Observation System Experiments with the Hourly-Updating Rapid Refresh (RAP) Model Using GSI Hybrid Ensemble/Variational Data Assimilation

A set of observation system experiments (OSEs) over 3 seasons using the hourly-updated Rapid Refresh (RAP) numerical weather prediction (NWP) model identifies the relative importance of the 11 various types of observations for 3-12h RAP forecasts.  Aircraft observations are shown the strongest impact for wind, relative humidity and temperature forecasts averaged over 1000-100 hPa. but many other observation types also have significant positive impact including satellite cloud and wind retrievals, surface, GPS precipitable water, radar, and rawinsondes.   Surface weather forecasts (including forecasts of 2-m temperature and dewpoint, 10-m wind, and cloud ceiling) are, not surprisingly, most strongly impacted by the assimilation of surface observations.    Separate OSEs were carried out for aircraft observations of wind only and moisture only and also for ascent/descent vs. enroute observations with broadening cross-variable and vertical-level impact with increase in forecast duration.  All of these observation system experiments were recently performed using the latest Rapid Refresh version being implemented at NCEP as RAP version 3 in late summer 2016.