Wood et al. (2025) Comparison of high-resolution climate reanalysis datasets for hydro-climatic impact studies

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-09-08
• Authors: Raul R. Wood, Joren Janzing, Amber van Hamel, Jonas Götte, Dominik L. Schumacher, Manuela I. Brunner
• DOI: 10.5194/hess-29-4153-2025

Research Groups

• WSL Institute for Snow and Avalanche Research SLF, Davos Dorf, Switzerland
• Climate Change, Extremes and Natural Hazards in Alpine Regions Research Center CERC, Davos Dorf, Switzerland
• Institute for Atmospheric and Climate Science, ETH Zurich, Zurich, Switzerland

Short Summary

This study comprehensively evaluates four high-resolution climate reanalysis datasets (ERA5, ERA5-Land, CERRA, CHELSA-v2.1) against gridded observations over complex terrain in Switzerland for hydro-climatic impact studies. It concludes that CERRA generally offers the most reliable representation of precipitation, temperature, and snowfall metrics, including their variability and extreme events, making it highly suitable for a broad range of hydrological analyses, particularly in regions where snow processes and daily to inter-annual precipitation variability are crucial.

Objective

• To conduct a comprehensive spatio-temporal evaluation of four state-of-the-art high-resolution climate reanalysis datasets (ERA5, ERA5-Land, CERRA, and CHELSA-v2.1) for various precipitation, temperature, and snowfall metrics over complex terrain in Switzerland.
• To determine which reanalysis products are most suitable for hydrological impact studies in mountainous regions by assessing their representation of mean and extreme climate metrics, daily to inter-annual variability, long-term trends, and specific extreme events (droughts and heavy precipitation).

Study Configuration

• Spatial Scale: Switzerland, focusing on 97 non-overlapping headwater catchments. Datasets' spatial resolutions: 31 km (ERA5), 9 km (ERA5-Land), 5.5 km (CERRA), 1 km (CHELSA), and 2 km (MeteoSwiss observations).
• Temporal Scale: 1986–2020 (35 years), daily resolution. Analysis includes daily, monthly, seasonal, and inter-annual variability, as well as long-term trends.

Methodology and Data

• Models used:
  • ERA5: Integrated Forecasting System Cy41r2 (ECMWF)
  • ERA5-Land: CHTESSEL (Carbon Hydrology-Tiled ECMWF Scheme for Surface Exchanges over Land) forced by ERA5
  • CERRA: HARMONIE-ALADIN model (regional reanalysis system), SURFEX v8.1 (land-surface model), MESCAN (regional precipitation analysis system with data assimilation)
  • CHELSA-v2.1: Statistical downscaling of W5E5 (bias-adjusted ERA5) using atmospheric lapse-rate downscaling for temperature and a wind-field-based algorithm for precipitation.
• Data sources:
  • Gridded observations (benchmark): MeteoSwiss daily mean temperature (TabsD, ≈2 km resolution, 90 stations) and daily total precipitation (RhiresD, ≈2 km resolution, 650 stations) over Switzerland.
  • Reanalysis datasets: ERA5, ERA5-Land, Copernicus European Regional Reanalysis (CERRA), Climatologies at High resolution for the Earth’s Land Surface Areas (CHELSA-v2.1).
  • Catchment outlines and elevations: CAMELS-CH dataset.
  • Climate metrics calculation: xclim Python package.

Main Results

• Mean Climatology: Most reanalysis datasets overestimate mean daily precipitation (except CERRA, with median biases around 0%), with no clear elevation dependence. Mean daily temperature is generally slightly underestimated or matches observations well, with ERA5 showing a warm bias and others a slight cold bias (<1 °C). All reanalyses exhibit clear cold biases in winter (median ≥ -1 °C), increasing with elevation.
• Extreme Climatology: Biases are most pronounced for extreme precipitation metrics. ERA5, ERA5-Land, and CHELSA underestimate annual maximum 1-day precipitation (Rx1d) and the fraction of total precipitation from very wet days (R99pTot), while overestimating the number of wet days. CERRA shows lower biases across these metrics. Extreme temperature indicator biases are less significant; ERA5 overestimates maximum daily temperature, while CERRA and CHELSA show smaller biases that increase with elevation.
• Solid Precipitation (Snowfall): CERRA and CHELSA best represent the fraction and total amount of solid precipitation, particularly below 1500 m. ERA5 underestimates both, with biases increasing with elevation. ERA5-Land consistently overestimates solid precipitation at all elevations.
• Temporal Variability: CERRA best represents precipitation variability across all timescales (daily, monthly, inter-annual) and seasons. ERA5, ERA5-Land, and CHELSA consistently underestimate precipitation variability. All datasets show more consistency with observations for temperature variability, with ERA5-Land generally exhibiting lower variability.
• Trends (1986-2020): CERRA shows the highest agreement with observed trends in precipitation and temperature metrics, though it tends to overemphasize the number of significant trends. ERA5, ERA5-Land, and CHELSA show inconsistent precipitation trends compared to observations. All datasets agree on a significant increase in mean daily temperature and maximum daily temperature.
• Extreme Events (Droughts): All reanalysis datasets capture the spatial patterns of the 2003 and 2018 droughts. ERA5, ERA5-Land, and CHELSA tend to overestimate drought severity and intensity, while CERRA generally underestimates them, sometimes simulating spurious local precipitation events that alleviate drought signals.
• Extreme Events (Heavy Precipitation/Floods): All datasets capture the intensity, severity, and spatial structure of the 1999 and 2005 heavy precipitation events. CERRA shows the highest agreement with observations. ERA5, ERA5-Land, and CHELSA tend to overestimate the spatial extent of these events and locally over/underestimate intensity.

Contributions

• Provides a comprehensive spatio-temporal evaluation of four state-of-the-art high-resolution reanalysis datasets (ERA5, ERA5-Land, CERRA, CHELSA-v2.1) for a broad range of hydro-climatic metrics (precipitation, temperature, snowfall, their variability, trends, and extreme events) over complex mountainous terrain (Switzerland).
• Highlights the superior performance of CERRA, particularly for precipitation metrics, variability, and heavy precipitation events, attributing its strength to the assimilation of additional precipitation observations.
• Identifies specific strengths and weaknesses of each dataset, including elevation-dependent biases and issues with spatial extent representation in coarser or statistically downscaled products.
• Offers practical recommendations for selecting suitable reanalysis datasets for various hydrological impact studies, especially in snow-dominated and complex terrain regions.

Funding

• Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (grant no. PZ00P2_201818)
• Swiss Federal Office for the Environment FOEN (HydroSMILE-CH grant)

Citation

@article{Wood2025Comparison,
  author = {Wood, Raul R. and Janzing, Joren and Hamel, Amber van and Götte, Jonas and Schumacher, Dominik L. and Brunner, Manuela I.},
  title = {Comparison of high-resolution climate reanalysis datasets for hydro-climatic impact studies},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-4153-2025},
  url = {https://doi.org/10.5194/hess-29-4153-2025}
}