Scale-Specific Variation in Daily Suspended Sediment Load in Karst Catchments

Abstract

Temporal variation in sediment load is generally scale dependent and affected by many factors operating at different timescales. Due to the complexity of sediment transportation mechanisms and the nonlinear behavior of hydrological processes, identifying multi-timescale effects of variables on daily suspended sediment load (SSL) is challenging. Using multivariate empirical mode decomposition (MEMD), the aims of this study were to investigate scale-dependent relationships between daily SSL and associated variables and to predict daily SSL based on these relationships. Data on daily SSL in four typical karst catchments of southwest China and five potential variables affecting daily SSL (i.e., runoff, precipitation, air temperature, potential evapotranspiration, and Normalized Differential Vegetation Index) were collected during 2009–2012. The results indicated that MEMD decomposed temporal patterns of daily SSL and five potential variables affecting daily SSL into eight or nine intrinsic mode functions (IMFs) and a residue. Short timescale oscillations of IMF1 to IMF4 contributed more than 72% of the total variance in daily SSL in the four catchments, indicating that temporal variation in daily SSL was dominated by short intervals of 4−23 days. Runoff was the dominant explanatory variable for the overall daily suspend sediment load prediction at the observation scale. The predictions of daily SSL based on the IMFs and their residuals greatly outperformed those based on the original time series data. This result implies that a single scale (e.g., observation scale) is not sufficient to capture the complex relationships that exist between daily SSL and associated variables at all timescales. The results demonstrate the superiority of MEMD in characterizing scale-specific relationships between daily SSL and associated variables and improving daily SSL predictions. Given the complexity, nonlinearity, and nonstationarity inherent in many hydrological and sediment transport processes, MEMD is recommended for identifying scale-specific variation in daily SSL in future studies.

Keywords: Soil erosion, Nonlinear and nonstationary, Multiscale correlations, Multivariate empirical mode decomposition, Ecohydrology