When and where higher-resolution climate data improve impact model performance
Johanna T. Malle 1,2,⋆ , Christopher P.O. Reyer 3 , Yael Amitai 4 , Andrey L. D. Augustynczik 5 , Yaron Be’eri-Shlevin 4 , Elad Ben-Zur 6 , Peter Burek 5 , Tarunsinh Chaudhari 7 , Jinfeng Chang 8,29 , Alessio Collalti 9 , Daniela Dalmonech 9 , Shouro Dasgupta 10,11 , Iulii Didovets 3 , Marc Djahanga · 2025
Read When and where higher-resolution climate data improve impact model performance by Johanna T. Malle 1,2,⋆ , Christopher P.O. Reyer 3 , Yael Amitai 4 , Andrey L. D. Augustynczik 5 , Yaron Be’eri-Shlevin 4 , Elad Ben-Zur 6 , Peter Burek 5 , Tarunsinh Chaudhari 7 , Jinfeng Chang 8,29 , Alessio Collalti 9 , Daniela Dalmonech 9 , Shouro Dasgupta 10,11 , Iulii Didovets 3 , Marc Djahanga (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-17739.txt.
Introduction
Johanna T Malle and a large team ask a money question for climate science: how sharp must climate data be before sharper stops helping? Impact models predict floods, crop losses, and heat harm from climate inputs, and finer grids cost dearly to compute. The team ran shared impact models across many sectors at grid sizes from 60 kilometers down to 1.
Most of the gain arrived in the first step, from 60 down to 10 kilometers. Finer grids helped only modestly, and unevenly. Temperature-driven impacts and mountain regions gained the most; rain-driven models on flat land sometimes got worse past 10 kilometers. The advice for data providers is blunt: fund the resolution jump that pays, and fix model physics before buying pixels.
Catalog record
| CREATOR | Johanna T. Malle 1,2,⋆ , Christopher P.O. Reyer 3 , Yael Amitai 4 , Andrey L. D. Augustynczik 5 , Yaron Be’eri-Shlevin 4 , Elad Ben-Zur 6 , Peter Burek 5 , Tarunsinh Chaudhari 7 , Jinfeng Chang 8,29 , Alessio Collalti 9 , Daniela Dalmonech 9 , Shouro Dasgupta 10,11 , Iulii Didovets 3 , Marc Djahanga |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-17739
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
When and where higher-resolution climate data improve impact model performance Johanna T. Malle 1,2,⋆ , Christopher P.O. Reyer 3 , Yael Amitai 4 , Andrey L. D. Augustynczik 5 , Yaron Be’eri-Shlevin 4 , Elad Ben-Zur 6 , Peter Burek 5 , Tarunsinh Chaudhari 7 , Jinfeng Chang 8,29 , Alessio Collalti 9 , Daniela Dalmonech 9 , Shouro Dasgupta 10,11 , Iulii Didovets 3 , Marc Djahangard 12 , Laura Dobor 13 , Louis François 7 , Simon N. Gosling 14 , Fred F. Hattermann 3 , Shaoshun Huang 15 , Heike Lischke 2 , Thomas Lorimer 16,26,27 , Katarina Merganicova 13,18 , Francesco Minunno 17,19 , Mats Nieberg 3,20,21 , Elizabeth J. Z. Robinson 11 , Martin Schmid 16 , Mikhail Smilovic 5,28 , Ritika Srinet 17,19 , Elia Vangi 9 , Xue Yang 22 , Rasoul Yousefpour 12,23 , Ana I. Ayala 25 , Daniel Mercado - Bettin 24 , Dánnell Quesada-Chacón 3 , Dirk N. Karger 2 1 Department of Evolutionary Biology and Environmental Studies, University of Zurich, Zurich, Switzerland 2 Dynamic Macroecology, Land Change Science, Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Birmensdorf, Switzerland 3 Potsdam Institute for Climate Impact Research (PIK), Member of the Leibniz Association, Potsdam, Germany 4 Kinneret Limnological Institute, Israel Oceanographic & Limnological Research, Israel 5 Water Security Research Group, Biodiversity and Natural Resources Program, International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria 6 Institute of Earth Sciences, Hebrew University of Jerusalem, Jerusalem, Israel 7 Department of Astrophysics, Geophysics and Oceanography (AGO), SPHERES Research Unit, University of Liège, Liège, Belgium 8 State Key Laboratory of Soil Pollution Control and Safety, College of Environmental and Resource Sciences, Zhejiang University, Hangzhou 310058, China 9 Forest Modelling Lab., Institute for Agriculture and Forestry Systems in the Mediterranean, National Research Council of Italy (CNR-ISAFOM), Perugia 06128, Italy 10 Centro Euro-Mediterraneo sui Cambiamenti Climatici (CMCC), Venice, Italy 11 Grantham Research Institute on Climate Change and the Environment, London School of Economics and Political Science (LSE), London, UK 12 Faculty of Environment and Natural Resources, University of Freiburg, Freiburg, Germany 13 Faculty of Forestry and Wood Sciences, Czech University of Life Sciences, Prague, Czech Republic 14 School of Geography, University of Nottingham, Nottingham, UK 15 The Norwegian Water Resources and Energy Directorate (NVE), Oslo, Norway 16 Surface Waters – Research and Management, Eawag: Swiss Federal Institute of Aquatic Science and Technology, Kastanienbaum, Switzerland 17 Institute for Atmospheric and Earth System Research (INAR) & Faculty of Agriculture and Forestry, University of Helsinki, Helsinki, Finland 18 Department of Biodiversity of Ecosystems and Landscape, Institute of Landscape Ecology, Slovak Academy of Sciences, Slovakia 19 Forest Modelling, Yucatrote 20 European Forest Institute, Bonn, Germany 21 Chair of Forest Growth and Woody Biomass Production, TU