# Generalized IOP algorithm (GIOP)

In 2008, NASA proposed to provide the data sets, processing framework, and international forum within which a new generation of global IOP products could be developed and evaluated. To initiate this effort, NASA organized an international IOP Algorithm Workshop, conducted in conjunction with the Ocean Optics XIX conference. Participants ultimately reached consensus on a processing framework, the Generalized IOP algorithm (GIOP), within which NASA has started producing IOP data products for community evaluation. A preliminary unified algorithm was proposed to initiate this process based on a specific configuration of GIOP, with alternative model parameterizations and features defined for subsequent evaluation.

#### Theoretical basis of GIOP

Remote-sensing reflectances () are related to absorption () and scattering coefficients via:

(1)

where () is the total backscattering coefficient and all terms are spectral. () varies with illumination conditions, sea surface properties, and the shape of the marine volume scattering function. The absorption coefficient can be expanded as the sum of all absorbing components and each component can be further expressed as the product of its concentration-specific absorption spectrum (eigenvector; ) and its concentration (eigenvalue; ):

(2)

where the subscripts , , and indicate contributions by water, dissolved organic matter (gelbstoff) + non-algal particles (detritus), and phytoplankton, respectively. The combination cannot currently be decomposed into its two components using remote-sensing methods. Total backscattering can be expanded to:

(3)

where the subscripts and indicate contributions by water and particles, respectively. Both and are known.

Using as input, eigenvalues for absorption and scattering ( and ) can be estimated via linear or non-linear least squares inversion of Eq. (1).

#### Configuration of GIOP

Put simply, GIOP allows on-the-fly configuration of the above ocean reflectance inversion model within the L2GEN data processing environment. For example, the three eigenvectors in Eqs. (2) and (3) can be assigned at run-time. The AOP-IOP relationship described by can also be configured (e.g., the quadratic form of Gordon et al. (1988) or the f/Q tables of Morel et al. (2002)), as can the inversion method (e.g., Levenburg-Marquardt or matrix inversion). A full description of GIOP configuration options is provided in its User's Guide.