Comparative analysis of rainfall erosivity estimation approaches and temporal resolution effects: A critical review supported by an observational case study.

Accurate estimation of the rainfall erosivity factor (R-factor) is fundamental for soil erosion prediction and informed land conservation. This study provides a comprehensive review of the historical evolution and current status of R-factor estimation, ranging from simplified empirical equations to complex process-based models. Special emphasis is placed on data requirements, model assumptions, and method transferability. We introduce an alternative process-based algorithmic procedure (AAP) for identifying erosive events that strictly aligns with the original Renard et al. (1997) criteria, effectively mitigating the overestimation of wet spell durations common in traditional approaches. Using a high-resolution (1-min) rainfall dataset from 134 stations in Sardinia, Italy, we evaluate the performance of USLE, RUSLE, and RUSLE2 under various event identification procedures and temporal resolutions (1 to 1440 min). Furthermore, we compare these process-based results against 51 widely used simplified empirical models from global literature. Our analysis reveals a significant dependence of erosivity estimates on temporal resolution; coarser data systematically smooth intensity peaks, leading to substantial R-factor underestimation. Additionally, simplified empirical models exhibit high epistemic uncertainty, with estimates often deviating by over 100% from process-based benchmarks due to local calibration constraints. The study concludes with recommendations regarding possible future research directions, based on important gaps identified in the existing literature.