State of the art research advances in climate and weather extremes, ranging from flash floods to heatwaves and droughts, were presented at CMIP Community Workshop 2026 in Kyoto. Session SCI3.2 brought together researchers from around the world to tackle one of the most pressing questions in climate science: how do we better assess, model, and project future changes in high-impact extremes?
Understanding and anticipating high-impact events remains a key focus in climate research. Extreme events lead to enormous tolls on human lives, societies, and ecosystems, and these impacts are projected become even larger under further global warming, as these events become stronger and strike more often. Yet understanding them is complicated by the fact that extremes, especially high-impact, worst-case or unprecedented extremes are driven by complex, and rare, combinations of factors. Through teleconnections and causal chains of atmospheric or oceanic interactions, extreme events can be linked to mechanisms triggered thousands of kilometres away. Untangling, and adequately modelling these remote influences from local drivers is one of the central scientific challenges the session addressed.
Single Model Large Ensembles: key tools to extreme event research
A significant thread running through the session was the use of Single Model Large Ensembles (SMLEs), large collections of simulations from a single model, each starting from slightly different initial conditions. By generating many realisations of the same model, SMLEs increase the chances of capturing rare extreme events in the model ensemble, thus improving the sampling of extreme events, and providing a richer diverse pool of events and background conditions to disentangling necessary from circumstantial mechanisms that may influence their development, and how they may change in a warming world.
The studies presented in the session explored aspects ranging from using single SMLEs to study specific extremes and their driving mechanisms in depth, to disentangling the forced responses, and in the mean climate response to external forcings such as greenhouse gases, from internal variability, the natural, chaotic fluctuations of the climate system. Quantifying the role of both factors matters enormously for understanding extreme conditions, as they occur under the backdrop of changing mean climate conditions and as the result of rare combinations of drivers, which can be affected by the internal chaotic variability in the climate system as well as experience forced changes themselves under warming. For example, the session explored whether future changes in ENSO teleconnections are robust to internal variability, directly addressing the question of how confidently can we project changes in one of the primary drivers of global climate variability in a changing climate.
Beyond single-model SMILE studies, the session also explored whether multi-model large ensembles can reveal things that single-model approaches obscure, a timely reminder that no single methodological choice is without trade-offs. Furthermore the session also explored how SMLEs can be used for model calibration, offering a more principled way to separate forced signals from internal noise.
Climate and weather extremes across the Globe
The session covered a remarkable geographic and thematic range, reflecting the global reach of the problem. Topics ranged from how mesoscale convective systems and desert dust interact to shape rainfall intensity in West Africa, to how recent extreme marine heatwaves around Japan interacted with atmospheric heatwaves, revealing coupled ocean-atmosphere dynamics driving compound extremes. A study investigating projected changes in atmospheric pathways driving western North American heatwaves underscored that the dynamics of extreme heat events themselves may shift under climate change, not just their frequency. Furthermore, the session also highlighted that studying regions in isolation is often insufficient to produce a comprehensive mechanistic understanding of extreme conditions. A striking example of non-local drivers presented evidence that Arctic sea ice decline is triggering heavier summer monsoon rains over South Asia, showcasing a pole-to-tropics connection with profound implications for one of the world’s most critical rainfall systems that billions of people depend on.
Lastly, the session explored potential worst-case European heatwave storylines generated using Ensemble Boosting. This technique is based on systematically re-running selected conditions in existing SMILE simulations to produce large samples of comparable trajectories emerge as new options to build upon the strength on currently modelling tools such as SMILEs, yet remain underexplored and applied to a limited number of climate models and event types, highlighting the potential for the international collaboration and intercomparison that the CMIP community excels at.
The challenge ahead
The breadth and depth of the science on display in this session underscored both the progress the community has made and the significant work that remains. Attribution systems moving toward operational use and regional studied painting increasingly detailed pictures of how extremes are affected by local and non-local drivers, and how they may change in a warming world are clear examples of the community’s successes in understanding high-impact extremes. SMLEs are solidified as central modelling tool for extreme event research, and while new techniques emerge to further amplify their strengths, they remain underexploited, highlighting exciting research and model development avenues. At the same time, the session highlighted the importance of honest confrontation with uncertainty, a robust assessment of what models can and cannot yet capture and how to communicate this information remains at the heart of high-impact extreme event research.
