Precipitation transformation in rainfall-induced landslide forecasting

Abstract

Precipitation is an important meteorological element widely used in rainfall-induced landslide forecasting. Most existing studies on rainfall-induced landslide forecasting focus on the determination of forecasting models and parameters. Five transformation methods of precipitation are evaluated in this paper to improve the forecasting models through a case study of Shenzhen, based on rain gauge observation data of a rainfall process between June 11 and June 13, 2008 which induced one hundred and sixty landslides. The square root transformation, cube root transformation, and logarithm transformation make the transformed precipitation curves more similar as the typical saturation curves of landslides by increasing the weights and sensitivities of small precipitation values and can improve the forecasting results among which the one of the logarithm transformation is the best. The square transformation and cube transformation that decrease the influences of small precipitation values worsen the forecasting results. This study can provide a guide to further research on landslide forecasting.