Dresden, Tharandt, Germany 22 State Key Laboratory of Water Engineering Ecology and Environment in Arid Area, Xi’an University of Technology, Xi’an, China 23 John H. Daniels Faculty of Architecture, Landscape, and Design, University of Toronto, Toronto, Canada 24 Spanish National Research Council, Centre for Advanced Studies Blanes, Blanes, Spain 25 Limnology Unit, Department of Ecology and Genetics, Uppsala University, Uppsala, Sweden 26 Laboratory of Molecular and Behavioral Neuroscience, Institute for Neuroscience, Department of Health Sciences and Technology, ETH Zurich, Switzerland 27 ETH Zurich 3R Hub, ETH Zurich, Switzerland 28 Institute of Environmental Engineering, ETH Zurich, 8093 Zürich, Switzerland 29 International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria ⋆ Corresponding author, email: [email protected] Abstract Climate impact assessments increasingly rely on high-resolution climate and forcing datasets, under the premise that finer detail enhances both the accuracy and the policy relevance of projections. Systematic evaluations of when and where higher resolution data improve model outcomes remain limited, and it is still unclear whether increasing spatial resolution consistently enhances climate impact model performance across application areas, regions, and forcing variables. Here we show that improvements in climate input accuracy and impact model performance are most pronounced when moving from coarse (60 km) to intermediate (10 km) resolution, while further refinement to 3 km and 1 km provides more modest and inconsistent benefits. Using the cross-sectoral model simulations from the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP), we demonstrate that higher resolution substantially improves model skill in temperature-sensitive impact models and topographically complex regions, whereas precipitation-driven and low-relief systems show less consistency to increase performance with resolution. For temperature, both climate inputs and model outputs improved most strongly at the 60 km → 10 km transition, with diminishing gains at finer scales. A similar result emerged for precipitation, although some models even exhibited reduced performance when resolution increased beyond 10 km. These results highlight that optimal resolution depends on sectoral and regional context, and point to the need for improving model process representation and downscaling techniques so that added spatial detail can translate into meaningful performance gains. For data providers, this implies prioritizing investments in resolutions that maximize improvements where they matter most, while for modeling groups and users, it underscores the need for explicit benchmarking of resolution choices. More broadly, this work advances the design of consistent, efficient, and policy-relevant multi-sectoral climate impact assessments by clarifying when high-resolution data meaningfully enhance outcomes. 1 Introduction Climate change is progressing rapidly and poses serious risks to both natural ecosystems and human societies ( 44 ; 9 ) . To assess these risks, climate impact models are widely applied across multiple sectors, including agriculture ( 30 ) , forestry ( 59 ; 43 ) , terrestrial biodiversity ( 24 ; 12 ) , hydrology ( 23 ; 74 ; 72 ) , energy ( 29 ; 76 ) , water quality ( 31 ) , labour ( 16 ) and human health ( 66 ; 8 ) . These models are inherently complex, integrating a wide range of biophysical and socio-economic processes. As a result, they carry substantial uncertainty, which is further amplified when driven by inherently complex future climate scenarios. To address and disentangle these uncertainties and enhance the comparability of results, international initiatives such as the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) ( 20 ) provide harmonized protocols and standardized input datasets to facilitate inter-model and cross-sectoral comparisons. This approach is inspired by practices from the climate modeling community, where coordinated model intercomparison projects like CMIP ( 50 ; 51 ) have long used ensembles of simulations to assess uncertainty and model performance ( 19 ) . However, truly interdisciplinary climate impact assessments that integrate models from multiple sectors under common forcing conditions are still in their infancy ( 21 ; 62 ; 26 ) , despite increasing evidence that climate impacts are interconnected and often occur simultaneously across different ecosystems ( 77 ; 60 ; 10 ; 55 ) . An important but often overlooked source of uncertainty in climate impact modeling stems from the discrepancy between the resolution of climate input data and the spatial scale at which impacts occur. Many impact models operate at coarse spatial resolutions (e.g., 0.5° or 50 km at the equator), which can mask fine-scale climatic variability, particularly in heterogeneous or mountainous terrain. This mismatch can significantly affect simulations of components of ecosystem dynamics such as water fluxes ( 71 ) , forest carbon fluxes ( 25 ) , growing seasons