Stratiform Cloud-Hydrometeor Assimilation for HRRR and RAP Model Short-Range Weather Prediction

Abstract

AbstractAccurate cloud and precipitation forecasts are a fundamental component of short-range data assimilation/model prediction systems such as the NOAA 3-km High-Resolution Rapid Refresh (HRRR) or the 13-km Rapid Refresh (RAP). To reduce cloud and precipitation spin-up problems, a non-variational assimilation technique for stratiform clouds was developed within the Gridpoint Statistical Interpolation (GSI) data assimilation system. One goal of this technique is retention of observed stratiform cloudy and clear 3D volumes into the subsequent model forecast. The cloud observations used include cloud-top data from satellite brightness temperatures, surface-based ceilometer data, and surface visibility. Quality control, expansion into spatial information content, and forward operators are described for each observation type. The projection of data from these observation types into an observation-based cloud-information 3D gridded field is accomplished via identification of cloudy, clear, and cloud-unknown 3D volumes. Updating of forecast background fields is accomplished through clearing and building of cloud water and cloud ice with associated modifications to water vapor and temperature. Impact of the cloud assimilation on short-range forecasts is assessed with a set of retrospective experiments in warm and cold seasons using the RAPv5 model. Short-range (1-9h) forecast skill is improved in both seasons for cloud ceiling and visibility and for 2-m temperature in daytime and with mixed results for other measures. Two modifications were introduced and tested with success: use of prognostic subgrid-scale cloud fraction to condition cloud building (in response to a high bias) and removal of a WRF-based rebalancing.