Prediction of cadmium concentration of soil using ANN and ANFIS models

Evaluation of soil contamination by heavy metals such as cadmium in soils is essential for human health and also for environment management. As direct measurement of soil cadmium is time-consuming and costly, in this study, the two methods of artificial intelligence, artificial neural network (ANN) and adaptive fuzzy neural network (ANFIS) used to estimate the amount of cadmium in soil as one of the dangerous heavy metals. For estimating of cadmium, soil readily available properties such as clay and sand percentage, organic carbon, EC, T.N and P used as input parameters and the relationship between these parameters and the concentration of cadmium established by ANN and ANFIS models. For training and testing the models, 250 soil samples collected from soils of Guilan province. For assessment of artificial intelligence models the statistical criteria such as the coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE) used. The results showed that ANN model with R2 = 0.83 and RMSE= 1.01, and MAE= 0.54 is superior to ANFIS model. Also the results of the sensitivity analysis on the input variables to the model showed that organic carbon and EC have the most and the least effect on the amount of Cd. The proposed model could be used to estimate the amount of cadmium in other parts of the studied area which the concentration of cadmium has not been measured, as well as for other areas with similar conditions.