Abstract: For about the past eight decades, high concentrations of naturally occurring fluoride have been detected in groundwater in different parts of India. The chronic consumption of fluoride in high concentrations is recognized to cause dental and skeletal fluorosis. We have used the random forest machine-learning algorithm to model a dataset of 12,600 groundwater fluoride concentrations from throughout India along with spatially continuous predictor variables of predominantly geology, climate and soil parameters. Despite only surface parameters being available to describe a subsurface phenomenon, this has produced a highly accurate prediction map of fluoride concentrations exceeding 1.5 mg/L at 1 km resolution throughout the country. The most affected areas are the northwestern states/territories of Delhi, Gujarat, Haryana, Punjab and Rajasthan and the southern states of Andhra Pradesh, Karnataka, Tamil Nadu and Telangana. The total number of people at risk of fluorosis due to fluoride in groundwater is predicted to be around 120 million, or 9% of the population. This number is based on rural populations and accounts for average rates of groundwater consumption from non-managed sources. The new fluoride hazard and risk maps can be used by authorities in conjunction with detailed groundwater utilization information to prioritize areas in need of mitigation measures.

*Abstract online at https://www.ncbi.nlm.nih.gov/pubmed/30052029