Temporal Scale Effects on Daily Sediment Yield in the Fu River Basin

[Objective] To analyze the scale-dependent relationships between daily sediment yield and key environmental factors and to explore effective prediction method to cope with the prediction challenges caused by the complexity of sediment transport mechanisms and the nonlinearity of hydrological processes. [Methods] Employing the Composite Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) approach, we analyzed the multi-temporal scale variability characteristics between daily sediment yield and five potential influencing variables (flow rate, precipitation, average temperature, potential evapotranspiration, and NDVI) in the controled regions of four hydrological stations in the Fu River basin of Southwest China from 2013 to 2018. Interactions between daily sediment yield and related variables at different temporal scales were identified, upon which predictions of daily sediment yield were predicted. [Results] The CEEMDAN method successfully decomposed the daily sediment yield and its potential influencing variables into 10-11 intrinsic mode functions (IMFs) and residual item, revealing the significant changes of the daily sediment yield at different temporal scales ranging from 3 to 730 days. After decomposition, daily sediment yield showed significantly enhanced correlation with howrate and precipitation across all temporal scales, whereas temperature, potential evapotranspiration, and NDVI were primarily significantly correlated with long-term scale sediment yields. By using stepwise multiple linear regression to predict the daily sediment yield, the prediction performance of the model is significant, and the R2 values exceeds 0.55 at all stations. [Conclusion] The CEEMDAN method effectively revealed the scale dependency between sediment yield and its influencing variables, offering a new perspective for understanding the dynamics of river sediment transport.