Application of Data-Driven Modelling to Flood Forecasting with a Case Study for the Huai River in China

The application of two data-driven modelling methods, M5 model tree and artificial neural network (ANN), to a flood forecasting problem is considered. The upper reach of the Huai River in China is used as a case study. It is shown that machine learning techniques could be an efficient tool in the problem of flood forecasting. M5 model trees, being analogous to piece-wise linear functions, have certain advantages compared to ANNs - they are more transparent and hence acceptable by decision makers, very fast in training and always converge. The accuracy of M5 trees is similar to that of ANNs. The improved accuracy in predicting high floods was achieved by building a modular model, in it the flood samples with special hydrological characteristics are split into groups for which separate M5 and ANN models are built. The hybrid model combining model tree and ANN, gives the best prediction result