Estimation of ocean surface emissivity in AMSU channels: comparison with existing models

The ocean emissivity model used for the Conical-scanning Microwave Imager/Sounder (CMIS) is based on Wentz (1995). In the past years, the model has been continuously evaluated and refined using data from TMI, AMSR and Windsat. This model is only valid for the Earth incidence angle range 49-53° that occurs with conically scanning instruments. In order to be able to test the CMIS atmospheric retrieval algorithm with AMSU data we examined different ocean surface models that are used operationally in weather prediction and are applicable from nadir to edge-of-scan. Comparison with the CMIS model at large scan angles showed that significant differences exist between the FASTEM3 model included in the RTTOVS package and the NOAA/NESDIS (Weng, personal communication) and CMIS models. This paper reports work to extend Wentz's approach to the retrieval of wind induced emissivity from NOAA-16 AMSU data and to compare the retrieved AMSU emissivities with those derived from the EOS/AMSR instrument using the same technique. A new formulation for the correction for non-specular reflection, derived from geometric optics, makes it possible to directly retrieve emissivities for mixed polarization measurements. Using this approach, the hope is to be able to assess validity of the FASTEM and NESDIS models, to verify the consistency between AMSU and AMSR measurements and to provide a model that can be used in our own AMSU retrieval experiments. For this work, we used the dielectric constant of sea water from Meissner and Wentz (2004) and the MonoRTM (Clough et al. 2005) atmospheric model. MonoRTM is derived from the LBLRTM (Clough et al. 1992) line-by-line model and has been extensively validated using up-looking radiometer data from the Atmospheric Radiation Measurement sites. Simulation of the observations involves NCEP Global Forecast System temperature, wind and sea surface temperature fields, and precipitable water data from AMSR-E. Sensitivity of the results to the choice of atmospheric model and source of precipitable water is provided.