Plan of Use of IR emissivity data to improve assimilation of GOES-R ABI in GFS

An accurate IR land surface emissivity (LSE) model is essential for IR window channels simulation. Initial assimilation of SEVIRI and GOES data indicates that the observation minus background (O-B) biases of water vapor channels (6.25 and 7.35 µm) and CO2 channel (13.4 µm) have near Gaussian distributions. However, other surface and cloud sensitive channels showed non-Gaussian distributions, especially for channel 4, 9 and 10 of SEVIRI sensor with heavy-tailed distribution. Under clear atmospheric conditions, large error sources that effect IR window channels radiance simulation comes from the infrared land emissivity error and the model forecast error of land surface temperature. In this study, we will investigate the potential impact of these two error sources on the assimilation of IR window channel radiances. We will evaluate the impacts of different IR land emissivity data bases on analysis fields and forecasts. GOES-R AWG land surface temperature (LST) retrievals are used to replace the model background LST in the simulation of IR radiances.