Evaluation of convection-permitting models for rainfall erosivity in an Alpine region

Rainfall erosivity is a major driver of soil erosion and is highly sensitive to short-duration precipitation extremes, which are expected to intensify under climate change. Great advancement on climate data have been seen in the last decade, and convection-permitting climate models (CPMs) offer new opportunities to simulate rainfall characteristics relevant to erosivity. However, their performance in complex terrain remains insufficiently quantified.

We evaluate rainfall erosivity (RUSLE R-factor) simulated by a nine-member CPM ensemble from the CORDEX Flagship Pilot Study on Convective Phenomena over Europe, focusing on the Great Alpine Region. CPM estimates are compared with long-term, high-resolution rain-gauge observations from ~500 stations spanning a wide elevation range. We quantify and apply a temporal adjustment to reconcile hourly model output with 10-minute observations, then model performance vs observations is assessed for key erosivity-related variables, including rainfall intensity, event depth, frequency of erosive events, and mean annual erosivity. The CPM ensemble reproduces the spatial variability of rainfall erosivity with good skill and overall low bias, but exhibits clear elevation-dependent biases. Erosivity is underestimated at low elevations and increasingly overestimated at higher elevations, reflecting biases in rainfall intensity and/or event frequency. While low-elevation biases are largely consistent with sampling variability, high-elevation biases are predominantly systematic.

These results highlight the potential of CPMs for rainfall erosivity assessment and the importance of accounting for elevation-dependent biases in mountainous regions.