Atmospheric Physics Building,Department of Physics, University of Oxford, Parks Road, Oxford, OX1 3PU
Professor Erik Kolstad, NORCE and the Bjerknes Centre for Climate Research
Andrea Simpson (andrea.simpson@physics.ox.ac.uk)
Abstract
Weather and climate forecasts are fundamental tools for managing risk—but turning model output into useful information is rarely straightforward. In this talk, I will reflect on my journey from working as a disciplinary atmospheric scientist to leading interdisciplinary collaborations aimed at bridging the gap between forecasts and real-world decision-making.
Drawing on projects such as Climate Futures in Norway and ACACIA and CATER in Africa, I will share examples of how collaboration across disciplines—meteorology, social science, policy, and practice—has enabled new kinds of climate services, tailored to different regional contexts and time horizons. This includes probabilistic 21-day forecasts used in agriculture and disaster risk planning, as well as public–private partnerships that integrate sub-seasonal predictions into operational decision-making.
I will also present key findings from two recent foundational studies that emerged from these collaborations. The first (Keane et al., accepted) demonstrates an asymmetry in forecast skill: mid-latitude systems are more predictable at short lead times, while tropical systems offer more skill at longer lead times. Surprisingly (to us at least), this pattern holds even for machine learning models, pointing to intrinsic properties of the atmosphere. The second paper (Dunn-Sigouin et al., accepted) shows how spatial aggregation can increase forecast skill and extend lead time, especially for sub-seasonal extremes. I will illustrate this with Storm Hans, the most damaging weather disaster in recent Norwegian history, where spatially aggregated rainfall forecasts could have improved early warnings.
These results highlight not just scientific insights, but also the value of collaborative approaches that embrace uncertainty, span disciplines, and engage with users. They suggest a growing need for climate scientists to navigate both the technical and social dimensions of their work—particularly in an era of rising climate risk and uneven global access to forecast information.