Abstract. This datapaper presents the methodology and the associated dataset of Sargassum detection from Sentinel-3 OLCI images developed at Météo-France. The dataset includes daily detections over the whole tropical Atlantic Ocean, the Caribbean Sea and the Gulf of Mexico, from May 2018 up to date. It is updated daily with the latest detections. Sargassum is detected through a physically-based approach relying on two indices (the AFAI and the MCI) and using adaptive thresholds. A comparison of this dataset with two existing ones reveals a reduction in masked areas and improved detection of Sargassum, particularly in areas of sunglint and at the edges of clouds. The entire dataset is freely available on the Odatis platform and can be downloaded in NetCDF format (DOI: 10.12770/1eb82d09-77ed-4f63-9f03-2c3516a9713d; Météo-France, 2026). Two variables are available: the status of detection and the deviation of the AFAI for pixels including Sargassum.
Received: 23 Jul 2026 – Discussion started: 24 Aug 2026
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
The authors present an open-access dataset of floating Sargassum observed using OLCI daily imagery from May 2018 to present. The underlying approach combines the AFAI and MCI, two indices that each may exhibit degraded performance under different water or atmospheric conditions. By requiring agreement between both indices, the algorithm improves the robustness of Sargassum retrievals to, e.g., cloud edge effects and turbidity, while allowing the authors to increase the observable area through using adaptive masking thresholds, which reduce data losses due to low-to-moderate glint, and through stricter detection thresholds near clouds, which allow detection in cloud-adjacent areas that would otherwise require masking.
The authors report larger detectable area and higher detection rates than existing datasets, with visual inspection suggesting this expanded detection does not confer a discernible increase in false detections. The challenge, which applies to this dataset and to the existing ones, is that ground truth data products to support quantitative validation of satellite retrievals are sparse. The sparsity of in situ data is not an argument against publishing this or other Sargassum maps. Rather, publishing open-access datasets produced with sound methods will support future validation efforts and also aid in the design of field sampling campaigns.
In my reading, this is a valuable contribution to the field. The reasoning behind the approach is well justified and logical, and the open-access dataset will likely become a useful asset to the community. Publication of this dataset report will help improve the use and documentation of this dataset.
--- Minor comments: L155: Removing aggregations smaller than 10 pixels likely reduces false positives, but also removes small but real aggregations. Is the trade-off motivated by a higher likelihood of false positives for small aggregations, or by challenges in verifying these smaller aggregations via visual inspection of the final product? Perhaps small aggregations might be retained with a flag, etc., as they could be of interest to some users. L162: For cases where values are averaged, how are the status of detection values handled? L187: Should "20000" also be: "120000"? Figure 4 Caption: "...(d) MCI near costs", I presume costs should be coasts. L272: Perhaps "intra-sensor" should be "inter-sensor".
This article presents a dataset of Sargassum detections and the methodology used to obtain it. The detections were made using images from the OLCI sensor aboard the Sentinel-3 satellite. They are freely available for the entire tropical Atlantic, the Caribbean Sea, and the Gulf of Mexico, covering the period from May 2018 to the present, with the latest detections added daily. These detections are intended for use in both operational and research applications.
This article presents a dataset of Sargassum detections and the methodology used to obtain it....
The authors present an open-access dataset of floating Sargassum observed using OLCI daily imagery from May 2018 to present. The underlying approach combines the AFAI and MCI, two indices that each may exhibit degraded performance under different water or atmospheric conditions. By requiring agreement between both indices, the algorithm improves the robustness of Sargassum retrievals to, e.g., cloud edge effects and turbidity, while allowing the authors to increase the observable area through using adaptive masking thresholds, which reduce data losses due to low-to-moderate glint, and through stricter detection thresholds near clouds, which allow detection in cloud-adjacent areas that would otherwise require masking.
The authors report larger detectable area and higher detection rates than existing datasets, with visual inspection suggesting this expanded detection does not confer a discernible increase in false detections. The challenge, which applies to this dataset and to the existing ones, is that ground truth data products to support quantitative validation of satellite retrievals are sparse. The sparsity of in situ data is not an argument against publishing this or other Sargassum maps. Rather, publishing open-access datasets produced with sound methods will support future validation efforts and also aid in the design of field sampling campaigns.
In my reading, this is a valuable contribution to the field. The reasoning behind the approach is well justified and logical, and the open-access dataset will likely become a useful asset to the community. Publication of this dataset report will help improve the use and documentation of this dataset.
--- Minor comments:
L155: Removing aggregations smaller than 10 pixels likely reduces false positives, but also removes small but real aggregations. Is the trade-off motivated by a higher likelihood of false positives for small aggregations, or by challenges in verifying these smaller aggregations via visual inspection of the final product? Perhaps small aggregations might be retained with a flag, etc., as they could be of interest to some users.
L162: For cases where values are averaged, how are the status of detection values handled?
L187: Should "20000" also be: "120000"?
Figure 4 Caption: "...(d) MCI near costs", I presume costs should be coasts.
L272: Perhaps "intra-sensor" should be "inter-sensor".