Abstract
Non-algal particles and chromophoric dissolved organic matter (CDOM)
are two major classes of seawater constituents that contribute
substantially to light absorption in the ocean within the ultraviolet
(UV) and visible (VIS) spectral regions. The similarities in the
spectral shape of these two constituent absorption coefficients, ad(λ) and ag(λ), respectively, have led to their
common estimation as a single combined non-phytoplankton absorption
coefficient, adg(λ), in optical remote-sensing
applications. Given the different biogeochemical and ecological roles
of non-algal particles and CDOM in the ocean, it is important to
determine and characterize the absorption coefficient of each of these
constituents separately. We describe an ADG model that partitions adg(λ) into ad(λ) and ag(λ). This model improves upon a recently
published model [Appl. Opt. 58, 3790
(2019)APOPAI0003-693510.1364/AO.58.003790] through
implementation of a newly assembled dataset of hyperspectral
measurements of ad(λ) and ag(λ) from diverse oceanic environments to
create the spectral shape function libraries of these coefficients, a
better characterization of variability in spectral shape of ad(λ) and ag(λ), and a spectral extension of model
output to include the near-UV (350–400 nm) in addition
to the VIS (400–700 nm) part of the spectrum. We
developed and tested two variants of the ADG model: the ADG_UV-VIS
model, which determines solutions over the spectral range from 350 to
700 nm, and the ADG_VIS model, which determines solutions in
the VIS but can also be coupled with an independent extrapolation
model to extend output to the near-UV. This specific model variant is
referred to as ADG_VIS-UVExt. Evaluation of the model with
development and independent datasets demonstrates good performance of
both ADG_UV-VIS and ADG_VIS-UVExt. Comparative analysis of
model-derived and measured values of ad(λ) and ag(λ) indicates negligible or small median
bias, generally within ±5% over the majority of the
350–700 nm spectral range but extending to or above 10%
near the ends of the spectrum, and the median percent difference
generally below 20% with a maximum reaching about 30%. The presented
ADG models are suitable for implementation as a component of
algorithms in support of satellite ocean color missions, especially
the NASA PACE mission.
Funder
National Aeronautics and Space
Administration