Related Experiment Videos
Predicting and assessing the distribution of high-fluoride groundwater in China using multiple machine learning
Mengyao Su1, Qingbo Wang1, Yunzhu Liu1
1Center for Endemic Disease Control, Chinese Center for Disease Control and Prevention, Harbin Medical University, Harbin 150086, PR China; NHC Key Laboratory of Etiology and Epidemiology, Harbin Medical University, Harbin 150086, PR China; Joint Key Laboratory of Endemic Diseases, Harbin Medical University, Guizhou Medical University, Xi'an Jiaotong University, PR China.
Ecotoxicology and Environmental Safety
|August 6, 2026
Summary
Drinking-water fluorosis is a major concern in China, affecting millions. This study uses advanced AI models and environmental data to map high-risk areas, identifying new hotspots and informing public health strategies for safer groundwater.
Area of Science:
- Environmental Science
- Public Health
- Geochemistry
Background:
- Drinking-water fluorosis is a significant public health issue in rural China, primarily linked to high-fluoride groundwater.
- The reliance on specific groundwater sources exacerbates the risk in vulnerable populations.
Purpose of the Study:
- To develop and compare predictive models for groundwater fluoride contamination risk in China.
- To identify high-risk zones and estimate the affected rural population.
- To understand the key environmental drivers of fluoride enrichment.
Main Methods:
- Utilized the largest harmonized national dataset (138,180 site records) with 38 environmental predictors.
- Developed and evaluated Artificial Neural Network (ANN), Random Forest (RF), and Logistic Regression (LR) models.
- Employed spatial block 5-fold cross-validation to ensure robust geographic generalization.
Main Results:
- All models showed high spatial generalizability, with PR_AUC values up to 0.992 (RF).
- Probabilistic risk maps identified known and new high-fluoride zones across China, including Northeast Plain, North China Plain, Inner Mongolia, and parts of northwest China.
- An estimated 40.78-54.77 million rural residents live in predicted high-risk areas.
Conclusions:
- Meteorological, topographic-soil, and volcanic-geologic factors are dominant controls on fluoride enrichment.
- The comparative modeling framework offers a validated, scalable tool for national fluorosis risk assessment.
- Findings provide actionable insights for targeted surveillance, mitigation, and groundwater management, applicable to similar regions globally.
