A parameterized forward model for WindSat ocean vector wind retrievals

WindSat ocean vector wind retrievals are retrieved in near-real-time and operationally assimilated into the numerical weather prediction global models of the U.S. Navy, the U.K. Met Office and NOAA NCEP. These retrievals are produced using the Naval Research Laboratory (NRL) retrieval algorithm. The retrieval algorithm utilizes a parameterized model for the modified Stokes components at the frequencies of the measured WindSat brightness temperatures. This forward model must be accurate to support wind vector retrievals because the wind direction signal is small relative to the magnitude of the measured brightness temperatures.

The forward model includes parameterizations for the sea surface emissivity and reflectivity. The emissivity varies with sea surface temperature, wind speed, earth incidence angle and sea surface salinity. The radiative transfer through the atmosphere is parameterized to match a one-layer slab model which includes absorption due to oxygen, water vapor and cloud liquid water. The applicability of the forward model is currently limited to non-precipitating conditions – scattering is neglected. The radiative transfer model is described and compared to earlier versions used for NRL WindSat retrieval algorithm development.

The radiative transfer model is parameterized for use in the physically-based retrieval algorithm to enable retrievals in near-real-time. The parameterization is developed to match the retrieval algorithm. We describe the parameterization methods used for the current NRL WindSat algorithm and contrast them to earlier versions of our algorithm and the other parameterizations described in the literature. We discuss the estimates of the accuracy of the forward model and the impact of the accuracy on ocean vector wind retrieval performance.