Estimation of Suspended Sediment Concentration by ANN and Regression Analysis in Brahmani River Basin,India

Abstract

The estimation of suspended sediment yield is basic towards understanding the mass harmony between the sea and land. Direct estimation of suspended sediment is troublesome as it needs adequate time and cash. The suspended sediment yield relies upon various factors, and the connections between them are exceedingly non-linear and complex in nature. In this investigation, soft computing based sediment yield estimation calculations are proposed for the Brahmani river basin. A multilayer perceptron (MLP) artificial neural network (ANN) with an error back-propagation algorithm utilizing month to month hydroclimatic data (temperature, water discharge, and rainfall) was used to predict the suspended sediment yield at the Panposh measuring station, Brahmani river. The outcomes showed that water discharge, rainfall are huge controlling parameters of suspended sediment in the Brahmani River. The results demonstrate that the feed-forward back-propagation with Levenberg–Marquardt (FFBP-LM) is the best model for suspended sediment yield estimation, and gives progressively good results for high and low values. The prediction of the sediment rating curve (SRC) model was below expectation. It is likewise seen that the multiple linear regression (MLR) model predicted negative sediment yield at low values; which is totally improbable as suspended sediment yield cannot be negative in nature. It was additionally seen that suspended yield prediction by ANN was better looked at than using MLR SRC models. The proposed model will be advantageous for sediment prediction where estimates of suspended sediment values are missing or unavailable.