Monday, May 25, 2015

Short Range Computer Projection vs Actual Radar

Its Memorial Day here at the station. I remember my dad, grandfather, uncle, brother-in-law who served in our Armed Forces over the generations.

I've been watching a few showers popping up on radar.  Nothing heavy, nothing widespread.  One helpful tool we use in determining where and when precipitation develops is the newly operational HRRR model (short for High Resolution Rapid Refresh).  In essence, this model assimilates 3 km radar data into it calculation each hour.  So rather than wait a full 6 hours until the next set of model outputs, the HRRR runs a NEW output every 60 minutes taking into account the ew radar information.

This morning, I ran a little experiment to see how accurate the HRRR was in determining the light showers it was showing for 9AM.  The NWS Cleveland radar was down for repairs so this was a perfect opportunity to see how this would affect the output.

Here is the HRRR Output each hour starting with the 5z (1am) run and concluding with the 11z (7am) run.  So the 5z run is 8 hours from 9am.  The 11z run is 2 hours from 9am.  So each successive model run brings us closer to the 9am target time. 

A light shower west of Cleveland...


By the 8z run, the HRRR started to shift the light shower further west into the main cluster of rain along the Ohio/Indiana border.  By 11z, it had merged with the western rain area.

Here is what the actual radar looked like at 9AM using the TDWR radar in Grafton since the NWS radar was still down as of this writing.  Hardly any precipitation!

Thinking outloud...

*  The HRRR overestimated the development 
*  Unknown whether Cleveland radar downtime had an affect
*  Unknown whether or not TDWR radar data from the FAA is incorporated into HRRR model.