NWP Background Adjustment Through 1DVAR Preprocessing of Satellite Radiances

Data assimilation systems rely partially on the use of background error covariances to remove discrepancies between the background (guess) fields and observations, in order to produce an analysis that more precisely characterizes the state of the surface and atmosphere at the model forecast initialization time. However due to the high degree of non-linearities in active weather regions (e.g. clouds and precipitation) in both time and space, background errors are difficult to prescribe. This often leads to unbalanced analysis fields when satellite observations in those regions are assimilated, resulting in model spin-up issues and degraded forecasts. In this study, we apply a 1DVAR algorithm to satellite observations in all-sky conditions and over all surfaces as a preprocessor that adjusts background fields prior to the full 3D/4DVAR data assimilation. The 1DVAR, known as the Multi-Instrument Inversion and Data Assimilation Preprocessing System (MIIDAPS) is applicable to multiple passive microwave and infrared space-borne sensors, and simultaneously inverts brightness temperature observations into geophysical parameters which include the temperature, humidity, hydrometeor, and trace gas profiles, along with surface parameters including skin temperature and surface emissivity. The 1DVAR analyses are blended given weights for close proximity to the analysis time as well as the level of convergence reached during minimization. The updated background fields are then passed to the full assimilation, in this case the Gridpoint Statistical Interolation (GSI) 4DEnsVar, with adjusted atmospheric and surface fields including an updated dynamic surface emissivity. The impact on the assimilation of cloud- and precipitation-affected observations as well as surface-sensitive observations will be shown. In addition, we will present the impact on overall analysis and forecast quality in the NOAA/NCEP Global Forecast System (GFS) model.