Abstract:

Massive sargassum influxes into the Wider Caribbean Region and West African coast have negatively affected both social and ecological systems since 2011. Current monitoring efforts using satellite data are being conducted but are mainly limited to offshore waters. This research attempts to address the literature gap by developing a method to characterize sargassum accumulations along the shoreline and nearshore waters, and to assess their spatial and temporal dynamics. Using the online Google Earth Engine platform, we analyzed Sentinel-2 MultiSpectral Instrument (MSI) satellite imagery for sargassum occurrence from 2015-09-16 to 2022-01-22 in La Parguera, Puerto Rico. A combination of MSI reflectance bands and several vegetation and water quality indexes were used with a Random Forest classification algorithm. Field data was collected to calibrate and validate the classification product. Our classification model was able to identify different stages of the sargassum decaying process in the shoreline (e.g., fresh sargassum, decomposing sargassum, and sargassum brown tide) along with other non-sargassum cover classes (e.g., water, mangroves, and clouds). Sargassum accumulation hotspots (SAHs) that persisted throughout the study period were identified and their spatial and temporal dynamics were assessed. The data package consists of the following datasets: - LeonPerez.et.al_TrainingAndValidation.shp: Shapefile showing the training and validation data used. - LeonPerez.et.al_CoverClassPersistence.tiff: Raster showing the cover classes that persisted for each pixel throughout the timeseries. - LeonPerez.et.al_SAHLocation.shp: Shapefile showing the location of the three SAHs analyzed. - LeonPerez.et.al_SAHTimeseries_IslaCueva.txt: Timeseries of the area covered by each of the six cover classes in Isla Cueva SAH. - LeonPerez.et.al_SAHTimeseries_IslaGuayacan.txt: Timeseries of the area covered by each of the six cover classes in Isla Guayacán SAH. - LeonPerez.et.al_SAHTimeseries_LaPitahaya.txt: Timeseries of the area covered by each of the six cover classes in La Pitahaya SAH.

Suggested Citation:

León-Pérez, Mariana C., Anthony S. Reisinger, and James Gibeaut. 2022. Pelagic sargassum accumulations along the shoreline and nearshore waters in La Parguera, Puerto Rico from 2015-09-16 to 2022-01-22. Distributed by: GRIIDC, Harte Research Institute, Texas A&M University–Corpus Christi. doi:10.7266/gw5dnwpd

Publications:

León-Pérez, M. C., Reisinger, A. S., & Gibeaut, J. C. (2023). Spatial-temporal dynamics of decaying stages of pelagic Sargassum spp. along shorelines in Puerto Rico using Google Earth Engine. Marine Pollution Bulletin, 188, 114715. https://doi.org/10.1016/j.marpolbul.2023.114715

Purpose:

To develop a method of detecting fresh and decaying sargassum using satellite data and to temporally and spatially characterize sargassum occurrences.

Data Parameters and Units:

Training and Validation (.shp): -date: Indicates the date of the Sentinel-2 scene used for digitizing polygons (format: month, day, year) -use: Indicates whether the polygon was used for training or for validating the model [training, validation] -class_: Cover class representing each polygon [fresh sargassum, decomposing sargassum, sargassum brown tide, water, mangroves, cloud] -source: Indicates whether the polygon represents data collected in-situ or visually interpreted data [collected in-situ, interpreted from imagery] Cover Class Persistence (.tiff): -Count: Number of pixels for each cover class -class_name: Indicates the most frequent (mode) cover class observed in the timeseries for each pixel [fresh sargassum, decomposing sargassum, sargassum brown tide, water, mangroves, cloud] SAH Location (.shp): -SAH: Name of the SAH [Isla Cueva, Isla Guayacán, La Pitahaya] SAH Timeseries (one file per SAH) (.txt): -date: Indicates the acquisition date of the classified Sentinel-2 scene (format: year, month, day) -fresh sargassum: Area covered by fresh sargassum within the SAH [m2] -decomposing sargassum: Area covered by decomposing sargassum within the SAH [m2] -sargassum brown tide: Area covered by sargassum brown tide within the SAH [m2] -water: Area covered by water within the SAH [m2] -mangroves: Area covered by mangroves within the SAH [m2]. Note: Some landward mangrove forest was masked, therefore this value does not represent total mangrove area within the SAH. -clouds: Area covered by clouds within the SAH [m2]

Methods:

Sentinel-2 MSI scenes (Level 1-C) from 2015-09-16 to 2022-01-22 from La Parguera, Puerto Rico, were preprocessed within Google Earth Engine (GEE) to remove clouds and land. A Random Forest (RF) supervised classification model was created combining Sentinel-2 MSI bands, and vegetation, water, and water quality indices, to classify six different cover classes. Training and Validation Dataset: Data for training and validation came from field collected data (from 2019 to 2022) as well as from the visual interpretation of Sentinel-2 MSI scenes. Cover Class Persistence Dataset: Once all Sentinel-2 scenes were classified by the RF model, the mode was calculated to determine the most frequent cover class for each pixel in the timeseries. SAH Location Dataset: The mode calculation was used to identify areas where fresh sargassum and sargassum brown tide classes coincided along the shoreline. Those areas were manually delineated and labeled as SAHs. SAH Timeseries Dataset: The area covered by each class within each SAH (e.g., Isla Cueva, Isla Guayacán, and La Pitahaya) was exported from GEE for each classified image in the image collection.

Error Analysis:

An accuracy assessment was conducted using the validation dataset. Two error matrices were created, one for the field collected data (overall accuracy = 97%) and another for the visually interpreted data (overall accuracy = 99%).

Individual Files: