Matthews et al. (2025) Dynamic assessment of rainfall erosivity in Europe: evaluation of EURADCLIM ground-radar data

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-10-20
• Authors: Francis Matthews, Pasquale Borrelli, Panos Panagos, Nejc Bezak
• DOI: 10.5194/hess-29-5299-2025

Research Groups

• Department of Science, Roma Tre University, Rome, Italy
• KU Leuven, Department of Earth and Environmental Sciences, Leuven, Belgium
• Department of Environmental Sciences, Environmental Geosciences, University of Basel, Basel, Switzerland
• European Commission, Joint Research Centre (JRC), Ispra, Italy
• Faculty of Civil and Geodetic Engineering, University of Ljubljana, Ljubljana, Slovenia

Short Summary

This study evaluates the ground radar-based EURADCLIM dataset for quantifying rainfall erosivity across Europe, finding that it initially overpredicts erosivity due to radar artifacts but significantly improves with an 80 mm h⁻¹ I30 threshold, offering unique spatial detail for soil erosion prediction.

Objective

• To evaluate the performance of EURADCLIM ground-radar compilations in estimating large-scale rainfall erosivity patterns in Europe at various timescales.
• To analyze the implications of imposing I30 threshold values to limit the influence of rainfall retrieval errors in EURADCLIM.
• To compare the R-factor derived from EURADCLIM with global rainfall erosivity products (GloREDa, GloRESatE, IMERG, CMORPH).
• To discuss the advantages and limitations of using EURADCLIM for rainfall erosivity estimation and its potential within ensemble rainfall erosivity predictions.

Study Configuration

• Spatial Scale: Pan-European coverage, with EURADCLIM data at a 2-kilometer grid resolution. Comparisons are made at country-average levels and point-scale for gauge data.
• Temporal Scale: Event-scale (EI30), 1-hourly (EURADCLIM input), 30-minute (disaggregated), monthly, and annual average rainfall erosivity (R-factor). Data periods: EURADCLIM (2013–2022, primarily 2016–2022 for calculations), GloREDa (predominantly pre-2013, with 2013 for event comparison and 2016–2020 for Slovenian stations).

Methodology and Data

• Models used:
  • Universal Soil Loss Equation (USLE) / Revised Universal Soil Loss Equation (RUSLE) methodology for calculating rainfall erosivity (EI30, R-factor).
  • Rule-based rainfall disaggregation scheme: 25 % of hourly rainfall in the first 30 minutes and 75 % in the second 30 minutes, used to derive 30-minute rainfall from 1-hourly EURADCLIM data.
  • Kling-Gupta Efficiency (KGE) for performance evaluation.
• Data sources:
  • EURADCLIM (EUropean RADar CLIMatology) v2.0: Ground radar-based precipitation grids (1-hourly, 2 km resolution) derived from OPERA (Operational Program on the Exchange of Weather Radar Information) and adjusted with daily rain gauge data from the European Climate Assessment & Dataset (ECA&D).
  • GloREDa (Global Rainfall Erosivity Database) 1.2: Gauge-based interpolations of average monthly and annual rainfall erosivity, and event (EI30) time series from over 1300 European stations.
  • Sentinel-2: NDVI data for assessing soil cover conditions.
  • Comparison datasets: GloRESatE, IMERG, CMORPH (satellite-based rainfall erosivity products).

Main Results

• EURADCLIM initially overestimates rainfall erosivity compared to GloREDa, with an average annual R-factor of 1470 MJ mm ha⁻¹ h⁻¹ versus GloREDa's 719 MJ mm ha⁻¹ h⁻¹, and a 96 % percent bias for country-averaged R-factors. This is primarily due to residual artifacts in some regions inflating instantaneous rainfall rates.
• Overprediction is most significant in regions with complex topography and lower radar/rain gauge coverage (e.g., Balkans), while flatter regions with better radar coverage show better spatial prediction but a tendency towards underprediction.
• Event (EI30) time series analysis showed reasonably good performance (KGE > 0.4) in 50 % of evaluated gauge locations, but significant overprediction by EURADCLIM was evident in the upper quantiles.
• Applying an 80 mm h⁻¹ threshold to limit the maximum I30 value during rainfall erosivity calculation significantly improves EURADCLIM's performance. With this adjustment, the annual country-averaged R-factor percent bias drops to 9 %, and monthly R² values range from 0.49 to 0.94, with percent biases from -15 % to 103 %.
• The adjusted EURADCLIM dataset shows the best agreement with GloREDa across Europe in July and August, with larger differences observed in June and winter.
• EURADCLIM offers unique spatial detail for detecting small-scale rainfall features (e.g., convective cells) critical for predicting erosion in susceptible fields, which is valuable for instantaneous erosion mapping when combined with soil cover data (e.g., Sentinel-2 NDVI).
• The inclusion of EURADCLIM in a multi-product ensemble (with GloREDa, CMORPH, IMERG, GloRESatE) adds significantly more spatial detail to the patterns of disagreement between datasets, highlighting its potential as a component for future updatable pan-European rainfall erosivity predictions.

Contributions

• First study to investigate ground radar-based estimates of rainfall erosivity in Europe using the EURADCLIM dataset.
• Comprehensive evaluation of EURADCLIM's performance for rainfall erosivity at event, monthly, and annual timescales across Europe, identifying key biases and their spatial patterns.
• Demonstration of a simple yet effective bias correction method (I30 threshold of 80 mm h⁻¹) that significantly improves EURADCLIM's accuracy, bringing its performance in line with or superior to some satellite-based products.
• Highlighting the unique value of ground radar data in resolving fine spatial detail of rainfall events for soil erosion modeling, especially in the context of multi-platform rainfall erosivity ensembles.
• Providing insights into the limitations of both ground radar and gauge-based datasets for rainfall erosivity estimation and suggesting future directions for ensemble-based approaches.

Funding

• Slovenian Research Agency (ARRS) through grants no. P2-0180, J6-4628, J2-4489, N2-0313.
• UNESCO Chair on Water-related Disaster Risk Reduction.
• Horizon Europe projects Soil O-LIVE (Grant No. 101091255).
• Horizon Europe projects AI4SoilHealth (Grant No. 101086179).
• Swiss State Secretariat for Education, Research and Innovation (SERI), grant agreement no. 101086179, AI4SoilHealth.

Citation

@article{Matthews2025Dynamic,
  author = {Matthews, Francis and Borrelli, Pasquale and Panagos, Panos and Bezak, Nejc},
  title = {Dynamic assessment of rainfall erosivity in Europe: evaluation of EURADCLIM ground-radar data},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-5299-2025},
  url = {https://doi.org/10.5194/hess-29-5299-2025}
}