Accurate measurements of water surface skin temperatures (WST) from satellite are required in numerous research and civil applications. Of particular importance are climate studies that
require a high degree of absolute accuracy (<0.3 K). In this work a physical retrieval methodology is developed for general application to any scanning spectrometer that obtains multispectral IR window radiance measurements. Such instrumentation includes high resolution Fourier transform spectrometer (FTS) systems as well as narrowband channel radiometers. To achieve this a quasi-specular surface reflection model is developed. The reflection model assumes the Kirchhoff approximation (valid for IR wavelengths and water surfaces) for rough surface scattering. The wave slope statistics of Cox and Munk (1954,1955) are then employed for the double integration of reflected rays over zenith and azimuth angle. The computation is performed via Gaussian quadrature using an uplooking fast transmittance model. The full hemispherical model
calculation is simplified by the concept of a reflection-diffusivity angle. Lookup tables of this parameter are computed as a function of wavenumber, viewing angle, wind speed and total uplooking transmittance. The model compares favorably against observed radiance spectra obtained from the Marine Atmospheric Emitted Radiance Interferometer (M-AERI) in the tropical Pacific ocean during the Combined Sensor Program (CSP) research cruise. The WST retrieval method then utilizes the reflection model and is shown to exhibit reduced random and systematic error in simulation. The method is applied to radiance spectra obtained from the NPOESS Airborne Sounder Testbed-Interferometer (NAST-I) during two field campaigns. High-resolution lake and sea surface temperature fields are measured which reveal mesoscale structures and large temperature gradients. Retrievals obtained within the vicinity of moored buoys compare favorably with the in-situ measurements.