Machine Learning and Citizen Science for Catchment-Scale Water-Quality Assessment

Assessing and managing water quality in data-scarce tropical basins remains challenging due to rapid land-use change, intensifying human pressures, and increasing climate variability. In the transboundary Mara River Basin (MRB), these factors strongly influence sediment and nutrient dynamics, yet traditional monitoring networks lack the temporal and spatial resolution needed to characterize pollution sources, transport pathways, and event-driven responses. To address these gaps, this study integrates citizen-science observations with satellite-derived hydro-climatic and land-use variables to model turbidity (NTU), nitrate (NO₃), and phosphate (PO₄) across 40 sub-basins. Two machine-learning approaches: Random Forests (RF) and Artificial Neural Networks (ANN) were employed to evaluate water-quality variability and identify dominant drivers under heterogeneous environmental conditions. RF outperformed ANN across all indicators, providing more robust predictions under noisy and nonlinear data constraints. SHAP analyses revealed that precipitation and river flow velocity dominate short-term, event-based fluctuations of turbidity, while population density represents persistent drivers of NO₃ concentration. These findings highlight the basin’s sensitivity to climate-driven changes in rainfall intensity and seasonality and demonstrate how hybrid monitoring–modelling frameworks can enhance the identification of nutrient hotspots, improve source attribution, and support adaptive water-quality management under land-use and climate-change scenarios.