Intercomparison between CRTM and RTTOV in the NCEP Global Model

The observation operator for satellite radiance assimilation is devised to provide the information on model biases and the sensitivity of the model background to the observation in the data assimilation system, leading to a correction towards an improved model states. The accuracy of the observation operator, therefore, plays a vital role in the overall performance of the analysis system. The Community Radiative Transfer Model (CRTM) is currently used in the operational NCEP's Grid-space Statistical Interpolation (GSI) analysis system for satellite radiance assimilation. The purpose of this study is to investigate and to validate the overall accuracy of the CRTM by using observations and by comparing another well-established radiative transfer model as a reference.

The capability of using RTTOV (Radiative Transfer for TOVS), in addition to CRTM, has been implemented in GSI. Having two RTMs in the same analysis system has the following advantages in: (1) obtaining a more consistent and flexible way to compare RTM by using the same model input; (2) spotting errors by cross-validating each other; (3) establishing a symbiotic relationship between the two RT models by exploring relative new features in each one and thus facilitating new developments.

The results of this inter-comparison had led to several improvements and developments in the CRTM for data assimilation application under all-sky conditions. An improvement in boundary condition at the surface has significantly increased the accuracy in simulating brightness temperatures under the multiple-scattering condition for microwave sensors. The development of two-column radiance calculation allows more realistic considerations for cloudy scenes with fractional coverage. In addition, the cross validation between CRTM and RTTOV in the Infrared spectral range discovered a systematic bias in the surface emissivity model in CRTM over the ocean in higher latitudes as well as the cause of this bias. A work plan has established based on the discovery, and the improved infrared surface emissivity model over the ocean is underway. A more recent investigation of the simulated GOES-16 clear-sky radiance (CSR) between CRTM and RTTOV revealed the regression-based prediction of optical depth for channel 9 is problematic in CRTM. As a result, the efforts in improving the optical depth calculation for channel 9 is undergoing. The process and results of the inter-comparison mentioned above will be summarized in detail in this presentation.