Aplicación de técnicas de aprendizaje automático a la monitorización de la calidad de masas de agua mediante técnicas de teledetección

Abstract

This thesis addresses the problem of monitoring the quality of continental water masses and the eutrophication episodes through space remote sensing data published openly by the European Space Agency (ESA) and NASA (National Aeronautics and Space Administration) for their Sentinel-2 and Landsat 8 space missions. In this work, recent automatic deep learning techniques are introduced, both in the preprocessing of the images and the inference of the water quality variables to be monitored. Thanks to these techniques, a new method is successfully proposed to monitor the temperature and chlorophyll concentration of a eutrophic reservoir, estimating both variables reliably both temporally and spatially. The final result is an automatic system for monitoring the trophic state of freshwater masses through the concentration of chlorophyll-a, which only uses open data provided by the main operational satellites.