al., 2018]] ; [[#Yumashev--2019|Yumashev et al., 2019]] ; [[#Chen--2020b|Chen et al., 2020b]] ). There is also a small literature that uses expert elicitation to gather subjective assessments of climate risks and potential economic impacts ( [[#Nordhaus--1994|Nordhaus, 1994]] ; [[#IPCC--2019a|IPCC, 2019a]] ; [[#Pindyck--2019|Pindyck, 2019]] ).

In addition to differences in methods, there are also differences in scope—geographic, sectoral and temporal. Global estimates are frequently based on an aggregation of independent sector and/or regional modelling and estimates; however, there are examples of estimates from global modelling that simulate multiple types of climate impacts and their potential interactions within a single, coherent framework (e.g., [[#Roson--2012|Roson and Van der Mensbrugghe, 2012]] ; [[#Dellink--2019|Dellink et al., 2019]] ; Takakura et al., 2019). Differences in scope also represent strengths and weaknesses between the methodologies, with narrower scope allowing for more detailed assessment, but missing potential interactions with the scope not covered (e.g., other geographic areas, sectors, markets or periods of time).

Comprehensive economic estimates are challenging to produce for many reasons, including complex interactions among physical, natural and social systems; pervasive climate, socioeconomic and system response uncertainties; and the heterogeneous nature of climate impacts that vary across space and time. Critiques and commentaries of global estimation methods ( [[#Pindyck--2013|Pindyck, 2013]] ; [[#Stern--2013|Stern, 2013]] ; [[#van%20den%20Bergh--2015|van den Bergh and Botzen, 2015]] ; [[#Cropper--2017|Cropper et al., 2017]] ; [[#Diaz--2017|Diaz and Moore, 2017]] ; [[#Pindyck--2017|Pindyck, 2017]] ; [[#Rose--2017|Rose et al., 2017]] ; [[#Stoerk--2018|Stoerk et al., 2018]] ; [[#DeFries--2019|DeFries et al., 2019]] ; [[#Pezzey--2019|Pezzey, 2019]] ; [[#Calel--2020|Calel et al., 2020]] ; [[#Warner--2020|Warner et al., 2020]] ; [[#EPRI--2021|EPRI, 2021]] ; [[#Grubb--2021|Grubb et al., 2021]] ; [[#Newell--2021|Newell et al., 2021]] ) include, among other things, concerns about statistical methods estimating weather but not climate relationships, making out-of-sample extrapolations, and model specification uncertainty, concerns about the observational grounding of structural modelling, and overall concerns about the lack of adaptation consideration, as well as representation and evaluation of potential large-scale singular events such as ice sheet destabilisation or biodiversity destruction, some questioning the ability to generate robust estimates (i.e., estimates insensitive to reasonable alternative inputs and specifications), and general concerns about methodological details, transparency and justification.

Additional methodological challenges to address (see, for instance, [[#EPRI--2021|EPRI, 2021]] ; [[#Piontek--2021|Piontek et al., 2021]] ) include how to capture and represent uncertainty and variability in potential damage responses for a given climate and societal condition, combine estimates from different methods and sources (including aggregating independent sectoral and regional results), assess sensitivity and evaluate robustness of estimates (including sensitivity to model specification), capture interactions and spillovers between regions and sectors, estimate societal welfare implications (versus gross domestic product [GDP] changes) of market and non-market impacts, consider distributional effects, represent micro- and macro-adaptation processes (and adaptation costs), specify non-gradual damages and nonlinearities, and improve understanding of potential long-run economic growth effects. Note that the treatment of time preference, risk aversion and equity considerations have important welfare implications for the aggregation of both potential economic impacts and climate change mitigation costs.

