Vol. 8, Issue 2 (2019)

Author(s):
Shreya Nivesh, Pravendra Kumar, Bhagwat Saran, Pragati N Sawant and Ramesh Verma

Abstract:
The present study was undertaken to estimate the suspended sediment load from the Vamsadhara river basin comprising of 7820 km² area, situated between Mahanadi and Godavari river basins. Three daily input data groups or cases were employed using Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS), Fuzzy Logic (FL), Multiple Linear Regression (MLR) and Sediment Rating Curve (SRC) to find the effect of different inputs on the suspended sediment load. Input 1 consists of P_t, Q_t, Q_t−1, S_t−1 as inputs to the model to predict S_t. Input 2 consists of P_t-1, Q_t, Q_t-1, S_t-1 and Input 3 consist of P_t-1, Q_t, Q_t-2, S_t-1. The developed models were trained and tested. Three statistical parameters: root mean square error (RMSE), correlation coefficient (r) and coefficient of efficiency (CE) were used to compare the results of the models. Based on the performance analysis results revealed that the ANFIS model (RMSE-44.02 kg/sec, r-0.995 and CE-99.06%) outperformed other soft computing and conventional models. It can be concluded that the ANFIS models are more preferable and can be applied successfully for the estimation of the suspended sediment concentration for the study area.