Developing empirical relationships to predict loess slide travel distances: a case study on the Loess Plateau in China

Abstract

To improve landslide hazard mapping quality, a functional relationship with travel distance prediction is essential. To obtain a more accurate empirical relationship for predicting loess slide travel distances, we developed a loess slide database for the central Loess Plateau using a combination of remote sensing image interpretations, existing datasets, and an intensive field survey. The loess slide travel distance was concentrated within less than 200 m, according to a cumulative frequency analysis. Our results reveal that the loess slide volume, slope height, and slope inclination of the sliding area control the travel distance, and this relation is well-described by a power law function. Furthermore, statistical analysis suggested that the equivalent coefficient of friction decreases with an increase in loess slide volume but increases with an increase in slope inclination. We compared the prediction performances of four empirical relationships proposed in this study using the mean absolute percentage error and Theil inequality coefficient methods. We discovered that the empirical relationship with three independent variables can more accurately predict the loess slide travel distance than the relationships with one or two independent variables.