The intoDBP project is excited to announce the publication of new research in Biogeosciences that marks a significant step forward in our ability to monitor and protect drinking water sources.
Predicting the Invisible
Dissolved Organic Matter (DOM) is a fundamental component of aquatic ecosystems, yet monitoring it in real-time can be challenging. The research team developed a machine learning workflow—specifically utilizing the CatBoost algorithm—to predict fDOM, an important proxy for carbon dynamics and water quality.
From Ireland to Spain
The study looked at two very different environments: the peat-dominated catchment of Lough Feeagh, Ireland, and the Mediterranean, human-influenced Sau Reservoir in Spain. The findings highlight that while soil temperature and moisture are universal drivers, their influence changes significantly depending on the local landscape.
Scalability: A Global Solution
Perhaps most importantly, the study proved that these models remain highly effective even when using globally accessible ERA5 reanalysis data. This suggests that the workflow can be scaled to sites around the world that lack expensive local monitoring equipment, providing a vital tool for global water management.
