Application of Machine-Learning Based Models for Prediction of Suspended Sediment Load in the Indian Peninsular River Basin

Suspended sediment load in rivers has a crucial impact on the river water quality, soil erosion, irrigation activities, and dam or reservoir operations. Dam construction in a river reduces the runoff, which increases the deposition of suspended sediment on the river course and ultimately leads to a change in the river channel morphology. Thus, suspended sediment load prediction is significant for planning and sustainable management of the riverine ecosystem. Researchers have used various physical models, such as sediment rating curves (SRC), SWAT, HEC-RAS, HEC-HMS, etc., for predicting suspended sediment load. Recently, researchers have used machine learning models to predict suspended sediment load in different hydroclimatic regions worldwide. In this study, we used five different machine learning models, such as ElasticNetCV, Multi‑Layer Perceptron (MLP) Regressor, Extreme Gradient Boosting (XGB) Regressor, Light Gradient-Boosting Machine (LGBM) Regressor and Linear Regression (LR), for predicting suspended sediment load in a downstream station of Godavari River Basin (GRB). The GRB is the largest Indian peninsular river basin, covering more than 0.3 million square kilometers of area. We used the 'Lazy Predict' Python library to achieve better results for machine-learning modeling. The data was collected for the period of 1970–2018 and divided into two parts, viz. pre-1990 and post-1990, to consider the dam effects on the downstream regions of the GRB. Performance evaluation revealed that the Multi‑Layer Perceptron (MLP) Regressor performed very significantly, with an r-square value of 0.71 and 0.74, respectively, for pre-1990 and post-1990. The developed models offer a valuable resource for decision-makers, environmental scientists, and water resource managers seeking to proactively manage sediment-related issues in river systems, ultimately fostering sustainable water quality and ecosystem health.