Predicting river sediment deposition using machine learning and InVEST-SDR modeling- A hydro electric perspective

Sediment transport significantly impacts hydropower reservoir efficiency, flood risk, and river morphology. However, accurate modeling remains challenging due to fluctuating sediment loads and a lack of standardized methods, highlighting the need for improved predictive approaches. This study focuses on evaluating the performance of rainfall prediction models and their implications for sediment transport and soil erosion using the InVEST-SDR framework. It aims to understand the impact of changing land cover and rainfall patterns on soil loss and sediment export over time in the Hydro Electric Project area. The study utilizes Random Forest Regression (RFR) for rainfall prediction, which achieved a high R² value of 0.78, indicating strong predictive accuracy. Additionally, RFR showed the lowest error rates, with a Root Mean Square Error (RMSE) of 5.55, Mean Absolute Error (MAE) of 2.07, and Mean Squared Error (MSE) of 30.84. LULC was classified using Sentinel-2. The InVEST-SDR model was used to assess soil loss and sediment transport. Pearson correlation analysis between model results and GIS-based predictions further validated the reliability of the models. The study found a decreasing trend in potential soil loss and sediment export from 1980 to 2028. Sediment export also showed a declining trend, from 1.79e-23 to 5.22e-11 tons/100m2/year in 1980 to 1.65e-23 to 4.77e-11 tons/100m2/year in 2023, with projections indicating a further reduction to 1.22e-23 to 3.55e-11 tons/100 m2/year by 2028. The strong Pearson correlation of 0.89 between soil loss in 1980 and sediment export in 2023 and 2028 indicates the predictive value of historical soil loss. In contrast, the declining correlations of 0.044 and 0.043 for sediment deposition highlight changing dynamics over time. The effectiveness of vegetation in soil retention also declined over time, highlighting the need for maintaining vegetation cover for effective soil conservation and erosion control.