Phytoplankton sustains marine ecosystems and influences the global carbon cycle. This study analyzes trends in surface chlorophyll‐ a concentration (Schl), a proxy for phytoplankton biomass, using six of the most widely used merged satellite products. Significant regional variations are observed, with contrasting trends observed among different products. To assess if these trends can be related to changes in the environment or to bias in radiometric products, a convolutional neural network is used to examine the relationship between physical ocean variables versus Schl. The results suggest that the merging algorithm of Globcolour‐GSM is not reliable for trend detection and that observed changes in Schl after 2012/2016 are not supported by changes of the physical ocean. These results emphasize the need for careful interpretation of satellite‐derived Schl trends and highlight the potential of machine learning in understanding the complex interactions in marine ecosystems.

Small lakes (<10 km²) provide a range of ecosystem services but are often overlooked in both monitoring efforts and limnological studies. Remote sensing has been increasingly used to complement in situ monitoring or to provide water colour data for unmonitored inland water bodies. However, due to spatial, radiometric, and spectral constraints, it has been heavily focused on large lakes. Sentinel-2 MSI is the first sensor with the capability to consistently retrieve a wide range of essential water quality variables, such as chlorophyll-a concentration (chl-a) and water transparency, in small water bodies, and to provide long time series. Here, we provide and validate a framework for retrieving two variables, chl-a and turbidity, over lakes with diverse optical characteristics using Sentinel-2 imagery. It is based on GRS for atmospheric and sun glint correction, WaterDetect for water detection, and inversion models that were automatically selected based on two different sets of optical water types (OWTs)—one for each variable; for chl-a, we produced a blended product for improved spatial representation. To validate the approach, we compared the products with more than 600 in situ data from 108 lakes located in the Adour–Garonne river basins, ranging from 3 to ∼5000 ha, as well as remote sensing reflectance (Rrs) data collected during 10 field campaigns during the summer and spring seasons. Rrs retrieval (n = 65) was robust for bands 2 to 5, with MAPE varying from 15 to 32% and achieving correlation from 0.74 up to 0.92. For bands 6 to 8A, the Rrs retrieval was much less accurate, being influenced by adjacency effects. Glint removal significantly enhanced Rrs accuracy, with RMSE improving from 0.0067 to 0.0021 sr⁻¹ for band 4, for example. Water quality retrieval showed consistent results, with an MAPE of 56%, an RMSE of 11.4 mg m⁻³, and an r of 0.76 for chl-a, and an MAPE of 47%, an RMSE of 9.7 NTU, and an r of 0.87 for turbidity, and no significant effect of lake area or lake depth on retrieval errors. The temporal and spatial representations of the selected parameters were also shown to be consistent, demonstrating that the framework is robust and can be applied over lakes as small as 3 ha. The validated methods can be applied to retrieve time series of chl-a and turbidity starting from 2016 and with a frequency of up to 5 days, largely expanding the database collected by water agencies. This dataset will be extremely useful for studying the dynamics of these small freshwater ecosystems.

Les séries temporelles de concentrations en Chlorophylle-a (Chl, un proxy de la biomasse phytoplanctonique) issues des observations satellites restant trop courtes pour étudier la variabilité à basse fréquence (ex : décennale) de la biomasse phytoplanctonique, cette thèse visait à explorer le potentiel des méthodes émergentes d’apprentissages profonds pour reconstruire des signaux passés de la Chl à partir de prédicteurs océaniques et atmosphériques, dans l’océan global. Dans un premier temps, est démontrée la pertinence des réseaux de neurones convolutifs (CNNs) pour prendre en compte la dépendance de la Chl aux conditions physiques environnementales. Dans un second temps, l’introduction de mécanismes d’attention a permis d’adapter ces modèles convolutifs classiques à la forte hétérogénéité spatio-temporelle des liens physique-biogéochimie observés à l'échelle globale. Ainsi, les processus régionaux sont mieux capturés par le modèle, qui permet également d’extraire plus facilement de nouvelles informations concernant les processus physiques sous-jacents. Enfin, le modèle statistique développé a été utilisé pour documenter l’impact de la prise en compte d’une source atmosphérique de nutriments (i.e. dépôt de poussières désertiques) caractérisée notamment par une variabilité spatio-temporelle différente des sources internes à l’océan. Les résultats montrent que la climatologie annuelle de ces dépôts semble améliorer la perception des motifs globaux de (co-)limitations en nutriments par le modèle statistique. De plus, des réponses contrastées aux variabilités saisonnières et interannuelles de ces dépôts sont observées selon les bassins océaniques.

GET-Pak is a standalone lightweight framework for reproducible generation and analysis of pixel-based inland water quality (WQ) maps and timeseries generation from atmospheric-corrected Sentinel-2 MSI images resulting from processors ACOLITE, GRS or SeaDAS. It supports command-line batch processing as well as interactive use as a Python library in Jupyter notebook environment. GET-Pak produces different optical water-type classifications (OWT) and maps of Chlorophyll-a (Chl-a), Suspended Particulate Matter (SPM) and Turbidity using validation algorithms published in the literature, with automated reporting over multiple regions of interest (ROIs). In this work we demonstrate resulting WQ maps and timeseries over the Madeira River basin, Brazil, and Liangzi Lake, China, running on both a HPC cluster, and an end-user machine at an average of 270 seconds per scene.