In addition to updated and new methods and estimates, newer literature has explored non-gradual damages, such as climatic and socioeconomic tipping points ( [[#Lontzek--2015|Lontzek et al., 2015]] ; [[#Méjean--2020|Méjean et al., 2020]] ), potential damage to economic growth (e.g., [[#Burke--2015|Burke et al., 2015]] ; [[#Moore--2015|Moore and Diaz, 2015]] ), valuing uncertainty in potential damages ( [[#Jensen--2014|Jensen and Traeger, 2014]] ; [[#Lemoine--2016|Lemoine and Traeger, 2016]] ; Cai and Lontzek) and representing adaptation (Takakura et al., 2019; [[#Carleton--2020|Carleton et al., 2020]] ; [[#Rode--2021|Rode et al., 2021]] ). Going forward, to help advance science and decisions, a key research priority is to understand and evaluate methodological strengths and weaknesses in damage estimation, and reconcile the differences affecting comparability in such a way that it informs use of the different lines of evidence. This will require greater transparency and assessment of details and assumptions in individual methods, communication and evaluation of alternatives for specifying or calibrating climate damage functional representations with respect to climate and non-climate drivers and potential nonlinearities, including evaluating data sufficiency for levels within and beyond observations and for characterising physical system dynamics, and evaluating the sensitivity of results to model specification and input parameter choices ( [[#Cropper--2017|Cropper et al., 2017]] ). Improving the robustness of economic impact estimates is an active area of research. Below we describe the latest estimates.

'''Global estimates of the economic costs of climate impacts'''
Since AR5, many new estimates of the global economic costs of climate change have been produced. Figure Cross-Working Group Box ECONOMIC.1 shows a wide spread of estimates, with growing variance at higher levels of warming, both within and across methodology types (i.e., statistical, structural or meta-analysis). Meta-analysis is used here to refer to studies that treat other studies’ estimates as data points in an attempt to derive a synthesised functional form.

Global aggregate economic impact estimates (Figure Cross-Working Group Box ECONOMIC.1) are generally found to increase with global average temperature change, as well as vary by other drivers, such as income and population and the composition of the economy. Most estimates are nonlinear with higher marginal economic impacts at higher temperature, although some recover declining marginal economic impacts, and functional forms cannot be determined for all studies. The drivers of nonlinearity found in economic impact estimates, and the differences in nonlinearity across estimates (e.g., convex versus concave, degree of curvature), are not well understood, with methodology construction, assumptions and data all being potential factors. Relative to AR5, there have been more estimates and greater variation in estimates, including some recent estimates significantly higher than the range reported in AR5. For most of the studies shown in Figure Cross-Working Group Box ECONOMIC.1, the visible variation within a study represents alternative socioeconomic projections and climate modelling, not economic impacts response uncertainty for a given socioeconomic and climate condition. Response uncertainty could be significant, as indicated by some of the results shown in the figure (e.g., [[#Burke--2015|Burke et al., 2015]] ; [[#Rose--2017|Rose et al., 2017]] ), but methodological differences in how uncertainty is characterised (model specification, errors and confidence intervals versus distributions of results) limit comparability and assessment. Note that modelling factors between global temperature change and the economic impact calculation, such as regional temperature pattern assumptions or assumed SLR dynamics, can also impact calculated estimates (e.g., Warren et al.., 2021 PAGE09 estimates versus those in Rose et al.., 2017, Chen et al.., 2020 PAGE-ICE estimates versus Burke et al.., 2015).

From Figure Cross-Working Group Box ECONOMIC.1, we find a large span of damage estimates, even without considering uncertainty/confidence in damage responses, including for today’s level of warming (about 1°C). There is also evidence that some regions benefit from low levels of warming, leading to net benefits globally at these temperatures. The size of the span of estimates grows with global warming level, with variation across statistical estimates larger than variation in structural estimates. The structural and meta-analyses estimates appear to be in closer agreement, but that outcome is contingent on the meta-analyses’ data considerations and approach. Meta analyses to date have not assessed the alternative methods and dealt with the lack of comparability between methods.

