The Roesler-Perry inverse reflectance model is employed to derive inherent optical properties (IOPs) from multi-spectral remotely sensed reflectance Rrs, produced from both radiometric measurements and Hydrolight radiative transfer modeling. These are derived from concurrent measurements of downwelling irradiance Ed (l, z), upwelling radiance Lu (l, z), total absorption a (l, z), and beam attenuation c (l, z) from the MISERY (Multiple Instrument Synoptic Educational Ruthless Youth) data set, from Puget Sound in August 1998. Closure between modeled and measured Rrs spectra is assessed with a view to future utilization of Hydrolight modeling for inverse model validation. Such an approach will allow limited future validation of inverse reflectance models from waters with only limited IOP data available. The Roesler-Perry inverse model is applied to corresponding measured and modeled Rrs spectra, providing independent tests for both forward Hydrolight modeling and the RP model. Both Hydrolight and the RP model compare well with measured Rrs, particulate and dissolved absorption determined spectrophotometrically, and Hydroscat-6 determined backscattering.
The RP model is then applied to Hydrolight modeled Rrs data, produced using three measured phytoplankton absorption spectra as endmembers. The bio-optical conditions simulated with Hydrolight are deliberately designed to be highly variable rather than an accurate representation of observed in situ conditions, as the objective is a rigorous test of RP model performance. The phytoplankton absorption spectra, originally measured with the filter pad technique in the Benguela system, were chosen for spectral variation: a mixed diatom/small flagellate population adiatom, a "red tide" sample composed primarily of Ceratium furca adino, and a "brown tide" sample composed primarily of Aureococcus sp. aaureo . These were ascribed chlorophyll a specific absorption values at 676 nm of a*diatom (676) = 0.009, a*dino (676) = 0.007 and a*aureo (676) = 0.012, all of which were assumed to be constant with depth. A smooth polynomial derived chlorophyll a vertical profile, varying from 7 mg m-3 at the surface to 25 mg m-3 at 6 m, was then used to produce profiles of phytoplankton absorption. Gelbstoff absorption was produced using the same vertical profile, assuming ag(400) =1 m-1 at 5 m, with the slope factor S=0.018 nm-1. Detrital absorption was assumed zero. Chlorophyll a specific scattering coefficients b* (l) were approximated from Ahn et al 1992 using H. elongata data for b*diatom (l), P.micans data for b*dino(l) and Synechococcus sp. data for b*aureo (l). Again, an emphasis is placed on rigorously testing the RP model under extreme bio-optical conditions rather than producing naturally representative data. Reflectance data were then produced with Hydrolight, using the Petzold coastal phase function, a surface wind speed of 5 m s-1 , and the Hydrolight semi-empirical sky model based on a sampling location in Puget Sound in August 1998. Data were modeled from a depth of 6 m to surface, corresponding to a minimum of 4 optical depths at 550 nm. Application of the RP model to the resultant reflectance data resulted in agreement to within ??20%?? between "measured" and returned af (l) data, despite poor agreement between "measured" and modeled ag and bb coefficients. The robustness of the RP model in determining phytoplankton absorption from high biomass water types with a range of phytoplankton assemblages is demonstrated.
Figure 1: Comparison of reflectance measured with a Satlantic SPMR and transmitted through the air-water interface, and that produced from Hydrolight using contiguous AC9 data.
Figure 2: Performance of the RP model in returning absorption, backscattering and reflectance for both measured and Hydrolight modeled reflectance data for a single station in Puget Sound August 1998. Returned absorption data are compared to spectrophotometrically determined absorption of discrete surface water samples. Returned backscattering data are compared to mean contiguous Hydroscat-6 measurements.
Figure 3: Performance of the RP model in returning phytoplankton absorption spectra from Hydrolight modeled endmember runs. Example profiled parameters are shown at right.
