Does higher-resolution climate forcing improve impact model performance?

Journal

Environmental Research Letters

Authors

Johanna Malle, Christopher P O Reyer, Yael Amitai, Andrey Lessa Derci Augustynczik, Yaron Be’eri-Shlevin, Elad Ben-Zur, Peter Burek, Tarunsinh Jayvirsinh Chaudhari, Jinfeng Chang, Alessio Collalti, Daniela Dalmonech, Shouro Dasgupta*, Iulii Didovets, Marc Djahangard, Laura Dobor, Louis Francois, Simon N Gosling, Fred Hattermann, Shaochun Huang, Heike Lischke, Thomas Lorimer, Katarína Merganičová, Francesco Minunno, Mats Nieberg, Elizabeth J. Z. Robinson*, Martin Schmid, Mikhail M Smilovic, Ritika Srinet, Elia Vangi, Xue Yang, Rasoul Yousefpour, Isabel Ayala, Daniel Mercado-Bettin, Dánnell Quesada-Chacón and Dirk Karger

*The starred authors are contributors from the Grantham Research Institute

Abstract

Climate-impact assessments increasingly use high-resolution climate forcing, often assuming that finer spatial detail improves impact projections. Yet this assumption has rarely been tested systematically across impact sectors, regions, and forcing variables. The authors of this paper use coordinated simulations from 16 impact models participating in the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) to evaluate how increasing forcing resolution from 60 km to 10 km, 3 km, and 1 km affects both climate-input accuracy and impact-model performance. The authors find that most gains occur when moving from 60 km to 10 km resolution. Further refinement to 3 km and 1 km yields smaller and less consistent benefits, and in some cases degrades performance. Improvements are strongest for temperature-sensitive models and in topographically complex regions, where finer forcing better captures spatial gradients. By contrast, precipitation-driven models and low-relief regions show weaker and less systematic gains. These results challenge the expectation that ever-finer climate forcing necessarily improves impact modelling. Instead, the value of resolution depends on the forcing variable, sector, region, and process representation of the impact model. Benchmarking resolution choices against impact-model performance, rather than climate-input accuracy alone, is therefore essential for designing efficient and decision-relevant climate-impact assessments.