i
Analyses of satellite ocean color retrievals show advantage of neural network approaches and algorithms that avoid deep blue bands
-
2019
-
-
Source: J. of Applied Remote Sensing, 13(2), 024509 (2019)
Details:
-
Journal Title:Journal of Applied Remote Sensing
-
Personal Author:
-
NOAA Program & Office:
-
Description:We have previously shown the advantage of using neural network (NN) inversion algorithms over other ocean color (OC) algorithms in Visible Infrared Imaging Radiometer Suite satellite retrievals of Karenia brevis (KB) in the west Florida shelf (WFS). We now extend NN retrievals well beyond the WFS, to include both complex coastal and open ocean waters along the Florida and Atlantic coasts with a large dynamic range of chlorophyll-a values. Most importantly, we add in situ radiometric measurements (which in contrast to satellite retrievals, are invulnerable to atmospheric transmission correction errors) as inputs to retrieval algorithms, permitting algorithm comparisons for in situ and simultaneous colocated satellite retrievals against sample measurements. Results unequivocally demonstrate the intrinsic efficacy and unfettered applicability of NN algorithms in widely varying waters beyond the WFS. Furthermore, they show that avoiding deep blue bands in retrieval algorithms significantly improves accuracies. Likely, rationales are that longer wavelengths (used with NN) are less vulnerable to atmospheric transmission correction errors and to spectral interference by colored dissolved organic matter and nonalgal particles in more complex waters than deeper blue wavelengths (used with other algorithms), thereby arguing for development of OC algorithms using longer wavelengths. Finally, quantitative analysis of temporal, intrapixel, and sample depth variabilities highlights their important impact on retrieval accuracies.
-
Keywords:
-
Source:J. of Applied Remote Sensing, 13(2), 024509 (2019)
-
DOI:
-
Document Type:
-
Place as Subject:
-
Rights Information:Other
-
Compliance:Submitted
-
Main Document Checksum:
-
Download URL:
-
File Type: