POSTER SESSION 2
10:30–12:30, Tuesday, September 15
Poster Session | 1 | 2 | 3 | 4 | Schedule at a Glance
ABSTRACT 1097 | POSTER 86
A HYBRID FRAMEWORK FOR ROBUST RETRIEVAL OF THE DIFFUSE UPWELLING RADIANCE ATTENUATION COEFFICIENT FROM HYPERNAV PROFILES
The diffuse attenuation coefficient for upwelling radiance, [image], is critical for propagating in-water measurements to the ocean surface with traceable accuracy. However, accurate retrieval of [image] from in situ radiance profiles remains challenging due to surface wave-induced variability in the near-surface light field, depth dependence caused by inelastic scattering and fluorescence, and low light levels at longer wavelengths.
We present a hybrid retrieval framework for estimating depth-integrated [image] from HyperNav radiometric profiles that combines measurement-derived regression with radiative transfer modeling. The approach first retrieves a depth-averaged [image] over a constrained vertical interval using a weighted robust log-linear regression applied to spectrally smoothed upwelling radiance profiles. Signal-to-noise–based quality control and robust loss functions are used to reduce the influence of instrumental noise and wave-induced variability.
To account for the depth dependence of [image] and to extend the retrieval into the red and near-infrared, where direct measurements are unreliable, the measurement-derived estimates are blended with HydroLight-modeled values that include inelastic scattering and fluorescence. Modeled inputs are constrained by environmental conditions at the time of measurement, including sun zenith angle and chlorophyll concentration. Uncertainties are propagated from the regression covariance and from model uncertainty estimated using Monte Carlo simulations.
Application to HyperNav deployments in clear waters (Crete, Hawaii, Puerto Rico, Tahiti) demonstrates improved spectral consistency and reduced bias relative to legacy processing, providing a robust pathway for operational [image] retrievals from autonomous radiometric profilers.
Nils Haëntjens, University of Maine, [email protected], https://orcid.org/0000-0002-7155-2721
Robert Frouin, University of California San Diego, [email protected]
Emmanuel Boss, University of Maine, [email protected]
Jing Tan, University of California San Diego, [email protected]
Andrew Barnard, Oregon State University, [email protected]
Poster Session | 1 | 2 | 3 | 4 | Schedule at a Glance
