A regional framework for spatio-temporal assessment of lake eutrophication using Sentinel-2/MSI imagery

In the context of climate change and increasing water scarcity, lakes provide essential services, such as irrigation or low-flow maintenance. However, they face eutrophication risk, threatening water quality and biodiversity in both lakes and downstream ecosystems. As traditional field monitoring concerns only 1 % of French lakes, remote sensing offers an effective alternative to assess lake eutrophication. Yet, its regional scale application is limited to large lakes and lacks standardized eutrophication assessment methodology. This study proposes a generalizable framework using Sentinel-2/MSI imagery to assess lake eutrophication at a large scale, including small lakes (>1 ha). The methodology involves 1) a missing data imputation scheme to gap-fill continuous monthly chlorophyll-a and turbidity time series (2018–2023) for small and data-scarce lakes using MissForest method with meteorological and land use predictors, 2) a satellite-derived trophic state linear model comparable to Carlson's index using exclusively satellite-derived Chla and turbidity and 3) eutrophication spatio-temporal trends and drivers analyses. Applied to the Adour-Garonne basin (Southwestern France; 118,000 km 2 ) from 2018 to 2023, the method achieved high accuracy for trophic state retrieval (r 2 = 0.7, n = 487). It allowed the eutrophication assessment of nearly 2500 unmonitored lakes, showing that >85 % of the lakes are eutrophic to hypereutrophic, with a mean TSI of 58 and stable trajectories since 2018. Higher TSI levels were associated with small lakes, low altitudes and high proportions of agricultural land uses such as corn. Results are provided as continuous and categorical indicators to support the identification of priority lakes for targeted monitoring and management. • A Sentinel-2-based framework was developed to monitor small, data-scarce lakes (2018–2023). • Chla, turbidity and derived Trophic State Index (TSI) were assessed in 2468 lakes from 1 ha. • Continuous monthly quality indicators were reconstructed using MissForest model. • 85 % of Adour-Garonne lakes were eutrophic to hypereutrophic during the monitoring period. • Environmental drivers' study showed higher TSI levels in small lakes and agricultural settings.