Subseasonal Convection-Permitting Global Weather
Prediction
Global weather prediction tends to fade rapidly after roughly 10
days, even when surface conditions (like warm sea surface
temperatures) should be providing some predictability.
A potential problem for extended forecasts (and shorter-period
forecasts as well) is the inability of the convective
parameterizations of current numerical prediction models to
properly simulate convection. The obvious solution is to run
globally at sufficiently high horizontal resolution
(convection-permitted grid spacing, 4 km or less), although this
takes an immense amount of computer resources.
In this project, sponsored by NOAA/NWS, we have run the NCAR WRF
model globally at 3-km grid spacing for four one-month
periods. We have also tried other configurations (e.g., high
resolution in the tropics along, no convective parameterization at
12-km grid spacing). The results have been very promising,
showing that convective-permitting resolution results in much
better precipitation statistics as well as improved general
predictability in the thirds week. We plan on repeating this
work, for more cases, using the new NOAA/NWS FV-3 system.
