Groundwater arsenic contamination risk prediction using GIS and classification tree method

Groundwater arsenic contamination is a serious health and environmental problem in Bangladesh. Sources or origin of this heavy metal is known to be natural but the process through which it is being released is poorly understood. In quest of mitigation of the problem it is important to predict area of probable contamination before it causes any damage to human health. In this research an attempt has been made to develop arsenic contamination risk map using a combination of classification tree and GIS technology. Six factors, which might have relation with arsenic contamination, namely geology, soil reaction (pH), elevation, soil organic matter, soil iron and proximity to river were selected as predictor variables. Prediction model was developed using classification algorithm of data mining technology in MATLAB using 80% training data. The developed classification trees were then validated using 20% test dataset. The best tree is found at pruning level 6 with cross validation accuracy result of 78.25%. The classification tree prediction model shows a complex relation of arsenic and the predictors. However, the root classifiers of the classification tree were found to be the surface elevation factor followed by geology. The best tree obtained from cross validation process was used to create the arsenic risk map in GIS environment. The developed risk map is found to be very close to reality where 87.9% of the data were found to be correctly predicted. The research shows that supported by geochemical process the surface elevation of an area determines the degree of accumulation of arsenic in the groundwater.

► Develop arsenic contamination risk map using classification tree and GIS technology.
► Geology, pH, elevation, organic matter, iron, and proximity to river as predictors.
► Surface elevation and geology are main factors related to arsenic contamination.
► 87.9 percent of the data were correctly predicted.