Geospatial analysis of groundwater contamination by heavy oil in the Dammam aquifer-Middle of Iraq

Abstract

This study aims to apply Artificial Neural Network (ANN) to map the groundwater contamination of the Dammam aquifer by heavy oil in the middle of Iraq. For this purpose, the inventory map of 139 groundwater wells (contaminated and non-contaminated with heavy oil) with the seven important factors playing a role in controlling contamination were used. The factors are distance to faults, faults density, groundwater depth, aquifer saturated thickness and hydraulic conductivity, elevation, and distance from Abu Jir fault. For the performance of ANN model, five statistical measures were used namely, accuracy, sensitivity, specificity, kappa, and the relative operating characteristic curve. Obtained results from applying the model in R statistical package indicated that ANN has a high accuracy (> 90%) in training and testing phases. The probability prediction of ANN model was categorized into five groundwater contamination classes: very low-low, moderate, high-very high. The averages of areas occupied by these zones were 5267 km² (65%), 488 km² (6%), and 2362 km² (29%), for very low-low, moderate, and high-very high, respectively. The contamination map developed in this study could be used to drill successful non-contaminated groundwater wells and avoid loss of many efforts in drilling contaminated wells in the study area.