Sargassum in the Florida Keys: Assessment using High Resolution Remote Sensing

Graduation Year

2024

Document Type

Thesis

Degree

M.S.

Degree Name

Master of Science (M.S.)

Degree Granting Department

Marine Science

Major Professor

Chuanmin Hu, Ph.D.

Committee Member

Brian Barnes, Ph.D.

Committee Member

Lin Qi, Ph.D.

Keywords

deep learning, Landsat 8, ResUNet, Sentinel 2, SST anomalies

Abstract

Pelagic Sargassum in the Atlantic Ocean, one type of floating macroalgae, has increased in biomass and extent since 2011, bringing environmental, health, and economic issues to coastal communities. The introduction of large quantities of Sargassum to the Florida Keys, a tourism hub renowned for coral reef and marine diversity, poses a potentially devastating threat. Currently, Sargassum mats are widely monitored using medium resolution satellite imagery such as those collected by the Moderate Resolution Imaging Spectroradiometer (MODIS, 1km) and spectral band difference algorithms such as the Alternative Floating Algae Index (AFAI), but the resulting maps lack the detail and nearshore certainty necessary for Sargassum detections in the optically complex study region along the Florida Keys. To overcome these limitations, high-resolution data collected by the Sentinel-2 Multi-Spectral Imagers (MSI, 10-20 m resolution but binned to 50-m resolution to increase signal-to-noise ratio, 2015-2023) was used together with a Res-UNet deep learning model for Sargassum feature extraction. After pixel unmixing, the model had an F1 score of 84% calculated over 7,796 Sargassum pixels and 2,147,840 non-Sargassum pixels. The extraction results were binned to provide monthly Sargassum density maps between 2015 and 2023. Over the common valid area, MSI provided Sargassum detections comparable to those of MODIS. However, in the nearshore environments (within 10 km of shore) where MODIS detection is unreliable, the MSI Res-UNet model offers valuable Sargassum detections, effectively distinguishing features in optically complex waters. Sargassum detections within the MODIS 10-km nearshore masked region account for 15.7% of MSI detections, equating to 0.08 km2 Sargassum daily.

Further, this study assessed satellite-based sea surface temperature (SST) anomalies over Sargassum mats. Landsat 8 and 9 Thermal Infrared Sensor (TIRS) high resolution (100 m) STB10 data were used with the corresponding Floating Algae Index (FAI) images to quantify temperature differences between Sargassum mats and their surrounding waters. A control experiment quantified the daily natural SST variability, with temperature changes outside the natural range being attributed to Sargassum presence. Those anomalous changes were observed in up to 21% of all Sargassum mats, with temperature differences ranging from -0.24°C to 0.99°C. Increased temperatures over Sargassum mats as compared to the surrounding water were present in 95% of all anomalies. SST differentials were found to increase with Sargassum mat density, with 90% of mats with over 45% Sargassum density showing SST changes outside of the natural range of variability.

The Sargassum extraction findings of this study support implementation of an automatic monitoring system to provide high-resolution bloom tracking whenever the satellite data is available. Such a system is useful for Sargassum management and supports ecological studies (e.g., juvenile sea turtle release). Additionally, the temperature findings offer valuable insight into the extent and seasonality of Sargassum-induced SST anomalies. These discoveries enhance understanding of Sargassum’s impact on local ecosystems and biodiversity, and they provide insight to associated species behavior in the face of global climate change.

This document is currently not available here.

Share

COinS