Application of artificial neural networks to represent the rainfall-runoff process for flow forecasting

There are several different approaches for modelling the rainfall-runoff process, which we can divide in two main groups: deterministic models and statistical models. This paper describes the neural rainfall-runoff model for flood forecasting in real time, and compares it with the linear ARMAX model (autoregressive moving average with exogenous inputs). The results of this study have shown that the non-linear neural ANN model can be considered as an advanced statistical approach. The advantages of a neural operational forecasting model increase with growing lead time interval.