IPCC emission scenarios: How did critiques affect their quality and relevance 1990–2022?

Long-term global emission scenarios enable the analysis of future climate change, impacts


Introduction
Because climate change and its impacts extend into the distant future (IPCC, 1990a;O'Neill et al., 2017), long-term global (emission) scenarios have influenced climate research and assessments for at least 30 years (van Beek et al., 2020;van Vuuren et al., 2012).These scenarios are projections of future greenhouse gasses (GHG), air pollutants and aerosols, and future land use based on underlying projections for energy and food systems (Riahi et al., 2017).The output of emission scenarios (emissions) is used as input for 1) climate change research, 2) impact Abbreviations: AR1, IPCC First Assessment Report (AR2 = Second Assessment Report); BECCS, Bioenergy with Carbon Capture and Storage; CCS, Carbon Capture and Storage; COPs, Conferences of the Parties (under the UNFCCC); GHGs, Greenhouse Gasses; IAMs, Integrated Assessment Models; IPCC, Intergovernmental Panel on Climate Change; NETs, Negative Emissions Technologies; UNFCCC, United Nations Frmework Convention on Climate Change; SA90, Scientific Assessments 1990 (first generation of scenarios informing IPCC assessment reports); IS92, IPCC Scenarios 1992 (2nd scenario generation); SRES, Special Report on Emission Scenarios (3rd generation); RCPs, Representative Concentration Pathways (4th generation -emission pathways); SPAs, Shared Policy Assumptions (4th generation -policy assumptions); SSPs, Shared Socioeconomic Pathways (4th generation -storylines); Post-SRES, SRES mitigation scenarios published and analyzed in IPCC AR3; Intervention scenarios, Scenarios including mitigation policy assumptions (also called mitigation scenarios); Baselines, Scenarios not including mitigation policy assumptions; Convergence, Scenarios describing increased world equality (often via GDP/income).assessment, and finally, 3) mitigation analysis.Thus, these scenarios play a key role in linking different climate research disciplines (IPCC, 2014a(IPCC, , 1990b)), the Intergovernmental Panel on Climate Change's (IPCC) assessment reports (ARs), and have supported national and international policymaking, reflected in the Paris Agreement on Climate Change (UNFCCC/COP, 2015), and referred to in national climate pledges (UNFCCC, 2021) and policies (Baranzelli et al., 2013;Fawcett et al., 2015).Additionally, emissions scenarios enable the assessment of the effectiveness of the Paris Agreement (UNEP, 2020), represent a crucial feature defining future sustainability thinking (Otero et al., 2020), and the need and range of possible sustainable development policy actions (Raskin et al., 2005).
Developing scenarios is not straightforward.They are typically created through qualitative assumptions and quantifications using integrated assessment models (IAMs).They require projections of underlying human activity levels over the long term (van Vuuren et al., 2010) and complex methodologies to significantly discern scenario differences within a framework (Schweizer and O'Neill, 2014).Thus, assessments need to be made on possible future changes for many factors such as socioeconomic development, technology advances, and lifestyle change (O'Neill et al., 2017).Similar choices are made on focus areas and definitions (Raskin and Swart, 2020).
The prominent role and the uncertainties and (subjective) choices involved in the work have led to multiple critiques caused by factors such as changing contexts and roles (Girod et al., 2009;Moss et al., 2010), different worldviews (Parikh, 1992;Schneider, 2001), methodological advances (Schweizer and Kriegler, 2012), and model-focused method-assessments by scenario developers/modelers (Schweizer and O'Neill, 2014).There has been quite some literature on the critique, assessments, and responses.However, no attempt has been made to assess the critique systematically.This paper provides the first comprehensive overview of emission scenario critiques, their responses, and possible impact on the scenarios and scenario developments.The review focuses on the scenarios informing the IPCC assessment reports 1990-2022, as these are also the most prominent scenarios in the scientific literature and policy assessment (O'Neill et al., 2020;Wilkinson and Eidinow, 2008).The history of IPCC assessments covers four generations of emissions scenarios.Three series were developed inside the IPCC, comprising the "1990 IPCC First Scientific Assessment" (SA90), (IPCC, 1990a), the "1992 IPCC Scenarios" (IS92) (Leggett et al., 1992), and the 2000 "Special Report on Emissions Scenarios" (SRES) (Nakicenovic and Swart, 2000), and the additional Post-SRES including intervention (IPCC, 2001a).The most recent emissions scenarios were developed outside the IPCC (Moss et al., 2010), i.e., the "Representative Concentration Pathways" (RCPs) (van Vuuren et al., 2011) and the "Shared Socioeconomic Pathways" (SSPs) (O'Neill et al., 2014;Riahi et al., 2017).The RCPs informed AR5, while the SSP-RCP combinations informed AR6 (IPCC, 2021).This paper does not analyze the scenarios but explores how others have evaluated and perceived the art of emission scenario development in an IPCC context, including the four subsequent emissions scenario series that have informed IPCC assessments (Moss et al., 2010;van Vuuren and O'Neill, 2006).This paper aims to neutral reflect the critique in line with a review paper.It does not judge the quality or content of critiques.Nor do we focus on the hundreds of scenarios published by IAM groups individually, the work of EMF (EMF, 2020), or the IEA World Energy Outlook.At the same time, a considerable part of the discussion in this paper is also relevant to the broader scenario literature.
First, we conducted a systematic literature search.Relevant peerreviewed critical literature and responses were identified via IPCC Assessments (e.g., IPCC, 2001a), primary scenario literature, and their bibliographies (e.g., Nakicenovic and Swart, 2000), SCOPUS, Google, and Google Scholar database searches.The search terms included all combinations presented in Fig. 1.We selected literature critiques or papers presenting scenario improvements, and responses to critiques by reading titles, abstracts, introductions, conclusions, and the full text.We also reviewed the relevant references in the papers.The listed papers may not be complete, but the method provides a solid basis for assessing the main critique topics (Table 1), their evolution, and contribution (SI Tables 1 and 2).
Second, we categorized papers based on their focus: 1) scenario assumptions, 2) scenario range, 3) methodological issues, and 4) scenario relevance.Within each category, we identified thirteen subcategories (Table 1).In addition, each paper was classified based on its primary and secondary topics: primary topic (value = 1) and secondary topics "closely related to or a consequence of primary topic" (0.75), "supporting but having a less close relationship to the primary topic/key message" (0.5), and "additional topics with an arbitrary relationship to the primary topic" (0.25).See totals in Fig. 5.

