Preprint version. Published in Journal of Geophysical Research, Volume 112, November 22, 2007.
An edited version of this paper was published by AGU. Copyright © 2007 American Geophysical Union. Further reproduction or electronic distribution is not permitted. The definitive version is available at http://dx.doi.org/10.1029/2007JC004136.
Many bioluminescence observations are made from the ocean's surface. However, the depth of the bioluminescent source is difficult to estimate on the basis of surface observations alone, given the variable light attenuation of unknown concentrations of water column constituents such as phytoplankton, colored dissolved organic matter, and detritus. Part 1 of this paper showed that bioluminescent water-leaving radiance signals are detectable, even in extremely turbid and dynamic coastal waters. Here, in part 2 of this paper, we analyze the water-leaving radiance patterns of bioluminescence modeled by HydroLight 4.2 to determine if the depth of the bioluminescent source can be estimated from its spectral signature. We find that the depth of the bioluminescent source is contained within the spectral signal and can be elucidated by simple neural networks. These networks can predict the depth of a bioluminescent layer with great accuracy, solely on the basis of the spectral shape of bioluminescent water-leaving radiance in a variety of water column and bottom type conditions. In addition, we found that as little as three wavelengths from the spectrum of water-leaving radiance are sufficient for an accurate determination of the depth of the bioluminescent source.