Wavelet and neuro-fuzzy conjunction model for predicting water table depth fluctuations

Abstract

The ability of a wavelet and neuro-fuzzy conjunction technique for groundwater depth forecasting was investigated in this study. The wavelet-neuro-fuzzy model was improved by combining two methods, the discrete wavelet transform and the neuro-fuzzy model. The conjunction model was applied to different input combinations of daily groundwater depth data of Bondville and Perry wells. Root mean square error (RMSE) and correlation coefficient (R) statistics were used for evaluating the accuracy of wavelet-neuro-fuzzy models. The accuracy of the conjunction models was compared with those of the single neuro-fuzzy models in one-, two- and three-day-ahead groundwater depth forecasting. Comparison of the results revealed that the wavelet-neuro-fuzzy models perform better than the neuro-fuzzy models especially for the two- and three-day-ahead forecasting cases.