Prediction of Water Availability by Using Tank Model and Artificial Neural Network (Case Study at Ciriung Sub-catchment Serang District)

Abstract

This study was conducted in Ciriung sub-watershed, Cidanau Watershed in Banten Province. total area of the sub-watershed is 118.01 ha. The land use is mostly dominated by mixen (88.27%), and dry paddy field (1 1.14%), settlement (0.59%). The purpose of this study predict water flow in Ciriung River in the future. Three steps were undertaken. The first p was to find the most effective evapotranspiration model for the area. The second step was to determine parameters of tank models. And, the third step was to forecast future rainfall Ond potential evapotranspiration values using Artificial Neural Network (ANN). The selected model is a standard tank model, which has four series of tank standing in a vertical mangement, and twelve parameters are involved, i.e., five parameters are in the surface, three parameters are in the intermediate, three parameters are in the sub-base, and one is in the base tank. One parameter and others are mutually interaction, and Marquardt algorithm was used for finding the optimum parameters. Three-layer of ANN w~th back-propagation were developed, trained and tested to forecast future rainfall and evapotranspiration. The climatic and the stream flow data were collected with digital instruments (logger).The results show that model Hargreaves along with Turc and Jensen-Haise models are the most effective evapotranspiration models for this location. The optimization technique to Tank model gained fast and accurate results of total flow and flow components. The ANN could forecast rainfall and evapotranspiration when trained on adequately representative data set. The result of forecasting of the future total runoff, there were various due to total rainfall and a-year daily rainfall distribution.