Paper No. 40-4
Presentation Time: 2:20 PM
PREDICTING THE DISTRIBUTION OF GROUNDWATER ARSENIC ACROSS INDIA USING ARTIFICIAL INTELLIGENCE: INSIGHTS FROM THE TECTONICS, HYDROLOGIC AND ANTHROPOGENIC INFLUENCING FACTORS
Occurrence of elevated groundwater arsenic concentration is a persistent problem in India and puts a large population at risk of severe arsenic-poisoning. The present study aims to develop a comprehensive understanding of the distribution of Arsenic in groundwater and delineate the factors that influence such distribution in India. For this purpose, we have compiled a large dataset of groundwater arsenic concentrations from about 2.7 million drinking water wells, along with geologic, hydrogeologic and anthropogenic factors that may have a significant influence on arsenic distribution. We developed a random forest algorithm based machine learning model to predict the high-resolution spatial occurrence of arsenic above 10µg/L in the country. The prediction model yielded accuracy values of 82% and 94% on the test and entire dataset, respectively, suggesting high accuracy. The risk map of elevated arsenic at 1 km2 generated from model-predicted probabilities indicated that about 20% of the study area in India is exposed to elevated ground arsenic. The model identified that tectonics, elevation (topography) and groundwater irrigation were the most critical factors influencing the occurrence of arsenic. Three major zones were identified, i) the Himalayan-source Indus-Ganges-Brahmaputra mega river basin, ii) Central Indian provinces, hypothesized to be linked with the Central Indian Tectonic Zone and iii) the chalcophilic gold deposit areas of Southern India. Thus tectonic framework seems to be a dominant influencing factor in regional-scale arsenic distribution, which is otherwise not visible in local-scale studies. The estimations indicate that about 268 million people are vulnerable to high groundwater Arsenic and Bihar, West Bengal, Uttar Pradesh and Assam are the states with highest vulnerable populations in India. This study would lead to developing an appropriate method for the delineation of groundwater arsenic in other parts of the world and formulations of groundwater management strategies.