Comparison of Techniques for Satellite Detection of Red Tides in the Gulf of Mexico

Toxic dinoflagellate Karenia brevis (K. brevis) blooms have regularly been reported in the Gulf of Mexico, particularly in the West Florida Shelf (WSF), and increasing efforts have gone into their detection, which, from space, still remains a challenge in the coastal waters due to the interferences from land and bottom reflectance. Detection algorithms that use blue-green reflectance ratios often have more uncertainties than the red or red-NIR algorithms, due to spectral interferences from colored dissolved organic matters (CDOM), which furthermore, does not correlate with chlorophyll in the coastal waters. Recently we proposed a bloom detection algorithm which uses the difference between two selected bands in the red (identified by us as the Red Band Difference (RBD)) technique, and a K. brevis classification technique, which uses the normalized difference of the two selected red bands (and identified by us as the K. brevis bloom index (KBBI)). In this study, we apply both of these approaches to satellite measurements by the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Medium Resolution Imaging Spectrometer (MERIS) ocean color sensors, which have a couple of bands in the red and NIR regions. We also make a comparison between our approaches and Fluorescence Line Height (FLH) approaches applied to bloom detection. Our analysis shows that although FLH can sometimes be used to detect blooms, it breaks down in highly scattering waters, often erroneously identifying other bloom like features, such as sediment plumes, as algal blooms, while our approaches generally correctly distinguish bloomed areas (give positive values) from non-bloom areas (give negative values) even in the highly scattering waters. Further analysis of the impact of CDOM on RBD and KBBI also shows that these techniques are less sensitive to CDOM. Thus, while there is a small reduction in the chlorophyll fluorescence signal due to the absorption of CDOM in the excitation wavebands, which slightly reduces the RBD and KBBI values, they still remain high enough to be detected and classified as blooms. Applications of the proposed approaches to satellite measurements are considered successful, since we were generally able to detect, trace and classify different K. brevis blooms documented in the literature with only minor ambiguities. The classification technique applied to MERIS data seems to give more false bloom alarms. The reason for this is not known at this point. However, this can probably be overcome by making agreement between both the detection and classification techniques a criterion for positive bloom detection.