A Global Rain-Driven Soil Erosion Investigation Based on Simulated Breakpoint Precipitation

Abstract

HighlightsAn international dataset of simulated breakpoint precipitation climate stations was used to overcome the limitations of fixed-interval precipitation in global soil erosion applications.The international simulated breakpoint dataset was validated against collocated high-quality, high-resolution precipitation data from a ground network and other data sources.The process-based Rangeland Hydrology and Erosion Model (RHEM) was used to predict erosion based on global climate classifications and soil properties.Critical precipitation factors in RHEM scenarios were identified based on their ability to predict erosion rates.Abstract. Recent research has highlighted problems with erosion modeling applications that use coarser fixed-interval precipitation data as opposed to breakpoint precipitation data, which better preserves precipitation characteristics such as intensity and duration. Due to their wider availability, erosion modeling applications and risk assessments are typically based on fixed-interval data. However, these applications could be subject to substantial erosion underestimation bias related to time-averaged precipitation. Alternatively, this manuscript presents a novel approach to global-scale erosion assessment based on simulated breakpoint precipitation data. A point-scale stochastic weather generator, CLIGEN, was used to generate precipitation events with characteristics more similar to breakpoint precipitation data than fixed-interval alternatives. An international CLIGEN dataset of climate parameters from more than 10,000 long-term climate stations in numerous countries was evaluated for use in parameterizing CLIGEN simulations. CLIGEN-simulated event characteristics and derived statistics such as annual rainfall erosivity were compared to high-quality, high-resolution NOAA-ASOS precipitation data where available. The Rangeland Hydrology and Erosion Model (RHEM) was used to predict runoff and soil loss based on the same CLIGEN inputs for simple bare soil scenarios with site-specific soil textures taken from the global 250m SoilGrids product. Erosion results were analyzed according to climate type, revealing that predicted distributions of sediment yield and runoff were statistically unique for most global climate types. A multivariate regression model was developed to explore and understand the importance of various precipitation input factors. Peak precipitation intensity was the most critical climate factor for determining sediment yield, and combined CLIGEN precipitation factors had approximately as much predictive power as soil texture. Average annual rainfall erosivity values were calculated for each location and were 25% and 20% greater than values from RUSLE2 and Panagos et al. (2017) estimates, respectively. This finding is in agreement with the latest research on the topic. Keywords: ASOS, CLIGEN, Erosivity, Global soil erosion, Machine learning, RHEM, Simulated breakpoint precipitation.