To date, 13 potential HAB species have been detected in coastal waters of Mississippi and Alabama, including representatives of the diatom genera Pseudo-nitzschia and Chaetoceros, and dinoflagellate genera Karenia, Karlodinium, Gonyaulax, and Prorocentrum. This study investigates the potential of satellite remote sensing (SeaWiFS) to predict environmental conditions leading to the formation of HABs in these turbid coastal waters. Phytoplankton populations and water quality were monitored in situ at 3 to 6 week intervals and 17 locations in Mobile Bay and the Mississippi Sound from July, 2005 through June, 2006. SeaWiFS imagery corresponding with in situ collections was acquired. Non-parametric multivariate analyses determined relationships between phytoplankton cell counts and satellite-derived water properties, including surface temperature, concentrations of chlorophyll-a, total suspended solids, colored dissolved organic material and remote sensing reflectance. This paper will describe an expert system decision tree analysis approach to prediction of ecological conditions necessary for the formation of HABs, using recorded cell counts from a post-Hurricane Katrina HAB event of Pseudo-nitzschia sp. in Mobile Bay and the Mississippi Sound as a testing framework. The model assumes unique ranges of remote sensing reflectance, Chla, TSS, and sea surface temperature must exist in order to allow the formation of HABs.
Results/Conclusions
Phytoplankton population surveys showed high diversity, with samples often being comprised of tens of genera and more than 100 species. Collections represented 13 phyla and 79 genera. However, some species were observed at notably higher numbers. Post-Hurricane Katrina Pseudo-nitzschia sp. bloom collections in May of 2005 were chosen for testing due to high numbers of cell counts per unique water sample (>1 X 106 cells/liter) and corresponding clear skies for sensor data retrieval. SeaWiFS-derived values exhibited poor linear relationships when directly compared with in situ measurements, expected in turbid near-shore waters of the northern Gulf of Mexico. However, peak population counts of Pseudo-nitzschia sp. exhibited Gaussian relationships when comparing cell counts and ocean color measurements such as Chla, TSS, and remote sensing reflection at 412, 443, 555 and 670nm wavelengths. Ratios of these values also showed narrow range values correlated with high population cell counts. MODIS Aqua sea surface temperature data was also used as input to the analysis, due to similar relationships with cell count data. In back testing accuracy of the decision tree analysis with in situ collections, a 78% accuracy of recorded in situ collection sites to decision tree predicted sites was achieved.
