Estimation of Chl-a in highly anthropized environments using machine learning and remote sensing

Coastal lagoons are ecosystems of great socioeconomic and environmental value. However, they are subject to great anthropogenic and environmental pressures, mainly due to climate change, which threatens their sustainability. High-resolution spatial and temporal monitoring systems are mandatory to (1) identify these threats, (2) understand the main problems affecting these ecosystems, and (3) predict how these ecosystems will behave in the future. In this paper, we present a monitoring system based on the European remote sensing service Copernicus that allows daily monitoring of chlorophyll-a (Chl-a) for the Mar Menor lagoon (Southeast Spain). Moreover, several machine learning (ML) models are analyzed to adapt the collected data to the particular context of the shallow and highly saline Mar Menor. The accuracy of the models are satisfactory, obtaining a global model with 0.9 value of R2 and 0.75 mg/m3 of mean absolute error. Also, this model is able to describe the algal bloom that provoke Chl-a peaks concentrations.