By Matt Balitewicz on Friday, December 14, 2007 - 06:29 am: Edit PostJohn,
Do you rely more on one than the other? Does reliance on surface observations become more important when forecasting for that day's weather, than using models?
Thanks.
By John Dee on Friday, December 14, 2007 - 10:08 am: Edit PostGood question Matt-
For almost all aspects of forecasting I rely on the model data far more than the surface data. The main reason is that the upper air conditions determine about 95% of what happens at the surface. Secondly, surface conditions will change with the warming of the day or with the approach of a weather system.
There are a few instances when looking at the surface conditions can aid in a forecast for the next few hours (nowcasting), but to use surface data for any kind of a forecast out beyond 6-8 hours is not very useful.
-John
By Matt Balitewicz on Monday, December 17, 2007 - 07:51 pm: Edit PostA follow up:
Are weather baloons still used? What are the various methods used to gather upper air conditions?
By John Dee on Tuesday, December 18, 2007 - 10:05 am: Edit PostAnother good question Matt.
They are, but unlike 20-30 years ago when that was about the only way they gathered upper air data, there are a lot of ways they do it these days, including satellite imagery, ground based wind profilers and even sensors on commercial aircraft.
-John
By Matt Balitewicz on Wednesday, December 19, 2007 - 08:46 pm: Edit PostAnother question, if you will indulge me? The computer models have shown a storm developing over the Southwest for a good week now. This storm has been predicted to affect the Midwest this weekend. I recall you mentioning it in your long range forecast several days ago. When did it become possible for this type of forecasting, in which a metoerologist can see a storm system develop several days before it actually forms?
Thank you.
By John Dee on Thursday, December 20, 2007 - 11:45 am: Edit PostMatt-
For at least 25 years they have been running a model out to 240 hours or 10 days. For quite some time it was the only model, now there are are at least 3 that go out that far. Some model data now goes out as far as 384 hours or 16 days.
-John
By FRNash/PHX, AZ on Thursday, December 20, 2007 - 02:24 pm: Edit PostMuch of the increase in reach of the long range models is also attributable to the significant advancements in the last decade or so in supercomputers such as the IBM Blue Gene series, among others.
Incidentally, here's a cute one, found while browsing: Plan 9 from Outer Space er
Plan 9 from Bell Labs 
By Pete W on Friday, December 21, 2007 - 07:26 am: Edit PostYou didn't mention the much cheaper little brother the PlayStation3 Gravity Grid. Same processor core('s) just very low relative cost.
Pete