The goal of this report is to provide an in-depth look at the HRRR, also known as the High Resolution Rapid Refresh RUC model compared to the actual radar data from the tornado outbreak that occurred in Southeastern Colorado and discuss the accuracy of the HRRR compared to real time data. This report will only highlight the Colorado domain because visual documentation of the tornadoes in the Texas Panhandle as well as Western Kansas was not collected.

April 22, 2010 will go down in the books as the most successful visual documentation tornado intercept by  Michael Carlson. A week prior to the event, the long range forecast was showing a potent upper level system moving into the plains area around April 20-23, 2011 period. As later model runs were generated the target became clear that Eastern Colorado would be a focus for strong supercell storms. All the ingredients were in place for strong tornadoes but the major question was if morning convection would clear out early enough so the atmosphere could heat up and destabilize.

Visible satellite showed nice clearing and heating going on in Southeast Colorado by 1715Z (11:15am MTN) this would be the primary area for thunderstorm initiation. (Fig 1.) The HRRR 12Z Composite Reflectivity model output run showed at this same time (17Z) that there was no convection occurring or forming in this area. (Fig 2.) Taking a look at real time composite reflectivity data from this same time would confirm this finding. (Fig 3.)

Fig 1. - Visible Sat Image

Fig 2. HRRR Composite Ref

Fig 3. Real Time Composite Ref

Moving ahead in time to the intonation period of the thunderstorm development, we can see on the Vis Sat that a nice cell was forming around 1815Z (12:15am MTN) just south of Las Animas, Colo. (Fig 4.) The HRRR 12Z Composite Reflectivity model output run showed at this same time no reflectivity (18Z)(Fig 5.) but the real time composite reflectivity data showed a small radar echo forming (17:54)(Fig 6.)

Fig 4. Vis Sat

Fig 5. HRRR Composite Ref

Fig 6. Composite Ref

At the time 1915Z of visual conformation of the first tornado the Vis Sat showed a nicely developed supercell south of Las Animas, Colo. (Fig 7.) The HRRR 12Z Composite Reflectivity model output run showed the same cell showing up near Las Animas, Colo. (Fig 8.) This same supercell was seen in real time on composite reflectivity (Fig 9.) as well as high resolution radar scans by the Pueblo, Colo. Radar.  (Fig 10.1. – 10.2.)

Fig 7. Vis Sat

Fig 8. HRRR Composite Ref

Fig 9. Composite Ref

Fig 10.1. High Res Velocity Radar from PUB.

Fig 10.2. Base Ref

Stepping back and looking at the other HRRR Composite Reflectivity model output runs it is noted that each run showed the exact same supercell forming in the region of the actual supercell event. There are some inconsistencies with each run on timing of supercell initiation and region of initiation. Though these inconsistencies were minor to the actual events that took place.

HRRR 13Z Run

HRRR 14Z Run

HRRR 15Z Run

HRRR 16Z Run

HRRR 17Z Run

HRRR 18Z Run

HRRR 19Z Run

It is clear that the HRRR is a powerful tool when it comes to forecasting. Not only did the HRRR predict the individual supercell that went on to produce a photogenic tornado but it dealt with the cloud cover that was the deciding factor in this particular setup. The HRRR should not be used as a single source forecasting tool but should be used in unison with other forecasting tools such as the RUC, WRF/NAM and GFS. Even though the HRRR was very accurate with this setup it is uncertain that the HRRR will be consistent with future setups, there are still many variables that need to be fixed before this is the “dream” model.

Michael Carlson