Genetic Programming for Modelling Long-Term Hydrological Time Series

In recent years, artificial neural networks (ANN) have emerged as a novel identification technique for the forecasting of hydrological time series. However, they represent their knowledge in terms of a weight matrix that is not accessible to human understanding at present. The purpose of this study is to develop a flow prediction method, based on the genetic programming (GP), which is an evolutionary computing method that provides `transparent´ and structured system identification. In terms of statistical characteristic of reservoir inflow sequence data, the autocorrelation function is employed to make certain amount of lagged input variables and the root mean square error is adopted as fitness of evaluation. The GP model is examined using the long-term observations of monthly river flow discharges. Through the comparison of its performance with those of the ANN, it is demonstrated that the GP model is an effective algorithm to forecast the long-term discharges.