Simulation of Soil Hydraulic Conductivity Using Adaptive Fuzzy Neural Inference System Model (Case Study of East Azarbaijan Province Soils)

Study of soil hydraulic properties such as saturated hydraulic conductivity in flow studies in porous media is necessary. Determination of saturated hydraulic conductivity using direct methods in spite of technological advances is still time consuming. In addition, due to the high spatial variability of ks, it is very difficult to determine this parameter especially in the case of large - scale studies such as a basin. Therefore, in the present study, Adaptive Neuro - Fuzzy Inference System (ANFIS) was used to estimate saturated hydraulic conductivity.For this purpose, 60 soil samples were taken from different parts of east Azerbaijan province and physical parameters including pH,Ec, organic carbon content, bulk density, sand,clay and silt percentages were measured, Between field and laboratory methods, a field method was used to determine water saturation of soil at the top of the water table. In the next step, input data to the model were defined in nine different models. Then 70% of the data were considered for model training and 30% for the test data. To evaluate the performance of ANFIS, statistical indices of mean deviation error (MBE), Nash Sutcliffe (NS) and root mean square error (RMSE) were considered. The results showed that the AFIS model with the sixth pattern has the best performance with statistics, MBE and RMSE equal to 2.45, 1.72 (cm/ h) and NS, equal to 0.96.