Comparing Performans of Fuzzy Logic, Artificial Neural Network and Random Forest Models in Transmissivity Estimation of Malekan Plain Aquifer

Estimating aquifer hydrogeological parameters is essential for the studies or management of groundwater resources. There are several different methods such as pumping test, simulation or modeling of groundwater, geophisical modeling to estimate these parameters. Although analysis and evaluation of pumping test data is the best way to achieve this purpose, it is costly, time consuming and the gained results are from limited points. Malekan plain aquifer is one of the marginal plains of Urmia Lake which suffered more ground water declination and Salinization in last decades and it needs qualitative and quantitative management. In this study, artificial neural networks, fuzzy logic and random forest models have been used to estimate the transmission of aquifers and the performance each of these models has been investigated. Inputs of presented models included related geophysical and hydrogeological variables to transmissivity such as transverse resistivity (Rt), electric conductivity (EC), alluvium thickness (B), and hydraulic conductivity (k). Based on the results of all models, random forest model has higher accuracy and ability to predict transmissivity parameter. According to this model, electrical conductivity (EC), aquifer environment (A) and hydraulic gradient (H) parameters were identified as the most important parameters to predict the transmissivity, respectively.