[[File:8d6b177d58ccc741c8088e43921be606 IPCC_AR6_WGII_Figure_16_Cross-Working_Group_Box_ECONOMIC-1.png]]

'''Figure Cross-Working Group Box ECONOMIC.1 |''' '''Global aggregate economic impact estimates by global warming level (annual % global GDP loss relative to GDP without additional climate change).''' Top row panels present estimates by methodology type: (a) statistical modelling, (b) structural modelling and (c) meta-analyses, with all estimates from a paper in the same colour and estimates from methodologies other than that highlighted by the panel in grey for reference. Second row left panel (d) presents AR5 estimates. Second row right panel (e) presents all estimates in one figure, with the same colours as panels (a–d) using outlined dots for the statistical modelling estimates, solid dots for structural modelling estimates, and triangles for meta-analysis estimates. In all panels, lines represent functions, with dashed and dotted lines 5th and 95th percentile functions from structural modelling. To avoid duplication, estimates from papers using the economic impacts estimates or model formulations already represented in the figure are not included (e.g., [[#Diaz--2017|Diaz and Moore, 2017]] ; [[#Chen--2020b|Chen et al., 2020b]] ; [[#Glanemann--2020|Glanemann et al., 2020]] ; [[#Warren--2021|Warren et al., 2021]] ). The exception is Burke et al. (2018), with the different estimates shown representing variation across climate scenarios for a given aggregate economic impacts specification from Burke et al. (2015)—the ‘pooled, short run’ statistical specification. Results shown for the latter are estimates with the author’s different statistical model specifications (and a fixed climate scenario, SSP5). From top to bottom, the Burke et al. (2015) estimates are for the ‘pooled, long run’, ‘differentiated, long run’, ‘pooled, short run’ (authors’ base case) and ‘differentiated, short run’ statistical specifications. For [[#Howard--2017|Howard and Sterner (2017)]] , the authors’ preferred function is shown. Overall, estimates shown in the figure can correspond to different future years, reflecting different socioeconomic conditions and climate pathways to a global warming level. Global average temperature change bars relative to the period 1850–1900 are shown below the economic cost estimates to provide context to potential future warming. Shown are the WGI AR6 assessed best estimates and 90% intervals for the illustrative emissions scenarios considered for the near term 2021–2040, mid-term 2041–2060 and long term 2081–2100.

Differences in methodology type and scope complicate comparison, assessment and synthesis ( [[#Cropper--2017|Cropper et al., 2017]] ; [[#Diaz--2017|Diaz and Moore, 2017]] ; [[#EPRI--2021|EPRI, 2021]] ; [[#Piontek--2021|Piontek et al., 2021]] ). In particular, structural economic modelling and empirical aggregate output modelling are fundamentally different, which has been identified as an issue affecting the comparability of results ( [[#Cropper--2017|Cropper et al., 2017]] ). The different methodologies affect outcomes, with global aggregate estimates based on statistical methodologies typically higher than those from structural modelling (Figure Cross-Working Group Box ECONOMIC.1). This is, in part, due to the relationships in observational data captured by statistical modelling, assumed persistence of impacts in statistical modelling, broader adaptation responses in structural modelling, and differences in the representation of future societies and how they might evolve, respond and interact. Within statistical modelling, results are also found to be very sensitive to the statistical model specification (e.g., [[#Burke--2015|Burke et al., 2015]] ; [[#Newell--2021|Newell et al., 2021]] ). Within structural modelling, differences in representations of biophysical changes and economic structural dynamics contribute to differences across structural estimates (e.g., [[#Rose--2017|Rose et al., 2017]] ).

The wide range of estimates, and the lack of comparability between methodologies, does not allow for identification of a robust range of estimates with confidence ( ''high confidence'' ). Evaluating and reconciling differences in methodologies is a research priority for facilitating use of the different lines of evidence ( ''high confidence'' ). However, the existence of higher estimates than AR5 indicate that global aggregate economic impacts could be higher than previously estimated ( ''low confidence'' due to the lack of comparability across methodologies and robustness of estimates).

