A weighting framework to improve the use of emissions scenario ensembles of opportunity

nature climate change Article https://doi.org/10.1038/s41558-026-02565-5 A weighting framework to improve the use of emissions scenario ensembles of opportunity Received: 10 December 2024 Accepted: 15 January 2026 Published online: xx xx xxxx Check for updates Hamish Beath 1 , Chris Smith2, Jarmo S. Kikstra Matthew J. Gidden 2,6 & Joeri Rogelj 1,2,3 , Mark M. Dekker 1,2,3 , 4,5 Integrated assessment models produce large ensembles of socioeconomic scenarios that are used profusely in climate change research. The Intergovernmental Panel on Climate Change (IPCC), non-governmental organizations or national climate committees often rely on ensemble statistics to identify mitigation strategies and set climate targets. A limitation of such evidence is the opportunistic nature of scenario ensembles: they are an unstructured, serendipitous collection of evidence. Drawing on concepts from physical climate science and ensemble analysis, we present an approach for the flexible, multidimensional weighting of emission scenario data that accounts for relevance, quality and diversity. Our illustrative application to the latest IPCC scenario database demonstrates a reduction in dominance of highly represented models and studies, and sees net-zero emission milestones differ to those originally reported. Our framework formalizes decisions otherwise made in an ad hoc manner, providing a tool contributing to the broader challenge of assessing ensembles of opportunity. Since the publication of the Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios1 in 2000, integrated assessment models (IAM) have become central tools for exploring emissions and climate futures in climate research. Despite already being part of the IPCC Third2 and Fourth3 Assessments in 2001 and 2007, it was the IPCC Fifth Assessment4, the Special Report on Global Warming of 1.5 °C (ref. 5) (SR1.5) and the IPCC Sixth Assessment6 (AR6) that solidified the use of large scenario ensembles for the assessment of global mitigation strategies. With their expanded use also came an improved understanding and communication of their limitations7,8. Scenarios that are collected as part of IPCC or other exercises9–11 represent ensembles of opportunity: a serendipitous collection of scenario data that is unstructured7 and in which the scenarios that are ultimately included vary in their purpose, design, comprehensiveness, coverage, quality and other characteristics. One key limitation of their use is that shortcomings or biases present in the collected ensemble can be propagated by subsequent secondary analysis7,8. Biases include dominance of specific models and intercomparison projects, which can represent a lack of diversity in organizational or regional composition12–14. Unless corrected for, this could lead to spurious or biased results. Typical shortcomings or biases relate to three main issues: scenario relevance, quality and diversity. Relevance refers to whether a scenario, through the structural properties of the underlying model, its design and outcome characteristics, is relevant to the question that is being investigated by the secondary analysis. This can include the Centre for Environmental Policy, Faculty of Natural Sciences, Imperial College, London, UK. 2International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria. 3The Grantham Institute for Climate Change and the Environment, Imperial College London, London, UK. 4PBL Netherlands Environmental Assessment Agency, The Hague, the Netherlands. 5Copernicus Institute for Sustainable Development, Utrecht University, Utrecht, the Netherlands. 6Center for Global Sustainability, University of Maryland, College Park, MD, USA. e-mail: ; 1 Nature Climate Change Article https://doi.org/10.1038/s41558-026-02565-5 Ensemble of opportunity Unstructured, serendipitous collection of emission scenario data Scenarios carry equal weight Relevance weighting Weighted ensemble Diversity weighting Quality weighting Prioritize scenarios Prioritize scenarios that meet relevant to the specific quality new research criteria questions Scenario ensemble with reduced sampling bias among relevant scenarios of acceptable quality Avoid duplicates and prioritize diverse scenarios Scenarios carry different weights Fig. 1 | A scenario-weighting framework for the analysis of scenario ensembles of opportunity. Schematic showing how the unstructured, serendipitous collection of evidence in a scenario ensemble of opportunity can be translated in a weighted ensemble, accounting for the relevance, quality and diversity of a scenario. estimated level of global warming avoided by the scenario15, assessments of the feasibility of its described transitions16 or even subjective— but transparently communicated—preferences about technologies or strategies. Quality refers to whether the implementation and execution of the modelling lives up to predefined standards set out by the secondary analysis. These standards typically refer to technical modelling aspects, such as the accuracy of historical data, time resolution or plausibility of near-term trends and resource use. Diversity refers to the degree of additional information a scenario communicates compared with other members in the ensemble of opportunity. Not accounting for the latter can result in statistics across a scenario ensemble being too narrow or overconfident towards the results of a single model, modelling team or modelling exercise17. In the past, such issues have been dealt with on an ad hoc basis. For example, the IPCC SR1.55 checked for the completeness of variables available in scenarios, whether data are reported until 2100 or whether reported historical- or near-future data are consistent with observations9,18. For the assessment of global emission characteristics of pathways aligned with 1.5 °C, SR1.5 also excluded 13 scenarios from a single modelling group19 because they included virtually no variation in emissions and their inclusion would have biased descriptive emission statistics. Criteria for including or excluding scenarios depend on context, which explains why these 13 scenarios were still included in the analysis of aspects other than the evolution of emissions in the SR1.5 report. Similar ad hoc considerations were applied in the AR6 mitigation assessment of the IPCC6,8. Other climate research communities have also grappled with similar issues. The Earth System Modelling community has established methods to down-weight models based on their similarity to other models20–22, as well as for model quality measured as performance relative to historical observations22,23. In emission scenario ensemble analysis, issues of scenario similarity have been considered24, but as of yet not systematically addressed. Here we use efforts from various communities20–24 as a starting point to develop and present a scenario-weighting method, applicable to scenarios from IAMs or e (...truncated)