Optimization of a semianalytical ocean color model for global-scale applications

Abstract

Semianalytical (SA) ocean color models have advantages over conventional band ratio algorithms in that multiple ocean properties can be retrieved simultaneously from a single water-leaving radiance spectrum. However, the complexity of SA models has stalled their development, and operational implementation as optimal SA parameter values are hard to determine because of limitations in development data sets and the lack of robust tuning procedures. We present a procedure for optimizing SA ocean color models for global applications. The SA model to be optimized retrieves simultaneous estimates for chlorophyll (Chl) concentration, the absorption coefficient for dissolved and detrital materials [a(cdm)(443)], and the particulate backscatter coefficient [b(bp)(443)] from measurements of the normalized water-leaving radiance spectrum. Parameters for the model are tuned by simulated annealing as the global optimization protocol. We first evaluate the robustness of the tuning method using synthetic data sets, and we then apply the tuning procedure to an in situ data set. With the tuned SA parameters, the accuracy of retrievals found with the globally optimized model (the Garver-Siegel-Maritorena model version 1; hereafter GSM01) is excellent and results are comparable with the current Sea-viewing Wide Field-of-view sensor (SeaWiFS) algorithm for Chl. The advantage of the GSM01 model is that simultaneous retrievals of a(cdm)(443) and b(bp)(443) are made that greatly extend the nature of global applications that can be explored. Current limitations and further developments of the model are discussed.

Generalized ocean color inversion model for retrieving marine inherent optical properties

“…Franz and Werdell [13] provide a detailed description of its use within the SeaDAS environment [20]. The general form of the GIOP inversion model (Section 2.A) is common to a number of published approaches (e.g., [3][4][5]8,9,30]) whose differences reside in the choice of eigenvectors employed, the number of eigenvalues resolved, the optimization method selected, and the number of sensor wavelengths considered in the optimization. A unique instance of GIOP is therefore defined by specifying eigenvectors for each optically significant constituent assumed to exist in the water column.…”

“…The latter normalization allows the spectral shape of a ϕ λ to change with an estimate of C a , but constrains its magnitude to an average value for oceanic water [3,8,14]. Our choice of default parameterizations served two purposes: to provide some consistency with previously developed SAAs and to acknowledge the emerging quality of variable, dynamically selected eigenvectors.…”

“…Many published SAAs attempt to retrieve spectral IOPs from spectral remote-sensing reflectances [R rs λ ; sr −1 ] and to further separate the total absorption and backscattering coefficients into subcomponents [e.g., the absorption of algal, nonalgal, and colored dissolved organic matter (CDOM)] [3][4][5][6][7][8][9][10][11]. These SAAs generally assume spectral shape functions (or eigenvectors) of the constituent absorption and scattering components and retrieve the magnitudes (or eigenvalues) of each constituent required to match the spectral distribution of R rs λ .…”

“…Briefly, GIOP allows construction of different IOP models at runtime by selection from a wide assortment of published absorption and backscattering eigenvectors [4,[6][7][8][14][15][16][17]. As such, GIOP permits isolation and evaluation of specific modeling assumptions, construction of new SAAs, development of regionally tuned SAAs (e.g., [18]), and ensemble inversion modeling.…”

Modelling ocean-colour-derived chlorophyll <i>a</i>

“…In fact, the differences in how CDOM and detrital matter absorb and scatter light has been used in algorithm development (e.g. Sathyendranath et al, 1989;Roesler and Perry, 1995;Maritorena et al, 2002). The largest discrepancies between algorithms that explicitly include or exclude the differences in the optical properties is most noticeable at high latitudes (Siegel et al, 2005b), and CDOM and non-algal particles are noted to be especially important (Brown et al, 2008).…”