Accurate and reliable groundwater level forecasting models can help ensure the sustainable use of a watershed’s aquifers for urban and rural water supply. In this paper, a Self-Organizing-Map (SOM)-based clustering technique was used to identify spatially homogeneous clusters of groundwater level (GWL) data for a feed-forward neural network (FFNN) to model one and multi-step-ahead GWLs. The wavelet transform (WT) was also used to extract dynamic and multi-scale features of the non-stationary GWL, runoff and rainfall time series. The performance of the FFNN model was compared to the newly proposed combined WT–FFNN model and also the conventional linear forecasting method of ARIMAX (Auto Regressive Integrated Moving Average with exogenous input). GWL predictions were investigated under three different scenarios.The results indicated that the proposed FFNN model coupled with the SOM-based clustering method decreased the dimensionality of the input variables and consequently the complexity of the FFNN models. On the other hand, the application of the wavelet transform to GWL data increased the performance of the FFNN model up to 15.3% in average by revealing the dominant periods of the process.
Financed by the National Centre for Research and Development under grant No. SP/I/1/77065/10 by the strategic scientific research and experimental development program:
SYNAT - “Interdisciplinary System for Interactive Scientific and Scientific-Technical Information”.