Characterization of AMSR2 and ATMS Observation Error Covariance in GSI under Cloudy Conditions

Numerical weather prediction (NWP) relies increasingly on direct assimilation of satellite radiance data to improve its forecasting accuracy. Currently, the radiances under clear and non-precipitation conditions are assimilated in most of NWP applications. Since satellite radiance measurements under precipitation conditions contain rich information on storm dynamics and thermal dynamics, we are developing a new methodology to characterize the observation error covariances for assimilation of satellite radiances under cloudy and precipitation conditions.

In this study, the simulation is provided by the Community Radiative Transfer Model (CRTM) developed by the US Joint Center of Satellite Data Assimilation (JCSDA), which has been widely used in NWP community for years. In cloud- or rain-affected regions, we enhance CRTM simulations including cloud scattering and the azimuthal variation of surface emission. For cloud scattering, the current CRTM version employs the Mie scattering theory, which assumes spherical liquid and ice water cloud particles for all cases in microwave frequencies. Under strong precipitation conditions where the large non-spherical particles are present, uses of Mie theory will result in large simulation uncertainty. Therefore, a more advanced model, the Discrete Dipole Approximation (DDA), which represents a particle as a three-dimensional array of polarizable points and provides a better model of the optical properties of non-spherical particles, is considered and tested in this study. For the emissivity over ocean surfaces, it is found that radiances from microwave imagers, such as the Advanced Microwave Scanning Radiometer 2 (AMSR2), exhibit a non-negligible variation that connects directly to the satellite azimuthal angle and surface wind direction. The azimuthal angle effect can be included through the ocean surface emissivity model, FAST microwave Emissivity Model 5 (FASTEM5) in CRTM. Simulation results of microwave imager channels with and without azimuthal effect are compared. After the test, it's found that the employment of DDA in CRTM for cloud scattering along with the consideration of the azimuthal angle effect in ocean surface emissivity, can improve the accuracy of microwave radiance simulation. Moreover, the above modified CRTM is used to simulate brightness temperatures from the AMSR2 and the Advanced Technology Microwave Sounder (ATMS) by using the atmospheric state profiles from European Centre for Medium-Range Weather Forecasts (ECMWF) and the retrieved hydrometeor profiles from collocated JAXA Global Precipitation Measurement/Dual-frequency Precipitation Radar (GPM/DPR) as inputs. Then, the two simulated brightness temperatures are compared against the AMSR2 and ATMS observations. Because the GPM is capable to measure the amount, size, intensity, and type of rain, and the DPR can provide the information of particle drop size distribution and rain types, the combination of GPM and DPR data makes possible to characterize the biases of observed and simulated microwave radiances under various types of cloud and rain. Such understanding of microwave radiance biases enables better characterization of observation error covariance under cloudy conditions, and prepares better for future direct assimilation of the microwave radiances under cloudy conditions in GSI.