Chlorophyll retrieval accuracies from satellite and in-situ radiometric measurements in Open Ocean and complex and bloom waters (Conference Presentation)

We examine the potential for ocean color (OC) retrievals using a neural network (NN) technique recently developed by us to make up for the lack of a 678 nm fluorescence band on VIIRS, previously available on MODIS and important for Karenia brevis harmful algal bloom (KB HABs) retrievals. NN uses VIIRS Remote Sensing Reflectance (Rrs) at 486, 551 and 671 nm to retrieve phytoplankton absorption at 443nm, from which both chlorophyll [Chla] concentrations and KB HABs can be inferred. NN retrievals are compared with retrievals obtained using different ocean color algorithms for both complex and open ocean waters. Retrievals were compared using both satellite match-ups and in-situ radiometric Rrs measurements against sample measurements of different field campaigns in the WFS and Atlantic coasts, 2014-18. VIIRS KB HABs retrievals in the WFS, using NN and other algorithms, are compared against retrievals obtained for different algorithms using satellite observations and in-situ radiometric Rrs measurements against sample measurements, 2017-18, for both the WFS and Atlantic coasts. Retrieval statistics showed (i) the important impact of temporal inter-intra pixel variations and sample depth considerations in complex bloom waters. These limit satellite retrieval accuracies and utility; and (ii) particularly for high chlorophyll bloom waters, better retrieval accuracies were obtained with NN followed by OCI/OCx algorithms. Likely rationales: the longer Rrs wavelengths used with NN are less vulnerable (i) to atmospheric correction inadequacies than the deeper blue wavelengths used with other algorithms, and (ii) to spectral interference by CDOM in more complex waters.