Development of empirical regression model for predicting monthly rainfall erosivity factors based on daily rainfall and its application to future rainfall scenarios in South Korea

Study region South Korea, with daily rainfall data retrieved from 58 meteorological stations across the country. Study focus To improve soil erosion risk assessments constrained by data scarcity, this study compares four monthly rainfall erosivity ( R -factor) prediction models using time-based walk-forward cross-validation. Spatial non-stationarity, regional erosivity regimes, and future shifts under climate change scenarios are also investigated. New hydrological insights for the region Across the country, effective rainfall amount (rain 10 ) consistently exerts a stronger influence on erosivity than rainfall frequency (days 10 ), indicating that future regional erosivity will be more sensitive to intensified rainfall than to frequency shifts. Spatial distribution of model coefficients reveals five distinct hydroclimatic regimes; inland areas are heavily influenced by rainfall frequency, whereas coastal and island zones are dominated by rainfall amount driven by typhoon activity and orographic enhancement. Climate projections further indicate a warming-induced snow-to-rain transition that increases January-February rainfall erosivity by upto 300% under high-emission scenarios, redistributing erosive rainfall from the traditional summer monsoon toward the winter non-cropping season. These findings indicate that future soil erosion risk in South Korea will be increasingly governed by the intensity and seasonal redistribution of erosive rainfall, rather than changes in annual rainfall alone.