| Water quality is of paramount importance for human health and agricultural production, influencing also economic outcomes and environmental sustainability. Rivers and their surrounding basins play a central role in transporting and distributing water toward downstream and coastal areas, making them key environments for monitoring water contamination levels. Thus, effective water-quality estimation requires capturing both temporal dynamics and spatial variations along river sides. In this work, we present a prediction approach for river water contamination using satellite-derived environmental data and water body location. Data are used as inputs to train multiple Long Short-Term Memory (LSTM) neural networks with different configurations, allowing us to compare their performance and identify the architecture that best predicts future contamination levels. The results showcased that satellite observations and hydrologic data, when combined with deep learning can provide reliable and scalable tools for monitoring water quality in river systems. The datasets used and produced during the current study are publicly available. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.