Sunday, February 6, 2011

NSSL WRF performance over OK

I grabbed some 2 meter temperature forecasts from NSSL's WRF 4 km model simulation from 00 UTC 5 February 24 hour forecast to compare with what happened on Saturday.

The model did not do terribly well but it highlighted the issue with snowpack. Lets go to the pictures:

The images depict the 2m temperature in 2 hour increments from 18 to 22 UTC; with the 22 UTC image being the warmest time of this day. What stands out is the warm air over the TX panhandle which does not expand rapidly into OK. Just east of the OK panhandle it warms rapidly, but it does not expand and penetrate eastward. Over OK, the cold patch, which aligns perfectly with the storm total snowfall and thus snow pack, does not appreciably change shape but it does warm a bit. Clearly the model had a poor representation of the snow cover, both in areal extent and depth.

From the previous post, the high temperatures even over the deep snow pack near Tulsa in the core of the model cold patch, got to near 40F ... a difference of 20 F!

The situation eases later as night arrives by 00 UTC. Below are the 24 hour forecast from the model and the initialization from the next cycle.
The differences between these 2 images is difficult to discern but they are still large, because of the eastward shift of the OK cold patch and the eastward extent of warmer air. This is an interesting case where I would expect this type of model to perform better. Diagnosing the evolution of the snow pack and the low level temperature tendencies both aloft and from within the model physics (boundary layer scheme) should shed some light on why the model performed poorly.