While Figure Cross-Working Group Box ECONOMIC.1 summarises global aggregate estimates, the literature exhibits significant heterogeneity in regional economic impacts that are also sensitive to methodology, model specification and societal assumptions (with, for instance, larger estimates due to the assumed size of society, but offsetting adaptive capacity improvements and adaptation responses). Regional results illustrate the potential for overall net benefits in more temperate regions at lower levels of warming with potential lower energy demand and comparative advantages in agricultural markets; however, at higher levels of warming, net losses are estimated. In addition, economic impacts for poorer households and poorer countries represent a smaller share in aggregate quantifications expressed in GDP terms than their influence on well-being or welfare ( [[#Byers--2018|Byers et al., 2018]] ; [[#Hallegatte--2020|Hallegatte et al., 2020]] ).

'''Social cost of carbon methods and estimates'''
The global economic impact estimates discussed in the previous section serve as a key input into the calculation of the value of potential net damages caused by a marginal ton of carbon dioxide emissions, or the SCC. To compute an SCC, damage estimates are commonly combined in a multi-century modelling framework with socioeconomic and emissions projections, a physical model of the climate, including a SLR component, and assumptions about the discount rate, with current frameworks having highly stylised representations of these components. Though we do not present quantitative estimates here, due to the challenge of comparability, for economic impacts methodologies (as discussed above) as well as other SCC estimation elements, large variations in SCC estimates are found in the literature assessed due to, among other things, differences in modelling component representations, input and parameter assumptions, considerations of uncertainty, and discounting, inflation, and emissions year (e.g., [[#Tol--2009|Tol, 2009]] ; [[#Tol--2018|Tol, 2018]] ; [[#Pezzey--2019|Pezzey, 2019]] ; [[#Iese--2021|Iese et al., 2021]] ). There are also different ‘variants’ of SCC estimates that differ conceptually, and in magnitude, depending on the reference condition for evaluating the impact of a marginal metric ton—is it being evaluated relative to a no-climate-policy baseline, an economically efficient pathway that weighs the benefits and costs of emissions mitigation, or a pathway based on a particular climate policy or goal such as 2°C or a concentration target ( [[#Rose--2017|Rose et al., 2017]] )? The variant of SCC has implications for its applicability to different policy contexts ( [[#Rose--2016|Rose and Bistline, 2016]] ).

In addition to the economic impacts methodological challenges discussed above with respect to aggregate economic impact estimates, the additional components needed for SCC calculations give rise to a new set of technical issues and critiques, including incorporation of uncertainties in the components beyond climate damages, links between components, and discounting ( [[#van%20den%20Bergh--2015|van den Bergh and Botzen, 2015]] ; [[#Cropper--2017|Cropper et al., 2017]] ; [[#Diaz--2017|Diaz and Moore, 2017]] ; [[#Pindyck--2017|Pindyck, 2017]] ; [[#Rose--2017|Rose et al., 2017]] ; [[#EPRI--2021|EPRI, 2021]] ). For component-specific discussions and assessment, see [[#Cropper--2017|Cropper et al. (2017)]] , Rose et al. (2017) and [[#EPRI--2021|EPRI (2021)]] .

