Estimation of groundwater depth using ANN-PSO, kriging, and IDW models (case study: Salman Farsi Sugarcane Plantation)

Appropriate management of groundwater resources requires accurate information about the characteristics of the groundwater table, spatial distribution of its characteristics, and the constant depth of the water table and its fluctuations. One of the most important issues in the quantitative management of groundwater resources is the estimation of water table using the data collected from the observation well network. In this study, to simulate the depth of groundwater Salman Farsi Sugarcane Plantation, three methods of Artificial neural network-integrated with particle swarm optimization algorithm, geostatistics (Kriging) and IDW was used. Inputs data include evapotranspiration, air temperature, precipitation and geographic location. The results showed that the highest simulation accuracy of groundwater depth in Salman Farsi Sugarcane Plantation was related to the ANN-PSO model with the highest R2 (0.95) index and lowest RMSE and MAE (to 1.05 and 1.11) values. Also, among the Kriging and IDW models used, the accuracy of the Kriging model was more than the IDW model. Due to the acceptable accuracy of the results of the three models, the water resource planner and -maker in this field can apply this optimum interpolated groundwater depth to monitor the spatiotemporal fluctuation of groundwater depth in this area by updating its data.