Evaluation of Combined ANFIS with Wavelet Transform to Modeling and Forecasting Groundwater Level

One of the most important factors, in a good management in any field, is having a proper perspective of the upcoming events. There is no exception in water resources management and the environment and awareness of the condition of water resources, in an area, plays a decisive role for planning water and agriculture. In this study, the Adaptive Neural Fuzzy Inference System (ANFIS) was used for the monthly forecast of Dalaki Basin groundwater levels in the province Bushehr in a 12-year period (2002- 2013). In order to improve the results of the model, the wavelet transform was used and the original signal was decomposed to sub-signals. Then, sub-signals were entered, as input, into ANFIS model to obtain the hybrid model, Wavelet-Adaptive Neural Fuzzy Inference System (WANFIS). To forecast the groundwater level of five observed wells has been used, using groundwater levels, precipitation, evaporation, andtemperature. Results showed that hybrid model, WANFIS, has better performance than ANFIS model. Also, it was showed that hybrid model has better performance in estimate extreme points. So, this method, using wavelet theory, increased the performance by 14%. At the end, groundwater levels were estimated by the best model in a year. The results of thepredicted groundwater levels showed that theincrease of having access to groundwater in the Dalaki area. This problem is noted to authorities of the area regarding the effects on water resources and the environment of the area.