Evaluation of estimation methods for rainfall erosivity based on annual precipitation in Korea

Abstract

► We examine relationships between rainfall erosivity and eight rainfall parameters. ► Twelve types of linear and nonlinear regression models were statistically evaluated. ► Annual precipitation is better correlated with annual rainfall erosivity than others. ► The simplified estimation methods based on annual precipitation are still useful. Rainfall erosivity ( R factor) has been used as a primary input parameter for soil erosion, sediment yield and water quality modeling. At least 20 years of continuous high resolution rainfall data for the study area are required to compute original rainfall erosivity in ( R )USLE, but such data are not available for many locations. Some simplified methods for estimating rainfall erosivity using readily available data have been presented to overcome this problem and are used in many countries. In Asia, annual precipitation is often used to estimate the R factor for ( R )USLE modeling. The purpose of the present study is to evaluate models for estimating rainfall erosivity based on annual precipitation and to identify the most applicable model for Korea. Over 20 years of actual rainfall erosivity data for 33 Korean weather stations calculated by high resolution precipitation data were used in this study. Correlation analyses between actual rainfall erosivity and eight rainfall parameters were investigated to identify appropriate estimators of rainfall erosivity. We found that 31 of the 33 stations indicated a strong positive relationship ( r > 0.5) between annual rainfall erosivity and annual precipitation with a 99% confidence level, but two stations indicated no correlation between the two variables. Twelve types of linear and nonlinear regression models were statistically evaluated for each weather station. Some parameters proved more useful to predict R factors for particular stations than annual precipitation. Nevertheless, the simplified estimation methods based on annual precipitation are still useful to predict long-term annual rainfall erosivity for the majority of locations in Korea. Finally, new simple regression models appropriate for each rain-gauge station were determined. Rainfall erosivities estimated by regression models for specific locations had limitations when used to predict actual rainfall erosivity in other locations due to site-specific conditions. Thus, simplified methods for estimating rainfall erosivity should be used with caution depending on location or period.