Coastal environments are subjected to rapid changes induced by global climate variations and human activities, hence requiring an efficient monitoring systems. In this context, autonomous surface and underwater vehicles offer a scalable and high-resolution approach to collecting water quality data. However, the reliability of these systems is often compromised by sensor faults and missing measurements, which can degrade both data integrity and system functionality. This study investigates the application of machine learning methodologies to estimate key water quality parameters in the presence of incomplete or erroneous data. Focusing on Biscayne Bay, Florida, we evaluate the performance of linear regression, random forest, support vector regression, and multilayer perceptron models in predicting dissolved oxygen, pH, and temperature. Initial results indicate the potential of these models to enhance data consistency and offer new perspectives for sensor fusion approaches.