Emission scenarios in the context of IPCC
Since 1990, four generations of emission scenarios have served as input to climate models and scenario-based literature informing successive IPCC Assessment Reports' (ARs) review of possible future climate change, impacts, and response strategies (IPCC, 1990a;Moss et al., 2010).Grounded in Working Group III (WG3, mitigation), emission scenarios are used by scientists in WG1 (climate science) and WG2 (impacts and adaptation) communities to analyze future outlookscutting across the three IPCC WGs (IPCC, 2014b(IPCC, , 1989a)).The SA90 scenarios were used directly for analyses in all AR1 WGs.Over time scenarios were more frequently analyzed in peer-reviewed literature informing IPCC assessments rather than being analyzed by IPCC authors.
IPCC WG3 facilitated the first three series, following IPCC procedures (Bolin, 2007;IAC, 2010).The first (SA90) was developed via scientific considerations (IPCC, 1990a, IPCC, 1989b).The second (IS92) and third (SRES) series were designed under explicit intergovernmental mandates (Leggett et al., 1992;Nakicenovic and Swart, 2000) adopted in IPCC sessions.Between 2003 and2006, the IPCC intergovernmental sessions decided to move scenario development outside the IPCC, leading to the fourth emission scenario generation (SSP-RCP).It was organized by IPCC but developed by the scientific community without constraining intergovernmental mandates.It was in line with the IPCC's aim to assess existing scientific knowledge (IPCC, 2006) rather than generate new data (Moss et al., 2010).

The four generations
The SA90s informed IPCC AR1 (IPCC, 1990a).They were developed between 1989 and 1990, led by the United States Environmental Protection Agency (USEPA) and the Dutch Environment Ministry (IPCC, 1990a).The IPCC was newly established by country delegates primarily from Environment Ministries.The contextual framing was that climate change is a real risk: the report aimed to explore emissions pathways and what can be done (Bolin, 2007).The scenarios comprised five GHGs and were constructed via two models, the USEPA's Atmospheric Stabilization Framework (ASF) supplemented by the Dutch Integrated Model for the Greenhouse Effect (IMAGE) (IPCC, 1990a(IPCC, , 1990c). 1 The four marker scenarios described a high emission (no-change) pathway called Business-as-Usual (BaU) (SA90-A), slow emissions growth via changed energy mix/efficiency (SA90-B), and two mitigation policies scenarios (SA90-C/D).An uncertainty range was defined by eight scenario variants describing higher and lower economic growth (IPCC, 1990a).
The IS92 informed AR2 (1995).They were an SA90-update, developed by the same models and team, which now also included economists.The period marked a political context shift.Two key parties debated opposing views with the Climate Convention adopted in 1992 (UNFCCC, 1992).The US 2 proposed an economic target-and-timetable approach to policy, while the EU delegation believed in a sciencebased-target approach, starting mitigation without fully understanding the problem (Bolin, 2007;Hecht and Tirpak, 1995;Oberthür and Ott, 1999).Intergovernmental delegates changed from environmental to more powerful departments within IPCC sessions.They asked new fundamental questions about climate change's reality and mitigation costs (Hecht and Tirpak, 1995;IPCC, 1990d).Several delegations, including the US 3 , argued that mitigation was premature.The session mandate (IPCC, 1991) excluded policy assumptions and higher emissions range (Edmonds et al., 1992;Pepper et al., 1992).The series included the full suite of GHGs (Alcamo et al., 1995;IPCC, 1996), more regional detail, and more diverse economy and population developments (IPCC, 1990a;Pepper et al., 1992).The series includes two high emission (IS92e/f), two low-emission (IS92c/d), and two nochange scenarios (IS92a/b) succeeding the SA90-A BaU (Leggett et al., 1992;Pepper et al., 1992).
2 Resistance towards climate mitigation was observed "from the energy industry, the US Senate, and Republican congressmen" in the United States (Bolin, 2007).During the 1982-1994 period, US officials were worried about the cost of an energy transition and favored less government regulation (Hecht and Tirpak, 1995).
Supplementary Information (SI) Tables 1 and 2 present overviews of the contemporary scientific and political context of the emission scenario generations, the scenario series' objectives, and the scientific and policy questions they generated.

