Neural Network Algorithms for Retrieval of Harmful Algal Blooms in the West Florida Shelf from VIIRS Satellite Observations and comparisons with other techniques, without the need for a fluorescence channel

New approaches are described that use of the Ocean Color Remote Sensing Reflectance readings (OC Rrs) available from the existing Visible Infrared Imaging Radiometer Suite (VIIRS) bands to detect and retrieve Karenia brevis (KB) Harmful Algal Blooms (HABs) that frequently plague the coasts of the West Florida Shelf (WFS). Unfortunately, VIIRS, unlike MODIS, does not have a 678 nm channel to detect Chlorophyll fluorescence, which is used in the normalized fluorescence height (nFLH) algorithm which has been shown to help in effectively detecting and tracking KB HABs. We report on using neural network (NN) algorithms, reported by us, and trained, using a wide range of suitably parametrized synthetic data typical of coastal waters, to form a multiband inversion algorithm which models the relationship between Rrs values at the 486, 551 and 671nm VIIRS bands against the values of phytoplankton absorption (a(ph)), CDOM absorption, non-algal particles (NAP) absorption and the particulate backscattering bb(p) coefficients, all at 443nm and permitting retrievals of these parameters. The NN retrieved a(ph443) in the WFS is in turn filtered by known limiting conditions on Chlorophyll concentration [Chla] and low backscatter properties associated with KB HABS in the WFS and used to identify, delineate and quantify them. Comparisons with in-situ measurements and other techniques including nFLH confirm the viability of both the NN retrievals and the filtering approaches devised.