Comparison of Using Different Systems of Artificial Intelligence in Subsurface Water Level Prediction (Case Study: Paddy Fields of Plain Areas between Tajan and Nekaroud Rivers, Mazandaran, Iran)

Novelty of the implementation drainage system in paddy fields and studies in this case led to the observation wells network with sufficient data does not exist in the northern region of the Iran and consequently does not provide access to subsurface water level of long-term data. Due to the complex and nonlinear behavior of subsurface water systems and to consider many factors affecting it, it seems to be difficult to predict the groundwater level. In this research, Artificial Neural Network (ANN) and Neuro- Fuzzy inference system (ANFIS) is used to predict the subsurface water level in paddy fields of Plain Areas between Tajan and Nekaroud Rivers. The results indicated by removal the wells that water depth is zero in them can be achieved reasonably accurate in predicting subsurface water depth in the study area and ANN with tangent sigmoid transfer function and with five neurons in the hidden layer and ANFIS with subtractive clustering and range of influence equal to 0.7 have almost the same accuracy in predicting the depth of subsurface water because the correlation coefficient of these two models are Respectively 0.8416 and 0.8593 and RMSE of them is 0.2667 and 0.2491.