Summer 2008 Validation

  • n220 severe hail cases
    • 69 AB, 107 SK, 44 MB
  • n266 radar cases (more since 2 radars could see same hail)
  • nUsed hail algorithm results interpolated to event location from scan time closest to event and 10 minutes previous upstream from event
  • nWhen in doubt about:
    • exact ¡location => use max value in area
    • ¡time => use max value over the point
    • nTherefore => built-in bias towards larger hail forecasts
Hail Total by Size
 
 
The NWS MESH appears to outperform the existing URP algorithm for smaller hail sizes < walnut-sized (35 mm), but with larger hail MESH under reports it. URP does a better job at the larger sizes. 
 

 
The POSH index seemed to perform better with larger hail, suggesting that perhaps a combination of the POSH and MESH could be used. Even with that, the errors are quite large, so in the summer of 2009, work was started on a brand-new hail algorithm for URP and eventually for NinJo.