Monitoring and quantifying of water turbidity in the Albufera of Valencia during the 2024 flood using a machine learning model

Monitoring turbidity during flood events is vital for assessing the extent of water contamination, understanding sediment transport dynamics, and guiding emergency response efforts to mitigate ecological damage and health risks. In this study, we aim to evaluate the application of a global turbidity model, developed using Sentinel-2 satellite data and machine-learning techniques, to map and monitor turbidity during the October 2024 flood event in Valencia, Spain. The model has been developed based on a comprehensive dataset including hyperspectral in situ measurements, turbidity data, and Sentinel-2 observations, enabling robust estimation across a wide range of turbidity levels (0-2500 FNU). The gradient boosting model used outperformed other techniques, achieving high accuracy in turbidity prediction. The model has been validated across diverse optical water types and water bodies (rivers, lakes, estuaries, and coastal oceans), demonstrating its versatility and scalability. The proposed application of this global model to the Valencia flood provides valuable insights into the spatial extent and severity of turbidity variations, sediment transport, and water quality degradation during the event. This study highlights the potential of combining satellite observations with advanced machine-learning models for real-time environmental monitoring, offering critical tools for disaster response and water resource management.