Thursday, 18 January 2007: 1:30 PM
The Effect of Reabsorption of Chlorophyll Fluorescence and Elastic Scattering in Coastal Waters on the Efficacy of Retrieval Algorithms
212B (Henry B. Gonzalez Convention Center)
The sun-induced chlorophyll fluorescence signal in the 685 nm spectral region is an important indicator of chlorophyll concentration, photosynthetic activity and algae blooms in seawater. Making use of satellite sensors with 3 spectral bands in the vicinity of the fluorescence peak, fluorescence line height (FLH) algorithms have been developed and shown to be effective for fluorescence retrieval in open ocean waters (Case 1). It is now well recognized that equivalent fluorescence measurements are more complicated in coastal waters because of the higher concentrations of chlorophyll a concentrations [Chl] and TSS and the resulting overlaps of the fluorescence spectrum with the elastic reflectance spectrum. Understanding the contribution of fluorescence to total water leaving radiance is particularly relevant to the analysis of bio-optical properties of coastal waters and the improvement of related algorithms which utilize NIR spectral features for retrieving chlorophyll-a concentration [Chl]. This work examines and reports on the impact of reabsorption by chlorophyll as well as the impact due to the presence of CDOM and mineral particles on the features and shifts in the peak of the combined water-leaving fluorescence and elastic scattering spectrum. The results were obtained through a combination of laboratory studies, field measurements and radiative transfer simulations for typical coastal water conditions. The laboratory work included comparison of the fluorescence retrieval from reflectance spectra using the polarization discrimination technique recently reported by us, now, however, combined with band pass filters to provide direct but independent confirmation of the derived fluorescence spectra. In these measurements, as well as in a series of experiments using laser illumination, the magnitude of water leaving fluorescence is examined for various types and concentrations of algae, showing the relationship between fluorescence magnitude and [Chl]. These experiments show that unlike some previous models, the linear relationship between [Chl] and fluoresence is maintained over a wide range of [Chl] concentrations. Our field work was carried out in the Chesapeake Bay in the summer 2005 and in the waters near Sapelo Island, Georgia in the summer of 2006. The campaigns utilized a WET Labs package containing an ACS instrument to measure hyper-spectral absorption a, beam attenuation spectra c, and a bb9 instrument for backscatter, chlorophyll [Chl] and CDOM concentrations measurements. A CTD was used for salinity, temperature and depth measurements, and a GER spectroradiometer was used for reflectance measurements. Data were collected at 42 stations in the Chesapeake Bay and 30 stations in Georgia. [Chl] varied from 2.2 to 236 mg/m3, absorption at 400 nm from 1.4 to 16.4 m-1, extinction from 5.7 to 28 m-1 and backscattering from 0.03 to 0.48 m-1 for these stations. CDOM spectra were measured directly at some stations by attaching a 0.2 Ám filter to the intake of ACS instrument, for the other stations they were retrieved from total absorption spectra as an exponential component and then compared with CDOM fluorescence data. To obtain an estimate of the fluorescence relationship to [Chl] concentration, the measured absorption and attenuation data from the acs instrument were used as inputs for reflectance simulations using the Hydrolight program without a fluorescence component inserted. The output of this simulation provided a reasonable approximation to the elastic component of the water signal except for a scale parameter which is needed to fit the simulated spectra to the measured elastic signal and is equivalent to the adjustment factor needed to correct the backscatter/scatter ratio from that used in Hydrolight. Once the scale corrected elastic signal is obtained, it was subtracted from the measured spectrum and the resultant fluorescence was fitted to a Gaussian with center at 685 nm and FWHM 25 nm, to obtain the magnitude. From these measurements, we found that in the range of [Chl] < 30 mg/m^3 fluorescence is generally proportional [Chl] and corresponds well to the MODIS FLH algorithm. For higher concentrations saturation is observed in the field measurements. However, by comparison, the laboratory measurements showed that fluorescence magnitude, both as retrieved from reflectance spectra using the polarization discrimination method, and from laser induced fluorescence showed only slight saturation with [Chl] for much higher [Chl] ~200 mg/m^3. Therefore, the observed saturation may be the result of more complex field conditions and is the focus of continuing analysis. By comparing the laboratory results with the field results and radiative transfer simulations it is possible to conclude that in coastal waters with high [Chl], fluorescence is a relatively small contributor to the total NIR reflectance peak, which is at slightly longer wavelengths than the fluorescence peak, and that the magnitude of this composite peak is due primarily to elastic reflectance by particulates (including minerals and algae) which is convolved with the significant [Chl] absorption and increasing water absorption. This overlap and dominance of elastic reflectance components in the observed peak, makes the retrieval of even strong fluorescence questionable, if measurements are available in only 3 spectral bands. To evaluate the impact of these conclusions, we tested the performance of the FLH algorithms using an IOCCG data set of reflectances created through Hydrolight for various coastal water compositions and [Chl] concentrations. Fluorescence was then superimposed on top of these reflectances as a Gaussian shape with the center at 685 nm and a wide range of magnitude dependence on [Chl]. The superimposed fluorescence was then retrieved using different FLH approaches. It was found that below 30 mg/m^3 [Chl] fluorescence is readily retrieved, but beyond 30 mg/m^3 [Chl], the error in the retrieval significantly increases. Other algorithm approaches better able to take into account the impact of algal chlorophyll absorption on reflectance spectra are now under consideration.