The IPCC context
Fig. 2 illustrates the scenario development periods (dashed horizontal lines) and the periods of their inclusion in the scenario-based literature informing IPCC working groups (colored horizontal lines).Often impact (red line) assessment literature was the last to include the newest scenarios.For instance, IS92 informed scenario-based literature included in WG1 (green line) and WG3 (blue line) until AR3 ( 2001), while it continued in WG2 (red line) until AR4 ( 2007).
The scenario development periods have increased over time, spanning from about one year (SA90, IS92), three years (SRES), six (RCPs), to 13-15 years (SSP-RCPs).Furthermore, the process has grown more complex, implying increasing variables, disciplines, researchers, stakeholders, and more complex methods and assumptions.Simultaneously, more scenarios are produced in the scientific community and energy sector, making the current literature review process more complex than the SA90 and IS92 periods.
Fig. 3 presents the four scenario generations' key characteristics and changes over time.The figure includes convergence scenarios, which are not an essential characteristic but highlighted because it is addressed in several critiques.The emissions ranges (upper and lower levels) have expanded over time, while there is high continuity in GDP and energy emission driver ranges across the four generations (Fig. 3b).Interestingly, the SSP-RCPs' low-end emissions range is below the SA90.Despite not having explicit descriptions of climate policy, the IS92 and SRES low emission pathways (IS92c and SRES-B1/A1T) have quantifications like the SA90C/D and SSP1-2.6 low-emission policy scenarios.

Overall critique topics and timeline
We identified 280 peer-reviewed emissions scenario critiques and responses, and selected 80 grey literature publications (Fig. 4).Critiques intensified with the publication of the SRES, with 93% of peer-reviewed critiques published after 2000.
Over time, several biases might influence the graph, including the increasing number of papers published and plausible earlier publications that are difficult to trace in 2022.Still, we believe that the first scenario set received little critique.During the 1990s, several critiques addressed model methodology (e.g., Oreskes et al., 1994).In the past decade, several critiques addressed IAMs (Bellamy and Healey, 2018) without addressing the IPCC scenarios and are thus not included.
Fig. 5a shows that most peer-reviewed critiques addressed assumptions (40%) as their primary focus, and additionally, scenario ranges (25%), methodology (24%), and scenario user relevance, comprising policy relevance, policy implications, and probability critique (11%).Almost all critiques (78%) addressed assumptions as primary or secondary topics, while 66%, 48%, and 40% addressed scenario ranges, relevance, and methods, respectively.Policy relevance and implications were crosscutting issues often related as secondary topics to assumptions and/or range critiques.Method critiques were the least cross-cutting topic, primarily addressed as a scientific issue rather than a (politically) heated topic, and were seldom replicated in public media.Several critique topics have persisted for three decades, like assumptions on energy and emission ranges (Fig. 5b).Over time, more scientific attention was drawn to scenarios (Fig. 4).Resolution critiques emerged with the SRES and storyline method assessments during RCP/ SSP preparations.Other topics, like MER-PPP, probability, and IPCC incrowd (process) critiques, have decreased or disappeared.However, qualitative likelihood critiques (i.e., identify best-guess scenarios) have recently emerged as secondary topics (Hausfather and Peters, 2020).
Finally, some critiques have become more important, like methods applied, policy implications, negative emission technologies (NETs) assumptions, and missing scenarios critiques.

Key scenario critiques
This section presents the critiques in more detail.

Income convergence
The most impactful IS92-critique addressed a limited development worldview.It emphasized an assumed growing inequality between the global South and North in the IS92a (continuation-of-historical-trends) scenario (Parikh, 1992).For the IS92 (and SA90), the regional level scenarios were less developed (IPCC, 1990a), and thus global (in-) equality assumptions were less explicit.Technically, the critique resulted in an explicit global convergence narrative principle in the third (SRES) and fourth (SSP) generations (Nakicenovic and Swart, 2000;Riahi et al., 2017).Methodologically, the critique led to an IPCC scenario evaluation, encouraging a more inclusive scenario design process (Alcamo et al., 1995), leading to the IPCC panel mandating the inclusion of non-Annex-I developing country researchers and stakeholders in the successive scenario developments (IPCC, 1996).Within the UNFCCC climate negotiations, the projected inequality (Parikh, 1992) became a governing negotiation issue (Gupta and Hisschemöller, 1997;Okereke and Coventry, 2016) based on economic interests (Hecht and Tirpak, 1995;Oberthür and Ott, 1999) and injustice (Bos and Gupta, 2019).The global convergence assumptions became necessary because they represent drivers of emissions projections and explain the subsequent role in shaping UNFCCC policy negotiations, e.g., for the mitigation engagement of developing non-Annex-I countries.
A decade later, two letters to IPCC, published in a peer-reviewed journal, argued that the SRES used the wrong economic metrics (market-exchange rates (MER)).They argued that using MER disrupted the conditional convergence quantifications in the SRES-A1 and SRES-B1 scenarios (global convergence), leading to excessive economic growth assumptions in low-income regions, resulting in unrealistic high energy and emission levels (Castles andHenderson, 2003a, 2003b).IPCC authors accepted the suggested use of purchasing power parities (PPP) (Nakicenovic et al., 2003).Others stated that the non-peer-reviewed critique was misused to discredit climate change research (Van Vuuren and Alfsen, 2006) or a weak IPCC response (Tol, 2006).Because of limited PPP databases (starting from 1990 (WB, 2021)), historical PPP could not have been used in the SRES.Expert meetings and IPCC AR4 assessed PPP vs. MER-based scenarios (IPCC, 2007b, IPCC, 2005c).Successive research concluded that economic metrics had no significant influence on emissions ranges (Dixon and Rimmer, 2005;Holtsmark and Alfsen, 2005;IPCC, 2007bIPCC, , 2005c;;Manne et al., 2005;McKibbin et al., 2004;Pearce et al., 2004;Tol, 2006;Van Vuuren and Alfsen, 2006).In addition, the SRES assumptions of absolute emissions intensity convergence were questioned (Tol, 2006), showing weak evidence for "absolute" but strong evidence for energy per unit income "conditional" convergence (Miketa and Mulder, 2005).
During UK parliamentary hearings, an expert stated that IPCC was politicized, e.g., that the SRES regional GDP projections were adjusted upwards under pressure from African governments (House of Lords, 2005a).

