Scale-specific controls of sediment yield in karst watersheds

Abstract

Variations in sediment yield are controlled by many variables that operate at different scales. The relationships between sediment yield and its influencing factors at multiple time scales are still unclear since the underlying processes overlap at diverse scales. The objective of this study was to use multivariate empirical mode decomposition (MEMD) to reveal scale-specific controls on sediment yield. Annual sediment yield and its five influencing factors including runoff, temperature, precipitation, potential evapotranspiration, and NDVI were collected from 1982 to 2015 from six typical karst watersheds in southwest China. Using MEMD, the different scales of controls on sediment yield were determined by decomposing the data into different intrinsic mode functions (IMFs) and residues. Sediment yield was found to be significantly correlated with its potential controlling factors at two or more scales for each watershed, although no significant correlations were observed between sediment yield and some of the influencing factors at the observation scales. The factors controlling sediment yield were found to be scale-specific and varied by watershed, and some factors controlling sediment yield were more deterministic at longer time scales. The scale-specific controlling factors at a particular scale or residue were able to predict sediment yield at each scale (IMF) or residue. Furthermore, sediment yield at the observation scale was predicted by summing up all the IMFs and the residues derived from the factor components at equivalent scales. Runoff and precipitation were the major predictors for sediment yield in all watersheds except in the Hongshui watershed where the temperature and NDVI were the major explanatory variables that predicted overall sediment yield. The predictions of overall sediment yield using MEMD were better (R2 ranges from 0.81 to 0.92) than the predictions of sediment yield made using the original data (R2 ranges from 0.32 to 0.78). This suggests that the complex relationships between sediment yield and associated factors are not captured by single scale (observation scale) analysis. This study shows that MEMD may be a useful tool for revealing the scale-dependent relationships between sediment yield and its controlling factors in karst watersheds.