Local model...summer examples.

• General impression of local model forecasts...

  They are most accurate when forcing is organized.

  Low level upslope, trough/dry line, etc.
  Mesoscale circulations like the Denver Cyclone.
  As would be expected, local model does well with terrain forced precipitation: mountain snowfall ( slide ) .
  Or, at the other end of the spectrum: Under weakly forced "diurnal upslope" model tends to be good for timing/location of initial convection.

• In general, it was found that convection became overly organized with time.
  This was not an unexpected result however for the local model used, which had a grid resolution of 10 km and employed explicit microphysics.
  So end up with a few larger sized storms vs. a greater number of smaller-scale storms. ( Forecast1 from 23 June 99 ) ( Forecast2 from 23 June 99 )

• So, given a resolution of 10 km, it is often better to look at the behavior of model generated convection on a given day, rather than the exact model solution. ( 2D loop ) ( 3D image )