Negative emissions technologies (NETs)
Because of continued policy delays, since the SA90 scenarios, it has become increasingly challenging to create low-emission scenarios aligned with the Paris Agreement.Thus, NETs gained a critical role in the modeling assumptions to achieve the Paris Agreement.NETs and assumptions on their costs became fundamental for the subsequent low emission pathway narratives (Gidden et al., 2019;IPCC, 2018a).The NETs ideas emerged in the late 1990s (Williams, 1998), describing that more CO 2 can be extracted from the atmosphere than released by humans.Throughout the 2000s, concepts like bioenergy with carbon capture and storage (BECCS) evolved further (Obersteiner, 2001), were picked up by models (Riahi et al., 2004(Riahi et al., , 2003;;van Vuuren et al., 2007), and included in the Paris Agreement (UNFCCC, 2015).Modelers and other researchers explored deep mitigation scenarios (Mori, 2000;Roehrl and Riahi, 2000) with and without BECCS (Edmonds et al., 2013), arguing that it could lighten mitigation costs (Edmonds et al., 2013;Kriegler et al., 2013;van Vuuren et al., 2013).The low-emission stabilization scenarios presented in IPCC AR1-3 (based on SA90, IS92, and SRES) were different from the scenarios presented in AR4-5 (based on RCP/SSP-RCPs) (Matsuno et al., 2012).The latter included negative emissions (Vaughan and Gough, 2016) with a broader mitigation range than previous assessments (Smith and Porter, 2018).Allowing net negative emissions in the RCP2.6 scenario made it logical to overshoot and subsequently compensate with negative emissions.As a result, several 1.5 • C and 2 • C-pathways rely on 'net negative' global carbon from 2050 (Workman et al., 2020), withdrawing between 260 and 1080 Gt CO 2 between 2020 and 2100 (IPCC, 2018a).

Emission ranges: Too-high or too-low
Emissions ranges are essential for assessing needs for mitigation (low emission pathways) and adaptation (high emission (and impact) pathways).The scenario ranges have been questioned for being too low and too high throughout the past three decades, e.g., reassessing the low and high emission scenarios, respectively.During the 1990s, global emissions grew at a similar speed as projected in medium-low emissions pathways (Pedersen et al., 2021).During that period, researchers argued that the IS92 emissions range was too high (Gray, 1998).Between 1999 and 2012, the World experienced a high emissions growth period (Pedersen et al., 2021), making researchers argue the SRES and RCP scenario ranges as potentially too low (Le Quéré et al., 2009;Peters et al., 2013;Raupach et al., 2007;Sheehan, 2008).During the successive period of overall slower growth (2013-2019) (Pedersen et al., 2021), researchers suggested that the SSP-RCP range was potentially too high (Hausfather and Peters, 2020).Based on the assumptions underlying the scenarios, SRES and RCP emissions ranges were questioned as too low (Anderson, 2015;Castles and Henderson, 2003a;Fuss et al., 2014;Pielke et al., 2008) or too high (Burgess et al., 2021;Castles and Henderson, 2003a;Christensen et al., 2018;Sanderson et al., 2011).These critiques including climate analyses informed IPCC AR3 and AR4 (Anderson et al., 2008;Ganguly et al., 2009;Reichstein, 2010;Romm, 2008).Modelers and others pointed out that RCP8.5 tracks cumulative historical CO 2 emissions (Pedersen et al., 2020;Schwalm et al., 2020) and that historical emissions are within emissions ranges (van Vuuren and Riahi, 2008) and tracking medium-high pathways (Pedersen et al., 2020).Modelers emphasized the fundamental differences underlying short-term fluctuations versus significant long-term trend breaks (Manning et al., 2010;van Vuuren et al., 2010;van Vuuren and Riahi, 2008).The range critiques did not impact SSP-RCP emissions ranges (Riahi et al., 2017).
Public media followed the fluctuations of scientific critiques.They first hinted that IPCC exaggerated temperature projections (Corcoran, 2002;Economist, 2003a), potentially compromising IPCC reports' reliability and policy relevance (Economist, 2003b, Economist, 2003a).Later, they questioned if IPCC climate projections were too conservative (Keulemans, 2020;Scherer, 2012).Such appraisals were critical from a policy perspective since emissions, and climate projections, inform the climate negotiations and national policies (Garnaut et al., 2008).
Several researchers argued for additional research on local risks and drivers of change (Cradock-Henry et al., 2018), such as institutional capacities (van Ruijven et al., 2013;Wilbanks and Ebi, 2014).SRES and SSPs do not explore conflict and security pathways (Nordås and Gleditsch, 2007).Civil war may reduce regional economic growth (Devitt and Tol, 2012).Here variables like equality, governance, and literacy may induce pacifying effects that can be implemented in scenarios (Andrijevic et al., 2020;Hegre et al., 2016).IPCC AR5 found climate impacts to increase conflict risks (IPCC, 2014a).SSP modelers argue that global conflict and governance extensions will support the SSPs (O'Neill et al., 2020).
Some researchers and modelers argue that scenarios preferred by policymakers might constrain scientific imagination and downplay structural discontinuity (Raskin and Swart, 2020).They problematize that economic growth is built into models (and policies) (Krakauer, 2014), despite also driving climate and environmental problems (Otero et al., 2020).To project sustainable development scenarios need, assumptions on nature-people relationships (Otero et al., 2020;Rosa et al., 2020) and the UN Sustainable Development Goals (SDGs) (Kriegler et al., 2018).Researchers advocate an increased focus on fundamental global system transformations (David Tàbara et al., 2018), lifestyles, values, institutions (Raskin, 2005(Raskin, , 2000)), and (weak) governance (Andrijevic et al., 2020).To guide policymakers, product developers, and consumers, modelers argued in favor of translating emission reductions into consumption levels (Girod et al., 2013).Additionally, international trade assumptions examining national emissions flows are less elaborated in the SSPs (Pedersen et al., 2021).SSP developers decided that narratives should inform analyses of global goals beyond those in the Paris Agreement (O'Neill et al., 2020).
Others advocated stakeholder inclusion on multiple levels (Girod and Mieg, 2008;Kok et al., 2007;Schenk and Lensink, 2007), contributing to adding locally relevant details (Cradock-Henry et al., 2018).Intensified scenario critiques after AR3 put pressure on IPCC delegates (IPCC, 2003) who decided that IPCC should facilitate rather than develop new scenarios (IPCC, 2005a) following scenario expert meeting recommendations (IPCC, 2007c, 2007a, IPCC, 2005d, IPCC, 2005c).Simultaneously, researchers argued that low funding support in developing countries limited regional scenario specifications (Wilbanks and Ebi, 2014).Thus there is a need for increased local stakeholder inclusion (Cradock-Henry et al., 2018) to improve scenario developments (Kok et al., 2007), support local decisionmaking (Cradock-Henry et al., 2018;Workman et al., 2020), and assess the feasibility of mitigation pathway solutions (Anderson and Jewell, 2019;Weber et al., 2018).Others warned that including a broader diversity of government and non-state actor viewpoints might compromise scenario credibility (Beck and Mahony, 2017), recommending improved systematic processes and formalized methods for stakeholder engagement (Carlsen et al., 2017).

