Abstract
As the rain infiltrates the ground, the runoff of the rain flows into the river. It is difficult to know accurately the amount of rain that infiltrated the ground and the saturated amount of rain under ground as well as a period of conversion from infiltrated rain into a runoff. Thus it is very difficult to forecast stream flow by modeling a runoff process. The runoff analysis (e. q. tank model) is useful to forecast stream flow. In this study, we construct a system of runoff analysis by using neural network, which can be effective to solve non-linear and pattern classification problems, using the same data used in the tank model by KOZUE, and then compare to the results of the tank model. We evaluated the efficiency of the system using neural network on runoff analysis.
