Testing of catchment module of integrated reservoir-based canal irrigation model for kangsabati irrigation project

Bhadra, (2007) developed Integrated reservoir based canal irrigation model (IRCIM). It consist of catchment, reservoir, crop water demand modules. In this study, IRCIM was applied on Kangsabati irrigation project, West Bengal, India for period of 1998 to 2003. Runoff was predicted using two techniques namely, Distributed SCS Curve Number (CN) with Muskingum routing and Artificial Neural Network (ANN) Backpropogation techniques available in catchment module. Distributed SCS CN method requires subbasin information, land cover characteristics, overland and channel information and daily rainfall on subbasin, whereas ANN method requires daily rainfall and runoff values. Catchment module was calibrated and validated using performance criteria modelling efficiency (ME) and coefficient of residual mass (CRM). ANN technique of runoff prediction involves extensive training of the network, where the unpredictable correlation of rainfall and runoff is also been taken into consideration which is not possible for conceptual model such as SCS CN method. Thus, results showed that for Kangsabati reservoir catchment, runoff values, predicted using ANN result in better match with observed runoff values compared to semi-distributed conceptual SCS CN method.

Runoff prediction by Empirical method ANN with Levenberg-Marquardt algorithm more accurate than Physical based distributed (GIS-based SCS CN method with muskingum routing)