Methods applied
Since 2000, qualitative scenario aspects have been expressed in narrative form (IPCC, 2000a;Schweizer and Kriegler, 2012), aiming to ensure scenario logic and internal consistency (Nakicenovic and Swart, 2000).SRES authors criticized the initial SRES approach, "story and simulation" (SAS), as being limited (Alcamo, 2008).Coupling a storyline to a quantitative simulation (SAS method) does not sufficiently check for internal consistency (Kemp-Benedict, 2012;Schweizer and Kriegler, 2012).Furthermore, the contemporary current global pathway SRES-A1FI ('coal-powered growth') was argued to be under-represented.Instead, consistent and robust scenarios with this theme could be identified via the new CIB method (Cross-impact Balance) (Kemp-Benedict, 2012;Schweizer and Kriegler, 2012).The CIB was used for SSP developments.It identified internal inconsistency in SRES storylines (Schweizer and Kriegler, 2012) and found internally consistent combinations in all five SSP challenge space domains.However, 85% of combinations lay along the diagonal for Low, Medium, or High mitigation-adaptation-challenges (SSP5-SSP2-SSP1), with most of these in Medium and High domains (Schweizer and O'Neill, 2014).More recently, an advanced 'linked CIB' technique enables the analysis of large CIB matrices and ensures internally consistent linking of scenario elements across scales and matrices (Schweizer and Kurniawan, 2016).
In parallel, modelers proposed a backward approach to support SSPstoryline developments, focusing on the most relevant emission drivers to distinguish between, e.g., equity and convergence scenarios (Rozenberg et al., 2014) and systematically identify scenario groups with similar outcomes (Guivarch et al., 2016).Additionally, a method for transparent scenario selection, revealing vulnerabilities of proposed policies and considering scenario diversity, was introduced (Carlsen et al., 2016).A collection of papers proposed to derive policy-relevant insights from scenario developments.They aimed to identify novel research questions, examine how scenarios reflect equity (O'Neill and Nakicenovic, 2008), and how scenarios are used in scientific fields to provide a common framework for coordinating studies across research communities (O'Neill and Nakicenovic, 2008).It was further examined via the Scenarios Forum Conference (O'Neill et al., 2019) and elaborated by scenario developers (O'Neill et al., 2020).
Lack of saliency across scenario series as regards the absence of intervention scenarios, storyline names, and labeling (Girod et al., 2009), was addressed by IPCC authors (van Vuuren et al., 2012) and more clearly labeled in the SSPs (Riahi et al., 2017).IPCC increased attention on assumptions and model approaches during the AR6 preparations (IPCC, 2017a).It published an SR1.5 database (IPCC, 2017b) without completely solving the IAM reproducibility and transparency challenges (Robertson, 2021).RCP/SSP authors provided more transparent descriptions of IAM assumptions on model structures, energy sectors, and bioenergy conversion chains (Bauer et al., 2020).

