Multiple input single output (MISO) linear model for sediment yield prediction from upper Ganga basin

Abstract

High rate of sediment outflow from catchments and high level of sedimentation in rivers and other water bodies is a serious cause of concern globally as it leads to physical disruption of the hydraulic characteristics of rivers and various water bodies. The sediment deposition in water bodies like lakes, reservoirs and ponds not only results in reduction of storage capacity but also adversely hampers irrigation potential, water supply to domestic and industrial use, hydroelectric power generation and drought mitigation programs. The present study was conducted to develop multiple input single output (MISO) model to predict sediment outflow from upper Ganga basin. In the development of this model the present day sediment yield has been correlated with previous days sediment yield, present and previous days runoff. Using the techniques of stepwise regression only significant variables were retained and the remaining variables were excluded at 5% level of significance. The model was developed considering data of 1980-2010. The applicability of the developed model was verified for future prediction using the data set of 2010-2011 which was not considered in the development of the model. It was found that the correlation coefficient values were 0.93 and 0.97 respectively, during training and cross validation period. The model performance was evaluated by performance evaluation indices such as MAPE, RMSE, NRMSE, ISE, CE, IOA and VE. The values of these performance indices were obtained as 50.51% and 4.22%; 1.09 and 1.97 tons/sec; 1.03 and 0.55; 0.03 and 0.09; 87.2% and 93.8%; 96.5% and 98.3%; and -0.02% and -7.65% respectively during training and cross validation.