Quantitative analysis of factors controlling sediment yield in mountainous watersheds

Sediment and turbidity are primary causes of impaired river ecosystems; remedial action for these impairments requires the identification of their sources and controlling factors. This paper examines the combined effects of watershed complexity in terms of land use and physiography on the specific sediment yield of the upper Du River watershed (8973km2) in China. The land use composition, land use pattern, morphometric variables, and soil properties of the watershed were calculated at the subwatershed scale and considered to be potentially influential factors. Because these watershed characteristics are highly codependent, a partial least-squares regression (PLSR) was used to elucidate the linkages between the specific sediment yield and metrics composed of 19 selected watershed characteristics. The first-order factors were identified by calculating the variable importance for the projection (VIP). The results revealed that the land use composition and land use pattern exerted the largest effects on the specific sediment yield and explained 65.2% of the variation in the specific sediment yield. A set of physiographic indices was also found to have a large effect on the specific sediment yield and explained 17.7% of the observed variation in the specific sediment yield. The following are the dominant first-order factors of the specific sediment yield at the subwatershed scale: the areal percentages of agriculture and forest, patch density, value of the Shannon's diversity index, contagion, value of the hypsometric integral, and saturated soil hydraulic conductivity. The watershed size exerted a substantial effect on the sediment delivery ratio (SDR). The VIP values also suggested that the Shannon's diversity index, contagion, and hypsometric integral are important factors in the SDR. With a readily available digital spatial database and rapid developments in geographic information system (GIS) technology, this practical and simple PLSR approach could be applied to a variety of watersheds. •We studied the sediment yield response to various watershed characteristics.•The PLSR approach is a useful tool for analyzing correlated data.•Land-use type and patterns have the largest effect on the specific sediment yield.•An SDR model as a function of watershed characteristics was developed.