Higher resolution for impact assessments
National detail is essential for policymakers (Pedersen et al., 2022), mitigation, and adaptation assessments (Kok et al., 2007).In 2002, a small number of scenario assessments called for higher resolution, down-scaling scenarios for regional climate impact assessments (Arnell et al., 2004;Gaffin et al., 2004) associated with the objectives of WG2.The SRES team refrained from downscaling because meaningful topdown downscaling is very difficult, and higher precision levels would misrepresent associated uncertainties.Researchers later requested finegrained climate data, incorporating geographic variation (Nordås and Gleditsch, 2007).The initial critiques led to a high-resolution database (i.e., population and GDP) developed by IPCC authors but independently of IPCC (CIESIN, 2002;Gaffin et al., 2004).National projections were prepared for the SSPs (Dellink et al., 2017;KC and Lutz, 2017).
Since scenarios mainly address the global scale (Zurek 2007), SSPs' ability to support national and local scale decisionmaking remains untested (Cradock-Henry et al., 2018).It is not always appropriate to tightly connect scenarios across scales (Biggs et al., 2007) since the global scale may alienate stakeholders at various administrative scales (Biggs et al., 2007;Kok et al., 2007).Also, the development of participatory scenarios at multiple scales (e.g., time scale, geographic scale) has a strong potential to contribute to decision making and coping with the existing tradeoff between maintaining relevance to stakeholders at different scales and maintaining consistency across scales (Kok et al., 2007).The global SSPs were prepared as a platform for developing extended SSPs substantive elaborations for specific sectors and regions, aiming to improve their usefulness for IAV studies (van Ruijven et al., 2013).Modelers encourage community consensus on methods for working with SSPs across scales (O'Neill et al., 2020).Furthermore, several IAMs are now open-source (e.g., MESSAGE, GCAM, and REMIND), and model description papers are available (Harmsen et al., 2021).

Policy relevance and implications
Policy relevance and implications represented crosscutting critique topics related to assessments of several assumptions, emission range, and process critiques.

Policy relevance
The earliest known scenario critique argued for extending emissions projections beyond 2100 to improve decisionmaking (Cline, 1991), which was included in the SSP-RCPs twenty-five years later (IPCC, 2007a).More recently, researchers argue that the translation of scenarios and scientific evidence into effective decisionmaking has been ineffective (Geden, 2016;Kok et al., 2007;Wilkinson and Eidinow, 2008).The model literature does not explain how researchers could more efficiently contribute to public discourses (Edenhofer and Kowarsch, 2015).On the one hand, scenarios need to be less complex and communicated in a simple manner (Pedersen et al., 2022;Schenk and Lensink, 2007).On the other hand, to ensure robust decisionmaking (Workman et al., 2020), they need regular updates (Garnaut et al., 2008;Peters et al., 2013), examining further the diverse regional emission growth (Anderson and Bows, 2011;Pedersen et al., 2020), including state and non-state viewpoints (Weber et al., 2018;Workman et al., 2020), identifying local policy interventions (David Tàbara et al., 2018;Pedersen et al., 2022), and including well-known mitigation benefits (not included in AR5) (Rosen and Guenther, 2016).According to SSP modelers, including the Paris goals and actual policies and their implications might improve low emission pathways (O'Neill et al., 2020).
During the past decade, the fourth generation IAMs were argued to be black boxes, unfit for policymaking, culturally biased, and comprising unresolved uncertainties (Ellenbeck and Lilliestam, 2019;Low and Schäfer, 2020;Workman et al., 2020).The NETs (and SRM) critiques also stretch beyond the IAMs, questioning IPCC neutrality (Anderson and Peters, 2016;Geden, 2016;Hansson et al., 2021;Low and Schäfer, 2020) and a need to inform policymakers (Fuss et al., 2014).The high policymaker demand for mitigation scenarios implies risks that models end up saying what policymakers want to hear (Anderson, 2015;Geden, 2016), presenting assumptions (Anderson and Jewell, 2019;Anderson and Peters, 2016) that differ from the actual policy actions (Rayner, 2016).Therefore, policy-driven researchers and advisors, including scenario developments, should critically evaluate how their work is interpreted and used in policymaking processes (Geden, 2016) to adequately inform policy (Beck andMahony, 2018a, 2018b).This also included implications regarding IAMs as boundary objects (Beck and Mahony, 2017;Hansson et al., 2021;Low and Schäfer, 2020).Public media replicated the critiques that IAMs contain unhealthy unproven doses of wishful thinking (Edwards, 2020b;Kruger et al., 2016).At the same time, the media also replicated scientific critiques of the scientific overuse of high-emissions pathways, which may mislead policy (Hausfather and Peters, 2020;Pielke and Ritchie, 2020).Additionally, that policymakers tend to focus on extreme scenarios (Höök, 2011).SSP developers announced a need for an increased focus on simplified communication (e.g., infographics and simpler IAMs) and better accessibility via developing an informative and user-friendly online database developed via stakeholder inclusion (O'Neill et al., 2020).

The role of scenarios
Since 2000, natural scientists have argued a need to include probability-based scenario designs (Allen et al., 2000;Schneider, 2001).Scenario developers defended using the explorative storyline approach (Grübler and Nakicenovic, 2001).The critics stated that the SRES does not sufficiently support decisionmaking, since policy analysts need probability estimates to assess the seriousness of the plausible climate impacts (Morgan and Keith, 2008;Schneider, 2001).Scholars argued that error bands and indications of likelihood might support decisionmaking (Schenk and Lensink, 2007;Schneider, 2001), simplify communication (Schenk and Lensink, 2007), and include an analyst's judgment about the probability of various futures (Morgan and Keith, 2008;Schneider, 2001).
SRES developers argued that natural scientific probability estimates might interfere with the scenario logic and the complex interconnection between emission drivers (Grübler and Nakicenovic, 2001).From a social science perspective, emission scenarios could not be represented by probabilities (Hulme 2004) because future emissions and aerosols fall into the category of "unknowable" knowledge, which depends on subjective judgments of unpredictable socioeconomic developments (Hulme 2000).To identify the most critical parameters (Webster et al., 2002), researchers explored probabilistic uncertainty in key drivers, such as population (Lutz et al., 2001) and technology (Gritsevskyi and Nakićenovi, 2000).Additionally, focusing on the output (radiative forcing) than on the input (emissions) may provide coverage of ranges and improve the probabilistic scenario design (IPCC, 2005c;Webster et al., 2002).At RCP/SSP expert meetings, developers discussed probability distributions and policymaker information.Probability was perceived as a subjective choice, potentially making policy choices expressed in probabilistic terms and probability assessment across storylines incorrect (IPCC, 2005b).AR4 compared what-if, probabilistic, and best-guess scenarios (IPCC, 2007b), while AR5 comprised results from 31 models and 1184 scenarios (IPCC, 2014c).Others, including SSP authors, found differences in long-term emission probabilities between expert estimates, which might result from factors like subjective assessments and model inability to foresee long-term disruptive changes (Ho et al., 2019).Researchers recently suggested qualitatively identifying the most likely (best-guess) scenarios based on current trends (Hausfather and Peters, 2020).
Besides a natural-social science opposition, the debate revealed disagreements between the climate and impact assessment communities.The first argued that probability analysis would support mitigation decisionmaking (Allen, 2003;Schneider, 2001;Webster et al., 2002) and more simple scenario communication (Hausfather and Peters, 2020;Pielke and Ritchie, 2020).On the contrary, the latter argued that robust adaptation policy solutions must be based on a wide range of plausible scenarios rather than best-guess (Lawrence et al., 2020;Lempert and Schlesinger, 2001).

