Sediment Load Prediction Using Combining Wavelet Transform and Least Square Support Vector Machine

In rivers and streams, sediment transport is a common occurrence that greatly contributes to ecosystem production and maintenance by replenishing essential nutrients and conserving the natural habitats of aquatic life. SSL prediction is a difficult task because of the intricacy and stochastic nature of sedimentation, and standard approaches frequently produce unreliable findings. Machine learning (ML) models are now frequently used to handle challenging issues like SSL modeling. In order to predict SSL in the Subarnarekha River, present work develops a reliable methodology based on a least square support vector regression (LS-SVM) model with wavelet transform (WT) as a preprocessing method. To increase the capability of ML models in SSL prediction, the WT technique was applied. Various combinations of these inputs were tested while estimating monthly SSL using discharge and sediment data. With WI of 0.9909, RMSE of 1.005, and NSE of 0.9871, the suggested WT-LSSVM model demonstrated superior and more reliable predictions. The findings of this investigation supported the applicability of the suggested methodology for accurate modeling of SSL.