Statistical and case study assessment of RAP and HRRR convective forecast skill for 2013

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 5 February 2014
Hall C3 (The Georgia World Congress Center )
Stephen S. Weygandt, NOAA/Earth System Research Laboratory, Boulder, CO; and C. Alexander, D. C. Dowell, E. P. James, S. G. Benjamin, M. Hu, T. G. Smirnova, J. B. Olson, P. Hofmann, J. M. Brown, and H. Lin

During the 2013 convective season, a frozen version of the RAP / HRRR prediction system was run in a real-time experimental configuration to support the creation of aviation guidance products. The frozen RAP and HRRR configurations used in this exercise are prototypes for the configurations planned for operational NCEP implementation (RAP in late 2013 or early 2014 and HRRR later in 2014) and include a number of important enhancements over previous versions. For the RAP, these include an upgrade from a traditional 3DVAR assimilation technique to a hybrid ensemble technique using covariance information from the 80-member global ensemble data assimilation system and upgrades from the MYJ planetary boundary layer (PBL) parameterization to a specially adapted MYNN PBL scheme and from a 6-layer to a new 9-layer version of the Smirnova (RUC) land surface model formulation (LSM).

For the HRRR, the enhancements include the same WRF-ARW model changes as well as a new storm-scale data assimilation procedure designed to significantly improve short lead-time predictions of convection and other weather phenomena. This procedure features a one-hour 3-km HRRR pre-forecast period, in which 4 15-min. cycles of radar reflectivity assimilation are completed, followed by application of the Global Statistical Interpolation (GSI) variational analysis package at 3-km to ensure close fit to the latest conventional observations. The radar reflectivity assimilation method is similar to that used in the RAP (specification of latent heating-based temperature tendencies derived from the radar observed reflectivity during a forward integration), but omits the digital filtering aspect of the RAP initialization. This procedure has greatly reduced the model spin-up time for initializing ongoing precipitation systems, leading to much improved short term forecast performance.

The presentation will include a detailed statistical assessment of RAP and HRRR forecasts from the 2013 warm season, with an emphasis on the prediction of convective phenomena. This statistical analysis will be complemented by specific case-study analyses to illustrate details of how the assimilation and model upgrades lead to forecast improvement.