Discussion
The review aimed to neutrally describe the criticism and how the scenario authors have addressed the criticism at the time.A neutral critique approach provides insights into the connection between critiques and responses and thus the scenarios' foundation and evolution.

Scenario changes
The review shows that scenario substance (assumptions and quantifications) and methodologies have changed over time.In the beginning, via intergovernmental arguments (IPCC, 1991).Later, changes occurred via scientific and IPCC evaluations (Alcamo et al., 1995;Parikh, 1992) guiding intergovernmental mandates (IPCC, 1996).Since 2000, the scenarios have evolved primarily via scientific critiques and assessments (IPCC, 2007a;O'Neill et al., 2020).Because of the nature of the IPCC, the IPCC panel agreed that experts should publish critique responses in peer-reviewed journals (IPCC, 2003).In addition, some key debates were addressed in IPCC sessions, expert meetings, and ARs, like economic metrics (IPCC, 2007b) and probability assessments (IPCC, 2014c).Seemingly the post-SRES scenarios were less visible as these SRES scenarios hardly attracted critique.
IPCC intergovernmental discussions affected scenario exercises at least three times and once raised the emission range's upper end (i.e., changing conditions for climate and response strategy assessments).We found no evidence that critiques significantly altered overall emissions ranges after 1992, although this is subject to a recurrent debate till today.
The critical letters sent to IPCC (Castles and Henderson, 2003a) addressed methodology and assumption critiques.It led to several scientific evaluations and changed the economic metric without significantly changing the non-OECD GDP range.Moreover, several missing scenario critiques were welcomed by SSP developers (O'Neill et al., 2020).
Methods have changed over time via assessments from SSP and other modelers.Also, the scenario development team has increased continuously.IPCC processes pushed the inclusion of economists in the second generation (IS92) (IPCC, 1991;Pepper et al., 1992), while critiques pushed the inclusion of non-OECD researchers and economic institutions in the third generation (SRES) (Castles and Henderson, 2003a;IPCC, 1996;Parikh, 1992).For the fourth generation (SSP-RCP), the IPCC panel recommended (IPCC, 2005a) including a wider variety and the number of non-governmental stakeholders, e.g., research communities, scenario user groups, and multilateral organizations (IPCC, 2007a).
The energy technology and fossil supply critiques drew low attention from modelers and did not affect assumptions nor ranges.The IPCC and developers have assessed critiques addressing policy issues, i.e., NETs, probability, and SRM.However, this did not lead to substantial scenario changes other than increasing transparency (e.g., improved databases).Only recently, user relevance and scenario communication have been explicitly expressed by modelers (O'Neill et al., 2020).

Imaginative capacity
The results demonstrate that substantial shortfalls in knowledge limit our understanding of the future.The future is explored partly on historical experiences, records, and trends and partly on our imaginative capacity.Several critiques advocate continuously exploring new possibilities within a series' chosen scenarios to remain science-and policyrelevant.Other critiques advocate being cautious and not too speculative.Some critiques implicitly targeted the (unrealistic) imaginative capacity of developers, like too optimistic regional GDP (Castles and Henderson, 2003a), global technology developments (Pielke et al., 2008), and some NETs critiques.However, historical non-OECD GDP and non-biomass renewable energy were within SRES ranges (Pedersen et al., 2021), and technology developments have been more rapid than expected (Creutzig et al., 2017).Despite this, such critiques play a role in 1) continuously challenging the modelers' perceptions, which shape assumptions, and 2) informing scenario users about plausible shortfalls.History will show how NETs will evolve, offering a plausible pathway toward reaching the Paris goals and informing about plausible mitigation tools.There are no indications that NETs assumptions will be excluded (Tanaka and O'Neill, 2018;Van Vuuren et al., 2017).Other critiques introduce alternative mitigation pathways and advocate increasing imaginative flexibility, e.g., degrowth and discontinuity scenarios (Otero et al., 2020;Raskin and Swart, 2020).The responses emphasize that not all scenarios in a series are realizable.Simultaneously, the critiques hint that scenario tools may inspire policy strategies via a wide band of plausible tools.

