Optimization of DRASTIC Model by Artificial Intelligence for Groundwater Vulnerability Assessment in Maragheh-Bonab Plain

Aquifer vulnerability assessment to define critical zones of pollution risk is an important method for groundwater resource management. By applying the DRASTIC model in this study, groundwater vulnerability in the Maragheh Bonab Plain aquifer was evaluated. The DRASTIC model uses seven environmental parameters (Depth to water, net Recharge, Aquifer media, Soil media, Topography, Impact of vadose zone, and hydraulic Conductivity) as seven layer in GIS media and finally a groundwater vulnerability map was created by overlaying the available hydrogeological data and categorized to low, moderate, and high risk. The DRASTIC index value was evaluated 81 to 116 for the study area. The vulnerability map created by DRASTIC is compared to nitrate data and the results indicate a relative correlation between the nitrate level and vulnerability index. In order to improve the model, four artificial intelligence (AI) models are adopted by optimizing the weights of the DRASTIC parameters. The four AI models are the Sugeno fuzzy logic (SFL), the Mamdani fuzzy logic (MFL), the artificial neural network (ANN), and the neurofuzzy (NF). For this purpose, the AI model input (the DRASTIC parameters), output (the vulnerability index), and nitrate concentration data was divided into two categories for training and test steps. The output of model in training step was corrected by related nitrate concentration, and after model training, the output of model in test step was verified by nitrate concentration. The results show that the four AI models are applicable to improve the correlation between nitrate level and vulnerability index using DRASTIC model for groundwater vulnerability assessment. The NF model by taking advantage of FL and ANN has the best results that high nitrate level at observation well location has high vulnerable index and was selected as a final model. According to the final model, the western areas of the aquifer are classified as high pollution risk. In conclusion, the AI approach proved to be an effective way to improve the DRASTIC model and provides a confident estimate of pollution risk for the study area.