A robust analytical algorithm for the diffuse attenuation coefficient of Downwelling Irradiance (Kd)

Charlotte Begouen Demeaux,  University of Maine,  charlotte.begouen@maine.edu
Emmanuel Boss,  University of Maine,  emmanuel.boss@maine.edu (Presenter)

A recent study (Begouen Demeaux and Boss, 2022) discovered a significant bias in empirical, implicit and semi-analytical algorithms commonly used to estimate the diffuse attenuation coefficient, notably in the clearest waters of the world. To address this issue, we gathered over 9000 in-situ Kd / Satellite Rrs matchups using BGC-Argo floats and tried to re-parameterized existing algorithms by fitting them with updated coefficients.

This improved the float-satellite matchup at a wavelength of 490nm, an independent wavelength of 412nm, and on independent in-situ datasets. However, we realized that while the floats/Satellite matchups constituted an improved dataset compared to cruise-acquired measurements, the dataset they were tuned on did not capture the full variability of the global ocean, making algorithms tuned to these data less robust, since algorithms who are empirically fit to data are only as good as the representativeness of the dataset with which they were constrained.

To address this issue, we hypothesized that using a physics-based analytical algorithm that depends solely on average cosine, absorption, and backscattering properties would avoid problems associated with training datasets shared by empirical algorithms. Indeed, Kd being an Inherent Optical Property (IOP), it is controlled to the first order by the absorption (a) and backscattering (bb) properties of a water parcel and to a smaller degree by the solar zenith angle. We use an average cosine parametrization devised by Morel and Prieur (1975) which accounts for water refraction, scattering near the surface and the sun and sky irradiance.

The new algorithm, called Kd-Morel, was evaluated on various databases, including in-situ databases spanning from ultra-oligotrophic to eutrophic waters, IOCCG hydrolight runs, and the synthetic database from Loisel et. al. We found that Kd-Morel performed relatively well with no bias for either very high or very low Kd. Therefore, we suggest using Kd-Morel when characterizing Kd at the global scale, as it seems more robust than current algorithms.

Poster Location ID: 1-52

Presentation Type: Poster

Session: Poster Session 1

Session Date: Tue (May 9) 5:00-7:00 PM

CCE Program: OBB