Comparison of Adaptive Neuro-Fuzzy Inference System (ANFIS), Inverse Distance Weighting and geostatistics methods for estimating the water table (Case study: Dehgolan Plain, Kurdistan Province)

The decline of water table is very important in management point of view and might cause negative impacts such as land subsidence, raising costs and reducing groundwater quality. Groundwater is the most important source of water supply in Dehgolan plain. Increasing water requirements and extractions, has declined water table. This plain with an area of about 780 km2 is one of the protected plains of the Kurdistan province and with decrease in water table about 37 meters, it has the most decline between plains of the province. The purpose of this study is to model the groundwater level and compare the performance of the method of Adaptive Neuro-Fuzzy Inference System (ANFIS) with Inverse Distance Weighted (IDW), Kriging and Cokriging methods. For this purpose, in September 2016, the water table data relating to the 44 Piezometer digged in Dehgolan plain has been used to modeling. The results show that the hydraulic head behavior is different across the aquifer, so the use of spatial data (h) for modeling doesn’t lead to satisfactory outputs. The water table in Dehgolan plain has the highest correlation with topography conditions and the ANFIS with a RMSE = 0.07, MSE = 0.005, MAE = 0.06, MBE = 0.04 and = 0.88 R2, has presented better performance than other methods.