Pelagic Sargassum is one type of marine macroalgae that is known to be abundant in the Gulf of Mexico and Sargasso Sea. It is also known to serve as a critical habitat for many marine animals. In the past few years, large amounts of Sargassum have been reported in the Tropical Atlantic and Caribbean Sea (CS), causing significant environmental and economic problems. The goal of this study is to improve the understanding of Sargassum distributions, quantity, transport pathways, and bloom mechanisms in the CS and Tropic Atlantic through combining a variety of techniques including satellite remote sensing, field and laboratory measurements, and numerical modeling.
The first question is where and how much Sargassum is in the CS and Tropic Atlantic. Previous field observations revealed strong seasonal and spatial variations of Sargassum, yet these observations are all limited in their spatial and temporal coverage. Satellite observations offer an effective means to measure their distributions with synoptic coverage and high sampling frequency, yet it is technically challenging to extract and quantify the small Sargassum features in coarse-resolution satellite imagery. Chapter 2 focuses on Sargassum detection and quantification algorithm development using Moderate Resolution Imaging Spectroradiometer (MODIS) data (Appendix A). The algorithm is based on MODIS alternative floating algae index (AFAI), which examines the red-edge reflectance of floating vegetation. The algorithm includes three basic steps: 1) classification of Sargassum-containing pixels through correction of large-scale gradient, masking clouds and cloud shadows, and removal of ambiguous pixels; 2) linear unmixing of Sargassum-containing pixels; and, 3) statistical analysis of Sargassum area coverage in pre-defined grids at monthly, seasonal, and annual intervals. The algorithm is applied to MODIS observations between 2000 and 2015 over the Central West Atlantic (CWA) region (0 – 22oN, 38 – 63oW) to derive the spatial and temporal distribution patterns as well as the total areal coverage of Sargassum. Results indicate that the first widespread Sargassum distribution event occurred in 2011, consistent with previous findings from the Medium Resolution Imaging Spectrometer (MERIS). Since 2011, only 2013 showed minimal Sargassum coverage similar to the period of 2000 to 2010; all other years showed significantly more coverage. More alarmingly, the summer months of 2015 showed mean coverage of > 2000 km2, or about 4 times of the summer 2011 coverage and 20 times of the summer 2000 to 2010 coverage. Analysis of several environmental variables provided some hints on the reasons causing the inter-annual changes after 2010, yet further multi-disciplinary research (including in situ measurements) is required to understand such changes and long-term trends in Sargassum coverage.
To better understand the potential ecological and environmental impacts of Sargassum, field and laboratory experiments are conducted to link the Sargassum areal coverage observations to biomass per area (density) and measure the nutrient contents and pigment concentrations (Chapter 3, Appendix B). An AFAI-biomass density model is established to derive Sargassum biomass density from the spectral reflectance, with a relative uncertainty of ~ 12%. Monthly mean integrated Sargassum biomass in the CS and CWA reached > 4.4 million tons in July 2015. The average % C, % N, and % P per dry-weight, measured from samples collected in Gulf of Mexico and Florida Straits in summer 2017, are 27.16, 1.06, and 0.10, respectively. The mean chlorophyll-a concentration is ~ 0.05% of the dry-weight. With these parameters, the amounts of nutrients and pigments can be estimated directly from remotely-sensed Sargassum biomass. During bloom seasons, Sargassum carbon can account for ~ 18% of the total particulate organic carbon in the upper water column. This chapter provides the first quantitative assessment of the overall Sargassum biomass, nutrients, and pigment abundance from remote-sensing observations, thus helping to quantify their ecological roles and facilitate management decisions.
To investigate the Sargassum transport patterns and potential bloom sources, a Lagrangian particle tracking model is established to track the Sargassum transport driven by surface currents and winds (Chapter 4, Appendix C). The mean Sargassum distributions derived from MODIS observations are used to initiate and evaluate a Lagrangian particle tracking model that tracks Sargassum advection under surface currents and winds. Among the thirty-nine experiments, adding surface currents alone improves model performance (i.e., by reducing difference between modeled and observed Sargassum distributions) in 82% of the cases after tracking Sargassum for one month. Adding 1% wind forcing to the advection model also shows improved performance in 67% of the cases. Adding a time- and location-dependent Sargassum growth/mortality rate (i.e., change rate), derived from time-series of the MODIS-based Sargassum abundance and the corresponding environmental data via a Random Forest regression, leads to further improvement in model performance (i.e., by increasing the matchup percentage between modeled and observed Sargassum distributions) in 64% of the cases, although the modeled change rates only explain ~ 27% of the variance of the validation dataset, possibly due to uncertainties in such-derived change rates. The Sargassum transport model, with the mean currents, winds, and change rates acting as the forcing, is applied to track the mean Sargassum distributions forward and backward. The results demonstrate the model’s capacity of simulating the Sargassum distribution patterns, with emphasis on the role of biological terms in determining the large-scale distributions. These tracking experiments also suggest that Sargassum blooms in the CS are strongly connected to the Central Atlantic regions, and blooms in the Tropical Atlantic show relatively weak connections to the Atlantic regions further north.
