Coastal environments, such as beaches, canals, and estuaries, are ecologically vital, supporting diverse species and regulating the interface between continental and oceanic systems. Despite their importance, these ecosystems face growing threats from anthropogenic activities that compromise their environmental balance. In this context, data-driven approaches such as machine learning offer promising tools for monitoring and managing water quality. However, many existing studies focus on limited indicators and often overlook temporal and spatial dynamics. This study addresses these gaps by applying machine-learning models to assess water quality in the Jacarepaguá Lagoon System, a complex of four interconnected lagoons in Rio de Janeiro, Brazil. While the analysis centered on biochemical oxygen demand (BOD), future work will incorporate additional indicators such as turbidity and dissolved oxygen, and adopt temporal modeling strategies. The models developed achieved a mean squared error of 17(mg/L) \(^2\) , constrained by the sparsity of monitoring data. These findings underscore both the potential and limitations of current predictive approaches and highlight the need for more comprehensive and continuous data collection to support evidence-based environmental management.