Prediction and controlling factors of high-fluoride groundwater in the Yellow River Basin based on machine learning model

Abstract

Long-term consumption of high-fluoride water (F > 1.5 mg/L) has significant negative effects on human health. In the Yellow River Basin of northern China, fluorosis resulting from geogenic groundwater fluoride contamination has been observed in several basins. In this study, the machine learning algorithm regression modeling was employed to predict the distribution of high-fluoride groundwater and potential population at risk using 30337 groundwater samples and 40 relevant environmental factors, with random forest (RF) was identified as the optimal algorithm. The model incorporated various environmental factors, including hydrogeology, climate, soil, topography, and human activities and the model performed well, with the value of AUC of 0.86. The climatic variables were identified as the primary factors influencing the model based on the ranking of their importance. The probability distribution map with a resolution of 250 m drawn from the modelling results shows that high-fluoride groundwater is mainly distributed within the basins, Loess Plateau, the front edge and the southern part of the Yellow-Huai-Hai River Plain (also known as North China Plain). The climate plays a vital role in regulating the distribution patterns of high-fluoride groundwater. Based on different probability cut-off values, it is estimated that approximately 7.307 and 8.899 million people in the study area may be at risk of direct consumption of fluoride-contaminated groundwater. High-fluoride groundwater primarily occurs in shallow pore aquifers of alluvial plains. Fine-grained sediments with high clay content and high levels of cations with exchangeable sites favor the enrichment of fluoride in groundwater. Alluvial and alkaline soils exhibit significant impacts on the enrichment of fluoride. Significant temperature differences and uneven precipitation are the main climatic factors affecting fluoride enrichment in groundwater. This study helps to enhance the understanding of the spatial differentiation and driving mechanism of high-fluoride groundwater, and provides a scientific basis for preventing endemic fluorosis and ensuring water supply security.