Impacts of Observation Errors of Hyperspectral Infrared Sounders under Partially Cloudy Regions for Hurricane Forecasts

Hyperspectral infrared (IR) sounders provide high vertical resolution atmospheric sounding information that can improve the forecast skill in numerical weather prediction (NWP) models. Usually only IR radiances under clear skies and radiances of clear channels under cloudy skies are assimilated in the operational NWP centers since hyperspectral IR sounder observations are highly affected by clouds and direct assimilation of IR radiances in cloudy skies is still challenging. To use the information under partially cloudy regions, a cloud-clearing (CC) method provided an alternative method to take advantage in data assimilation of the thermodynamic information from cloudy skies. The CC method removes the cloud effects from an IR cloudy field-of-view (FOV) and derives the cloud-cleared radiances (CCRs). This CC method has already been successfully applied on AIRS/MODIS to obtain the AIRS CCRs, and CrIS/VIIRS to obtain the CrIS CCRs. The CCRs are clear sky equivalent radiances and contain the temperature and moisture information under cloudy regions. In general, the observation errors of radiances in cloudy regions are set to be much larger than clear skies for direct cloudy radiance assimilation due to the large uncertainties of both the NWP and RTM. Due to noise amplification in cloud-clearing, it is not appropriate to use errors of clear radiances for assimilating CCRs. In this study, the observation errors of CCRs are inflated from the original clear observation error covariance based on the fractional cloud coverage of FOVs to be cloud-cleared. The observation error inflation method is developed and discussed; Hurricane Harvey (2017) is used as a case study to demonstrate the impacts of a more realistic observation error when assimilating CrIS CCRs.