Calibration of Electromagnetic Induction Device (EM38) in Order to Estimate Vertical Variation of Soil Salinity Using Machine Learning Techniques

Soil salinity is a serious environmental problem especially in arid and semiarid areas. Therefore, it is vital to generate and update soil salinity maps in order to determine early stage of salinization. Electromagnetic induction instrument is an alternative to traditional methods for assessing soil salinity. Different methods have been used to calibrate electromagnetic induction instrument. At present research, an attempt was made to calibrate EM38 in pistachio orchard located in Ardakan city using multi-linear regression (MLR), artificial neural network (ANN) and neuro-fuzzy (ANFIS). To calibrate and predict soil salinity in nine standard depths, the best result was obtained by ANFIS model with R2 of 0.06, 0.11, 0.30, 0.59, 0.69, 0.64, 0.70, 0.74 and 0.74; and RMSE of 4.09, 3.66, 2.87, 2.22, 2.26, 2.62, 2.46, 2.38 and 2.50, respectively; which showed the accuracy of ANFIS was higher than other models (ANN and MLR) to predict soil salinity and calibrate EM38.