Cloudy and Cool

Since I’ve been talking about the computer forecast for Thanksgiving Day for 16 days, the only responsible thing to do is review how well the forecast turned out.

Here’s the map from Thanksgiving morning (7 a.m. EST) (Steve at Weather Talk also talked about the weather for Thanksgiving in Thanksgiving Forecast: The Final Verdict):

There was a weak storm in the upper Midwest, which was accompanied by a little snow in the northern Plains, along with some rain along the East. There was also a little rain in the Pacific Northwest. There were no strong storms, which is in stark contrast to the original computer model we looked at (from November 10):

Five days later, the computer model had begun to back off on the strength of the storm, as is indicated by this map, which shows no big storm–a much better forecast than the original map but still not accurate on the details:

Forecast for Thanksgiving Day (12z run of GFS from November 15, 2009)

The next map we looked at (from a week out) was reasonably accurate, but that’s to be expected:

Conclusion

Consider this an exercise an example of the big changes that often take place in the long-range operational computer models, but remember, computer models designed for long-range forecasting, often months in advance, are not based on the same mathematical equations, so the same rules do not apply. (See December Forecast.)

–Paul Yeager