Short-term water level prediction using neural networks and neuro-fuzzy approach

A comparative study on a short-term water level prediction using artificial neural networks (ANN) and neuro-fuzzy system is addressed in this paper. The performance of the traditional approaches applied for such a hydrological task can often be constrained by data availability and simplifying assumptions in the processes description. In this paper, the ANN and neuro-fuzzy approaches are used for handling the situations with scarce data, where the predictions are based on the upstream hydrological conditions only. The models have been tested on two different river reaches in Germany. Moreover, the obtained results are compared to those of linear statistical models. Both ANN and neuro-fuzzy systems have performed comparably well and accurate for the purpose, explicitly outperforming the linear statistical models for a longer prediction horizon. The trained neural networks are partly implemented on-line, as a prototype of a web-based water level predictor.