The Sargassum is an invasive species of algae that aggregates and drifts in the open ocean. Since the last decade, Sargassum is observed at unusually high quantities from the Caribbean Sea to Brazil up to the coast of Northwest Africa. Remote sensing is a powerful mean to detect these brown algae in the open ocean and estimate their abundance. The Maximum Chlorophyll Index (MCI) is able to detect Sargassum at the sea surface using their specific red and NIR reflectances. However, the Caribbean Sea, rough waters and strong winds (>4m.s-1) can often immerse the aggregations to more than 1 meter deep. However, water strongly absorbs in the red and NIR domain, then the MCI could fail to detect the immersed Sargassum aggregations. One previous study (Descloitres et al 2021) adapted the semi-analytical model developed by Lee et al. (Appl. Opt., 1999) to estimate the abundance and the depth of the Sargassum aggregation from the surface reflectance. A neural network was trained here using simulated data based on the physical model, to estimate parameters (coverage and depth) of the Sargassum aggregations from Sentinel-3/OLCI surface derived reflectances. The neural network enables to analyze large datasets of Sentinel-3/OLCI scenes for a short computing time. In this study, more than 300 Sentinel-3/OLCI scenes centered in the Caribbean Sea were analyzed using the neural network, to determine the Sargassum coverage and the depth of the aggregations. The work focuses on the year 2021 for which the Caribbean Sea was strongly impacted by Sargassum strandings. Satellite data will enable to better understand the spatial distribution and the seasonal cycle of the immersed Sargassum aggregation in the Caribbean waters. The purpose of the study is also to identify existing patterns and factors leading to the immersion of Sargassum. As an example, a correlation was found between the wind speed (> 4m s-1) and the amount of immersed Sargassum aggregation. The temporal variability of the immersed Sargassum aggregation would improve the stranding forecast.