Groundwater Level Fluctuation Simulation Using Support Vector Machines and Adaptive Neuro Fuzzy Inference System (Case Study: Maragheh Plain)

In order to optimal management of groundwater resources, accurate estimate of groundwater level fluctuations is required. In recent years, the use of artificial intelligence methods based on data mining theory has increasingly attracted researchers' attention. The purpose of the present study is to compare the performance of adaptive neuro-fuzzy inference system (ANFIS) and support vector machine (SVM) methods to simulate groundwater level fluctuations. A 22-year dataset (1996-2018) including hydrological parameters such as monthly precipitation (P) and groundwater level (GL) from 25 observation wells in some parts of Maragheh plain located in East Azarbaijan province were used as models input data. The average groundwater level in the study area is 1321 m and the annual precipitation and temperature was calculated 294 mm and 14 ◦C, respectively. Mean values of statistical indices of correlation coefficients and root mean square error were calculated 0.91 and 0.38 m for the ANFIS model and 0.92 and 0.40 m for the SVM model, respectively. Results showed that the addition of monthly precipitation parameter to the input data had no significant effect on the accuracy of the ANFIS model, however, the model prediction accuracy increased by 14% for the SVM model. In general, the simulation accuracy of both models was acceptable. However, it can be stated that the ANFIS model has a slight advantage over the SVM model.