All-Sky Assimilation of ATMS Observations with Correlated Error

Research with correlated satellite observation error has been ongoing at NCEP and has primarily focused on estimating and accounting for inter-channel error correlations in Atmospheric Infrared Sounder (AIRS) and Infrared Atmospheric Sounding Interferometer (IASI) observations that are taken over the sea and are free from cloud contamination. In the Gridpoint Statistical Interpolation (GSI), it is assumed that the observation errors of different satellite channels are uncorrelated. To compensate for the neglected correlations, the errors are inflated. Modifying the representation of these errors in the GSI to include correlated error has proven to give a better weighting to these observations in the assimilation and yield a positive forecast impact in the troposphere. Therefore, neglecting error correlations and inflating errors do not give an optimal representation of these observations in the analysis.

The all-sky assimilation of Advanced Microwave Sounding Unit-A (AMSU-A) radiances has been implemented in the operational Global Forecast System (GFS) at NCEP in 2016. In early 2019, all-sky assimilation of Advanced Technology Microwave Sounder (ATMS) radiances will also be included, with the transition to the Finite Volume Cubed-Sphere (FV3) dynamical core. In this configuration, observation errors are assigned to AMSU-A and ATMS observations based on the cloud amount. This presentation will discuss the all-sky assimilation of ATMS observations with correlated error. Observation errors and error correlations are expected to be larger in cloudy scenes. The estimation of the error correlation matrix will be detailed, and two approaches will be tested to account for the correlations in the all-sky configuration. First, error correlations will only be used when observations are free from cloud contamination. In the second approach, an error covariance matrix will be modeled as a function of cloud amount. The forecast impact will be verified against ECMWF analyses and against satellite and conventional observations in either case.