Transparency and communication
The critiques reveal a need to improve scenario communication and transparency to serve scenario uses in research and policymaking.Low transparency has led to critique.Already in 1984, energy models were accused of being hardwired, reaching specific outputs (Keepin and Wynne, 1984).Similarly, NETs critiques declared that models unintentionally risk masking model inconsistency (Anderson and Peters, 2016;Fuss et al., 2014) and that simpler tools may be more relevant for policymaking (Ellenbeck and Lilliestam, 2019;Pindyck, 2017).To facilitate a 'correct' use of scenarios, modelers propose improving scenario results via new approaches, like infographics, cartoons, and simplified illustrations of system dynamics and IAMs (O'Neill et al., 2020).Here the following could be emphasized: More simple accessibility and overview of input and output data (transparency) and simple communication of the relationships between assumptions, drivers, and future developments (ensuring that users understand and use scenarios 'correctly').Policy relevance and actionability may increase by highlighting policy tools and plausible implications.As an add-on, modelers could consider specific communication of assumed policy roadmaps with timetables of needed technology funding and implementation (to support monitoring policy actions and delays).

Conclusion
The review shows that scenario assumptions, quantifications, and methods have changed over time, inspired by political considerations and scientific critiques.
The subsequent scenario generations used in IPCC assessments have passed the test of criticism over time.Many critiques have scrutinized the scenarios, led to scenario improvements and enhanced their credibility.From a scientific perspective, the credibility may have been compromised because of excluding mitigation scenarios in IS92 and SRES.However, from a political perspective, this reduced scope was necessary to have the scenarios also accepted for consistent use in IPCC by countries that still questioned the need for mitigation.Later the mitigation need was globally accepted.As the RCP/SSP developments moved outside the IPCC, the scenarios' scope expanded to include mitigation as a component of sustainable futures.
Critiques can be grouped into various primary and secondary focus topics, revealing that half of the critiques addressed assumptions.In total, we identified 280 emission scenario critiques.They can be grouped into four main categories emerging from the literature: assumptions and scenario ranges (substance), and methodology and user relevance/policy issues (process).Some of the critical themes in the critiques, MER-PPP (2003-2007) and the IPCC in-crowd (1998-2013), have been intense during specific periods but seem to have disappeared, while probability/best-guess have decreased in intensity.Scenario improvements took away some critique topics, like narratives including explicit income convergence and changed economic metrics.Improved development processes, such as increased author teams and stakeholder inclusion, took away several process critiques (while the IPCC critical literature continued in the grey literature).The probability critiques evolved during the transition period between the second and third generations (IS92 and SRES) and faded after 2013.However, critiques recently advocated adding qualitative likelihoods or best guesses to the SSP-RCP framework.
Some themes have continued to be relevant.The most prominent examples are assumptions, emission ranges (since IS92), resolution, and applied methods (since SRES).Although experts and stakeholders have increasingly been included in scenario developments, stakeholder inclusion in scenario preparations and local extensions continues to be addressed in the literature addressing resolution/local extensions and non-government mitigation actions.
Policy implications and transparency critiques have emerged more recently.These critiques were also addressed as secondary topics in NETs critiques.Furthermore, has missing scenario critiques (adding new aspects to the narratives) become more frequent.These critique topics might continue to be relevant in the future.
The scenario critiques do emphasize the importance of communication and transparency.Although probability critiques did not significantly change the scenarios, they advocated for more uncomplicated scenario communication, which developers recently considered.Scenarios have grown more complex over time; thus, it may be valuable to include user perspectives (e.g., policymakers, sectorial stakeholders) to develop effective scenario communication in the future.
Not only scenarios include subjective choices.Also, the assessed critiques have (implicit and explicit) politically motivated aims, such as convergence assessments, critiques questioning the IPCC status, if policy regulation is needed, and missing Solar Radiation Management.Others were more neutral, contributing to later scenario developments, e.g., probability critiques focusing on outputs (radiative forcing) or effective communicating scenarios.To further improve the knowledge of IPCC assessments' effectiveness and the role of emissions scenarios, more research would be required into the sources of sponsorship of critiques and the grey literature.

Fig. 2 .
Fig. 2. Historical overview of United Nations Framework Convention on Climate Change (UNFCCC) history and key processes, Intergovernmental Panel on Climate Change establishment and assessment reports (IPCC ARs), and the four generations of emission scenario series and their inclusion in scenario-based literature informing IPCC ARs three Working Groups.Data sources: IPCC ARs 1990-2022.

Fig. 4 .
Fig. 4. Historical development and weight of scenario critique 1990-2022: Peer-reviewed (green) and grey literature (grey) critiques of emissions scenarios in the IPCC context.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5 .
Fig. 5. Distribution and development of critique topics.a The number of peer-reviewed scenario critiques by primary topic (intense colors) and secondary topics (light colors) analysis topics.Primary topics are given a score of 1. Secondary topics are given a score of 0.25, 0.5, or 0.75, equal to their weight and relevance in the paper.One primary topic was identified per paper, while it may address several secondary topics.b Primary topics & weighted secondary topics grouped by publication year.Based on 280 peer-reviewed articles published between 1990 and 2021 assessing SA90, IS92, SRES, SSP-RCPs, or SR1.5 emission scenario series (See SI excel, Sheet 1).

Table 1
Four main categories (and 13 subcategories) of emission scenario critique topics.
-Too little spatial resolution (energy systems, land-use, etc.) 4. Scenario relevance 4.1 User/Policy implications -Not scientific; unreliable to guide policy 4.2 User/Policy relevance -Aspects needed to increase policy-relevance -Scenarios are not addressing the right questions 4.3 Role of scenarios (scenario type) -Explorative (storyline/quantification) vs. probabilistic approaches (frequency distributions) vs. Qualitative best-guess scenarios