Real-Time Flood Forecasting Using Self-Organizing Linear Output Network

Abstract

Self-Organizing feature map with Linear Output mapping (SOLO) algorithm was applied for real-time forecasting of flood discharge in Kuroki Dam Basin, Okayama. The SOLO algorithm is like a neural network algorithm, which is known for its advantage in quick calibration of a model and flexibility in forecasting of inexperienced pattern to the original neural network algorithm. The results show that the SOLO model provided accurate forecasts of one to three-hour ahead runoff with using feature vectors composed of principal components of hydrological data although the forecasts were not accurate enough by using feature vectors composed of the latest rainfall and runoff dataas an input to the SOLO model.