Substantial progress has been made in recent years to better reflect complexities in the global economy, the climate system, and their interaction. For example, recent studies have explored damages to natural capital ( [[#Bastien-Olvera--2021|Bastien-Olvera and Moore, 2021]] ), the influence of imperfect substitutability between environmental services and market goods ( [[#Sterner--2008|Sterner and Persson, 2008]] ; [[#Weitzman--2012|Weitzman, 2012]] ; [[#Drupp--2021|Drupp and Hänsel, 2021]] ), the implications of heterogeneous climate change impacts across income groups ( [[#Dennig--2015|Dennig et al., 2015]] ; [[#EPRI--2021|EPRI, 2021]] ; [[#Errickson--2021|Errickson et al., 2021]] ), the potential for persistent climate impacts to economic growth instead of effects on levels of economic output ( [[#Dietz--2015|Dietz and Stern, 2015]] ; [[#Moore--2015|Moore and Diaz, 2015]] ; [[#Ricke--2018|Ricke et al., 2018]] ; [[#Kikstra--2021|Kikstra et al., 2021]] ; [[#Newell--2021|Newell et al., 2021]] ), valuing the risks of climate tipping points ( [[#Cai--2019|Cai and Lontzek, 2019]] ; Rising et al., 2020), valuing uncertainty under risk aversion ( [[#Jensen--2014|Jensen and Traeger, 2014]] ; [[#Lemoine--2016|Lemoine and Traeger, 2016]] ), and modelling a distinction between intertemporal inequality aversion and risk aversion in the social welfare utility function ( [[#Crost--2013|Crost and Traeger, 2013]] ; [[#Jensen--2014|Jensen and Traeger, 2014]] ; [[#Daniel--2015|Daniel et al., 2015]] ). These new studies have, in general, raised estimates of the SCC ( [[#Crost--2013|Crost and Traeger, 2013]] ; [[#Jensen--2014|Jensen and Traeger, 2014]] ; [[#Gerlagh--2015|Gerlagh and Michielsen, 2015]] ; [[#Moore--2015|Moore and Diaz, 2015]] ; [[#Faulwasser--2018|Faulwasser et al., 2018]] ; [[#Guivarch--2018|Guivarch and Pottier, 2018]] ; [[#Budolfson--2019|Budolfson et al., 2019]] ; [[#Cai--2019|Cai and Lontzek, 2019]] ; [[#Dietz--2019|Dietz and Venmans, 2019]] ; [[#Kalkuhl--2020|Kalkuhl and Wenz, 2020]] ), in some cases by an order of magnitude ( [[#Ricke--2018|Ricke et al., 2018]] ). However, challenges persist in terms of moving from conceptual to practical application, such as pinning down parameter specifications, modelling specific mechanisms for impacts, and more fully representing adaptation.

Despite these scientific advances, SCC estimates vary widely in the literature. Technical issues with past and current modelling (e.g., [[#Pezzey--2019|Pezzey, 2019]] ; [[#Pindyck--2019|Pindyck, 2019]] ; [[#EPRI--2021|EPRI, 2021]] ) and the challenge of comparability across methodologies imply that many estimates are not robust ( ''high confidence'' ). Also, as a result, the issue of directional bias of past estimates remains unsettled. Better representation of uncertainty in methods can improve robustness, while detailed methodology assessment and comparison will help define the relative biases of methods ( ''high confidence'' ).

'''Application to decision making'''
The literature has also assessed the application of aggregate economic impact cost and SCC estimates ( [[#Rose--2016|Rose and Bistline, 2016]] ; [[#Rose--2017|Rose et al., 2017]] ; [[#Kaufman--2020|Kaufman et al., 2020]] ) and identified conceptual and technical issues that need to be considered when using results to inform policy decisions. These issues include: accounting for endogenous marginal benefits and socioeconomic conditions in evaluating policies with non-incremental global emissions implications; consistency in assumptions and treatment of uncertainty across benefit and cost calculations; fully accounting for the streams of both mitigation costs and benefits over time; avoiding inefficiently valuing or pricing emissions more than once across policies and jurisdictions; and accounting for emissions leakage to capture net climate implications. Furthermore, concerns about the robustness of estimates have led some to recommend considering alternatives, such as using marginal mitigation cost estimates based on modelling of policy goals instead of the SCC (e.g., [[#Rose--2012|Rose, 2012]] ; [[#Pezzey--2019|Pezzey, 2019]] ; [[#Kaufman--2020|Kaufman et al., 2020]] ), although this comes with its own set of assumptions and technical challenges.

Cross-Working Group Box ECONOMIC

Cross-Working Group Box ECONOMIC

Cross-Working Group Box ECONOMIC

Cross-Working Group Box ECONOMIC

<div id="16.6.4" class="h2-container"></div>

<span id="summary"></span>