Soil dispersivity coefficient (α) is a required input parameter in solute-transport models based on the Advection-Dispersion Equation (ADE). Soil dispersivity coefficient is typically difficult to model due to the complexity of the phenomenon. With respect to importance of this parameter, this paper presents the MLP- Artificial Neural Network (ANN) approach to predict the soil dispersivity coefficient. For training and testing of MLP model, the experimental data which measured in the rectangular tank with 1550mm length, 100mm width and 600mm height, were used. The collected data related to sandy soil with five sizes of very coarse, coarse, medium, fine and very fine and five distances of 25, 50, 75, 100 and 125 cm. NaCl was used as persistent pollutants with five velocities. The measured data such as transport distance (L), bulk density (ρb), porosity (n), hydraulic conductivity (K), average diameter of particles (D50), the pollutant velocity (Vc) were used as input data for predicting soil dispersivity coefficient (α). For comparison of results statistical criteria such as R2, RMSE and MAE were used. The ANN model was performed by different structures to minimize the prediction error and determine the optimum network configuration. The results show that the best architecture for the MLP model is comprised of 4 neurons with two hidden layers and transfer function of hyperbolic Tangent. The proposed MLP/ANN approach produced excellent results (R2 = 0.987, RMSE = 0.00036m and MAE= 0.00047m) compared to the MLR model (R2 = 0.943, RMSE = 0.00063m and MAE= 0.00043m) in testing set. Results show that the performance of ANN model was better than MLR model.
