Towards the development of bias-corrected rainfall erosivity time series for Europe

Rainfall erosivity maps for (near) real-time soil erosion predictions require the integration of (combinations of) reanalysis products and satellite-retrievals of rainfall, and the overcoming of potential bias related to their simplified spatial and temporal scale. Across Europe, we evaluate: 1) the European Meteorological Observations (EMO) dataset to simulate the localised characteristics of rainfall erosivity at the event scale (EI30), and 2) different implementations of quantile delta mapping (QDM) bias correction to improve the prediction skill. Between 1990 and 2014, evaluations were made at several spatial (location-specific, climatic zone and pan-European) and temporal (event, annual and long-term annual average) scales. The uncorrected EMO predictions demonstrated: 1) a slight overprediction of the number of EI30 events, 2) a reduced coefficient of variation in the EI30 (CV EMO = 1.57, CV REDES = 2.5), and 3) a relatively low (R2 = 0.22, n = 139,306) location-specific predictive skill, with higher discrepancies in all cases in Southern Europe. Following QDM, the EI30 predictions significantly better represented the large-sample variability of EI30 per climate region and improved the monthly correspondence. At specific locations, station-wise bias correction was the only implementation to improve the event (R2 = 0.24, n = 139,306), annual (R2 = 0.51, n = 14,248) and average annual (R2 = 0.76, n = 1142) predictions. While bias correction can improve rainfall erosivity predictions, applications should consider possibly large error propagation into subsequent predictions, regional disparities in performance, and the potential to improve the large-sample statistical correspondence but degrade the location-specific time series prediction.

MATTHEWS Francis;  MEDVED Anže;  BORRELLI Pasquale;  LIAKOS Leonidas;  VERSTRAETEN Gert;  PANAGOS Panos;  BEZAK Nejc; 

2025-01-17

ELSEVIER

JRC138482

1879-2707 (online),   

https://www.sciencedirect.com/science/article/pii/S0022169424018560,    https://publications.jrc.ec.europa.eu/repository/handle/JRC138482,   

10.1016/j.jhydrol.2024.132460 (online),