Brenner et al. (2026) GeoDS (v.1.0): a simple Geographical DownScaling model for long-term precipitation data over complex terrains

Identification

• Journal: Geoscientific model development
• Year: 2026
• Date: 2026-02-03
• Authors: Jean-Baptiste Brenner, Aurelien Quiquet, Didier M. Roche, Didier Paillard, Pradeebane Vaittinada Ayar
• DOI: 10.5194/gmd-19-1075-2026

Research Groups

• Laboratoire des Sciences du Climat et de l’Environnement, LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, France
• Earth and Climate Cluster, Faculty of Earth and Life Sciences, Vrije Universiteit Amsterdam, Amsterdam, the Netherlands

Short Summary

This paper introduces GeoDS (v.1.0), a simple, topography-based geographical downscaling model for long-term precipitation data over complex terrains, designed for paleoclimate studies. It demonstrates GeoDS's ability to capture fine-scale precipitation patterns and improve statistical agreement with observations in the European Alps and Greenland, while being computationally inexpensive and robust.

Objective

• To develop and evaluate a simple, computationally inexpensive, and robust topography-based model (GeoDS) for downscaling long-term precipitation fields in complex terrains, particularly adapted for multi-millennia paleoclimate simulations, by physically representing climate-relief interactions.

Study Configuration

• Spatial Scale: European Alpine region (4.8–17.5° E / 43–49° N) and Greenland. Input coarse resolution: 50 km (degraded APGD, ERA-5) or spatially homogeneous. Output high resolution: 5 km (Alps), 15 km (Greenland). Digital Elevation Model (DEM) resolution: 1 km, 5 km, 10 km.
• Temporal Scale: Calibration and testing period: 1971–2019 (Alps), 1990–2020 (Greenland). Primary temporal resolution: monthly. Sensitivity tests included daily resolution. Designed for long-term (millennial or longer) paleoclimate simulations.

Methodology and Data

• Models used:
  • GeoDS (v.1.0): Geographical DownScaling model (developed in this study).
  • MAR (Modèle Atmosphérique Régional): Regional climate model (used as target for Greenland).
• Data sources:
  • Alpine Precipitation Grid Dataset (APGD): Target high-resolution precipitation data for the Alps (5 km grid, 1971–2019).
  • ERA-5: Eastward (u) and northward (v) wind components at various pressure levels (10 m, 950 hPa, 900 hPa, 850 hPa, 800 hPa, 700 hPa, 500 hPa) (31 km grid, 1971–2019).
  • Shuttle Radar Topography Mission (SRTM) dataset: Digital Elevation Model (30 arc-second native, aggregated to 1 km, 5 km, 10 km).
  • Degraded APGDm: Coarse resolution precipitation input (50 km grid) derived from APGDm for consistency.

Main Results

• GeoDS significantly improves the representation of fine-scale precipitation patterns in complex terrain compared to bilinearly interpolated coarse data.
• For the Alps (1971–2019, 10 km DEM), GeoDS reduced the Mean Absolute Error over quantiles (MAEquantiles) from 14.68 mm to 1.73 mm per month and increased the coefficient of determination (R² quantiles) from 0.978 to 0.999.
• The model effectively reproduces key topographic features such as rain-shadow effects in deep valleys (e.g., Grenoble, Sion) and enhanced precipitation on windward slopes.
• GeoDS performs better during summer than winter in the Alps, attributed to the nature of precipitation and moisture advection.
• The model demonstrates robustness when applied to a different climate (Greenland) and topography, reducing MAEquantiles from 8.5 mm to 2.59 mm per year (using Alps-tuned parameters) and increasing R² quantiles from 0.995 to 0.999.
• Increasing the temporal resolution of input data (e.g., daily instead of monthly) can significantly improve performance in regions affected by short, intense rainfall events.
• GeoDS is computationally inexpensive, with simulations taking less than 20 seconds on an Intel Core i7-1365U processor.

Contributions

• Development of GeoDS (v.1.0), a novel, simple, and computationally inexpensive topography-based downscaling model specifically designed for long-term (multi-millennia) paleoclimate simulations.
• Demonstration of the model's robustness and physical basis, making it suitable for applications outside its calibration framework and in contexts of changing climate and surface features.
• Provision of a flexible, open-source, and well-documented downscaling tool for the climate community.
• Addressing limitations of existing statistical downscaling methods (stationarity issues) and dynamical downscaling (high computational costs) for paleoclimate studies.

Funding

• Research was conducted at the Laboratoire des Sciences du Climat et de l’Environnement, LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, France. No specific project or grant codes were listed.

Citation

@article{Brenner2026GeoDS,
  author = {Brenner, Jean-Baptiste and Quiquet, Aurelien and Roche, Didier M. and Paillard, Didier and Ayar, Pradeebane Vaittinada},
  title = {GeoDS (v.1.0): a simple Geographical DownScaling model for long-term precipitation data over complex terrains},
  journal = {Geoscientific model development},
  year = {2026},
  doi = {10.5194/gmd-19-1075-2026},
  url = {https://doi.org/10.5194/gmd-19-1075-2026}
}