Improving Forecast Skill by Assimilation of Quality Controlled AIRS Cloud Cleared Radiances

ECMWF, NCEP, and GMAO routinely assimilate observed infra-red radiances as part of their operational data assimilation procedures. Radiances for select channels are passed into the data assimilation module, which accepts radiances for a given scene only for those channels whose radiances are thought to be unaffected by cloud cover. As a result of this, observed radiances for most tropospheric sounding channels are rejected under partial cloud cover conditions. AIRS Version-6 retrievals generate cloud cleared radiances for each retrieval which are derived parameters that represent what the AIRS instrument would have observed if the entire AIRS Field of Regard (FOR) were completely clear. We have conducted data assimilation experiments using a recent version of the GMAO Data Assimilation System (DAS) in which we assimilate Quality Controlled (QC'd) values of AIRS cloud cleared radiances in place of AIRS observed radiances as is done operationally, everything else being the same. AIRS cloud cleared radiances are accompanied by channel-by-channel, case-by-case error estimates which we use for QC purposes. The DAS accepts many more QC'd values of cloud cleared radiances than it does observed radiances because it thinks "those cloud cleared radiances are not affected by clouds". We have found that 7 day forecast skill is improved considerably, globally, as well as in the Northern Hemisphere Extra-tropics, and in the Southern Hemisphere Extra-tropics by the assimilation of AIRS cloud-cleared radiances instead of the assimilation of observed radiances. Results of these experiments will be shown.