Although it is straightforward to apply the transport model to predict Sargassum blooms, such long-term prediction could suffer from large error accumulations and unable to achieve satisfactory performance. Therefore historical Sargassum distributions derived from MODIS are used to provide an alternative way to realize the bloom prediction. Chapter 5 proposes such a prediction based on a hindcast of 2000–2016 observations from MODIS, which shows Sargassum abundance in the CS and the CWA, as well as connectivity between the two regions with time lags (Appendix D). This information is used to derive bloom and nonbloom probability matrices for each 1° square in the CS for the months of May–August, predicted from bloom conditions in a hotspot region in the CWA in February. A suite of standard statistical measures is used to gauge the prediction accuracy, among which the user’s accuracy and kappa statistics show high fidelity of the probability maps in predicting both blooms and nonblooms in the eastern CS with several months of lead time, with an overall accuracy often exceeding 80%. The bloom probability maps from this hindcast analysis will provide early warnings to better study Sargassum blooms and prepare for beaching events near the study region. This approach may also be extendable to many other regions around the world that face similar challenges and opportunities of macroalgal blooms and beaching events. Using this forecasting scheme, the summer blooms in the CS in 2017 were successfully predicted. Since February 2018, we have also generated monthly-updated 1-page Sargassum outlook bulletins to help these regions to better prepare for potential beaching events.
Currently, the mean Sargassum distribution statistics used in this study are derived from MODIS, which has been operating well beyond the designed mission life, arousing concerns as to whether the Sargassum observation statistics can be continued in the future. As a follow-on sensor, the Visible Infrared Imaging Radiometer Suite (VIIRS) has the appropriate spectral bands to detect and quantify floating macroalgae. Based on previous works on MODIS, Chapter 6 presents an improved procedure to extract floating algae pixels from VIIRS AFAI imagery, with image filtering used to suppress noise and adjusted thresholds used to mask sun glint, clouds, and cloud shadows. The overall extraction accuracy is about 85%. Simultaneous daily observations from MODIS and VIIRS over the CWA show consistent spatial patterns, but VIIRS estimations of the algae coverage (in km2) are consistently lower than MODIS (around – 19% mean relative difference or MRD), possibly due to lower sensitivity of the VIIRS near-infrared (NIR) bands than the corresponding MODIS bands. Similarly, at monthly scale VIIRS also shows lower coverage than MODIS, and their difference (around – 29% MRD) is larger than the difference between MODIS-Aqua and MODIS-Terra estimates (around – 14% MRD). Despite these differences, the spatial and temporal patterns between VIIRS and MODIS observed algae distributions match very well at all spatial and temporal scales. These results suggest that VIIRS can provide continuous and consistent observations of floating algae distributions and abundance from MODIS as long as their differences are accounted for, thus assuring continuity in the future.
In summary, this study has worked on four connected topics regarding Sargassum distributions, biomass and nutrients, transport pathways, and bloom predictions through combined efforts in satellite remote sensing, field and laboratory measurements, physical modelling, and statistical analyses. To my best knowledge, this is the first comprehensive and multi-disciplinary study to investigate pelagic Sargassum at synoptic scale in the Intra-Americas Sea (IAS) and Atlantic Ocean. Although several questions remain to be answered (e.g., “What cause the inter-annual variations of Sargassum blooms?” and “Where are the bloom origins?”), the outcomes of this study (remote sensing algorithms, Sargassum distribution and abundance maps, established bio-physical model, and a bloom forecast model) are expected to make significant contributions in both scientific research (including new critical baseline data) and management decision support.
| Identifer | oai:union.ndltd.org:USF/oai:scholarcommons.usf.edu:etd-8913 |
| Date | 03 July 2018 |
| Creators | Wang, Mengqiu |
| Publisher | Scholar Commons |
| Source Sets | University of South Flordia |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Graduate Theses and Dissertations |
| Rights | default |
Page generated in 0.0448 seconds