Abstract
Mapping vulnerability of water resources (VWR) is crucial for the sustainable management of water resources, particularly in freshwater-scarce coastal plains. This research aims to construct a coupled novel framework technique for assessing VWR using hydrochemical properties and data-driven models, e.g., Boosted Regression Tree (BRT), Random Forest (RF) with Support Vector Regression (SVR) as a classic model through k-fold cross-validation (CV). A total of 380 groundwater samples were collected during the dry and wet seasons to construct an inventory map. The models were used to demarcate the vulnerable zones from sixteen vulnerability causal factors using a 4-fold CV approach. Obtained results were validated using the area under the curve (AUC) of receiver operating characteristic (ROC), sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). The results showed the excellent capability of the models to identify the VWR zones in the coastal plain. The RF model showed higher performance (AUC = 0.93, NPV = 0.89, PPV = 0.86, specificity = 0.85, sensitivity = 0.90) than others models. The south-central and southwestern areas had a higher VWR due to salinity, NO3−, F− and As pollution in the coastal plain. Groundwater As, NO3− and F− pollution should be urgently monitored and possibly controlled in areas of high VWR. Decision-makers and water managers can utilize the VWR maps, derived usinga coupled novel framework, to achieve sustainable groundwater management and prevent anthropogenic activities at the regional scale.
[Display omitted]
•Vulnerability of water resources was assessed using a novel framework.•South-central and southwestern coastal plains of Bangladesh are found high vulnerable due to salinity, NO3−, F− and As.•Salinity, depth variation and nitrate are the most influential factors of water resources vulnerability.•Random forest was the optimal model for achieving high prediction accuracy.