Multi-model high-resolution projections of rainfall erosivity in Austria

Water erosion poses a significant threat to agricultural systems, causing soil degradation, loss of fertile topsoil, and adverse impacts on water quality. Climate change is expected to exacerbate these effects, as increasing rainfall intensities enhance water-induced soil erosion. Potential soil erosion by water is commonly assessed using the Universal Soil Loss Equation (USLE) and its revised version (RUSLE), which combine information on soil properties, topography, land cover, conservation practices, and rainfall erosivity, expressed by the R-factor. While the R-factor is typically derived from high-temporal-resolution precipitation measurements and extrapolated using regression-based approaches, its projection into future climates remains challenging. Uncertainties arise primarily from the strong sensitivity of rainfall erosivity to short-duration precipitation extremes, which are poorly represented in conventional climate projections. As a result, existing methods either apply present-day relationships to future conditions or rely on convection-resolving simulations that are limited in temporal coverage and resolution, as well as ensemble size, hindering a robust assessment of future erosion risk and associated uncertainties. 

In this study, we determine the R-factor on a monthly basis for Austria at a spatial resolution of 1 km × 1 km for a reference period (1995–2015), the mid-century period (2036–2065, RCP8.5), the far future (2071–2100, RCP4.5 and RCP8.5), and under a global warming level of 3 °C. We first develop multiple linear regression models using rainfall data of 184 stations in Austria. For future periods, the regression models were modified applying Clausius-Clapeyron scaling and Austrian climate projection data (ÖKS15). This framework allows us to spatially extrapolate rainfall erosivity across Austria, while retaining high temporal resolution and explicitly accounting for model and scenario uncertainty. 

Our results indicate that rainfall erosivity is highest in August and increases most strongly in alpine regions, where the R-factor already has the highest present-day values. The largest increases, but also the greatest model uncertainties, are associated with the RCP8.5 scenario in the far future. In this scenario, the median increase in the R-factor in August is at least 72% - in some regions, even over 100%. Overall, all climate scenarios consistently project an increase in rainfall erosivity in the future. Incorporating the projected R-factors into the RUSLE model enables the estimation of future water-induced soil erosion and supports more robust risk assessment and adaptation planning.