Using Artificial Neural Network for Outflow Estimation in an Ungauged Area

This research employs an artificial neural network with a variable mathematic structure that is capable of simulating a nonlinear structural system. A back-propagation neural network (BPN) is adopted to estimate outflow for an ungauged area by considering temporal distribution of rainfall-runoff and the spatial distribution of watershed environment. The nonlinear relationship among the physiographic factors, precipitation, and outflow of the specific watershed was established to estimate the outflow of the sub-watershed where no flow gauge has been settled. The model was tested at Bei-Shi watershed of Hou-Long River, Taiwan. Three typhoon occurrences were used for model calibration and verification that indicates the model validity and proves the model suitable for estimating the outflow of an ungauged area.