Ecological Indicators (Nov 2024)
Analyzing watershed system state through runoff complexity and driver interactions using multiscale entropy and deep learning
Abstract
Quantifying watershed state is crucial for ecological management and sustainable development. Traditional methods, often based on multiple indicators and subjective weighting, struggle to objectively and comprehensively capture the inherent complexity of watershed ecosystems. Runoff, a key hydrological component, reflects watershed functioning, but most studies focus solely on runoff volume, limiting the ability to link runoff dynamics with surface conditions and broader system processes. This study introduced a novel approach that uses runoff complexity to represent watershed system-level state, bridging the gap between runoff dynamics and ecosystem functioning. The Hydrological Refined Composite Multiscale Entropy (Hydro_RCMFE) method was developed to quantify runoff complexity across various time scales. Combined with the Hydrological Complexity Transformer Model (HydroC_Trans) and Shapley Additive Explanations (SHAP), this method explored interactions between runoff complexity and influencing factors. It was applied to the Yanhe Watershed in China’s Loess Plateau, known for severe soil erosion and large-scale ecological restoration. The results revealed significant fluctuations in runoff complexity after 2005. Vegetation cover from the Grain for Green Program enhanced self-organization, buffering climatic variability while introducing instability. Urbanization further amplified runoff complexity, while landscape factors, such as aggregation and hydrological connectivity, had spatially and temporally varied effects, highlighting the need for tailored management strategies. Upstream, efforts should enhance climate resilience, increase vegetation cover—particularly grasslands—and improve landscape aggregation. Midstream and downstream strategies should prioritize optimizing hydrological connectivity, limiting impervious surface expansion, and restoring ecological functions.
