As an additive in toothpaste, it protects our teeth from decay. But when fluoride occurs in nature in larger quantities and accumulates in groundwater, it can become a hazard for our health. For the first time, Eawag scientists have produced a detailed map of global fluoride contamination in groundwater and shown which regions of the world are particularly affected.
Not every pollutant in man-made. Some occur naturally in rock and thus also in groundwater. For example fluoride, which has a toxic effect when ingested in large quantities and causes bones and joints to degenerate. Besides geology, climate is a decisive factor for fluoride accumulation in groundwater. Concentrations are particularly high in hot, dry regions: On the one hand, because higher temperatures favour weathering and thus the dissolution of fluoride from the rock, and on the other hand, because the fluoride remains in the groundwater for a longer period of time, as this is renewed only very slowly due to the low amounts of precipitation. The assumption is obvious that with climate change and the advancing desertification in many parts of the earth, the fluoride problem could also become more acute.
To make matters worse, fluoride often goes undetected because it is odourless and invisible. Only water analyses provide information about excessive concentrations. In many countries of the global south, however, groundwater is hardly tested, and quite a few people get their water directly from a groundwater pump. The risk areas are therefore not widely known, and there are many gaps,” says Joel Podgorski, who is doing research on water resources and drinking water at Eawag. “We wanted to fill these gaps with our work to create a basis for better groundwater monitoring.”
Machine-Learning closes gaps in knowledge
Joel Podgorski and his colleague Michael Berg have developed a model based on machine learning. Based on certain soil properties as well as topographical, geological and climatic factors, this model calculates the probability of the fluoride content being above the health-relevant threshold limits of 1.5 milligrams per litre. A prerequisite for the model to function is sufficient measurement data to train the computer algorithm. “Our first trial a few years ago failed because we had too little measurement data for a reliable model,” says Podgorski.. In the meantime, however, more and more countries are making their data publicly accessible. That is why the two researchers were able to create a database of 400,000 fluoride measurements in groundwater - the largest global dataset to date.