Monday, January 31, 2011

The Extended Range


Below is a link to the NWS Hydrometeorological Prediction Center’s (HPC) extended range forecast discussion. Although it is technical, it serves as a good representation of what really goes on in weather forecasting/predicting/guessing 72+ hrs onwards. If you read it, you will note that most of the discussion written by the meteorologist centers on forecast models and their behaviors/biases/handling of major weather features and the impact in terms of sensible weather that could result. These are the 5 models primarily used for extended range forecasting :
  • GFS (US)
  • ECMWF (European)
  • CMC (Canadian)
  • UKMET (UK)
  • JMA (Japanese)
 A GFS Ensemble run uses the same data input into the operational/deterministic model, but differs because instead of 1 (operational) forecast, an "ensemble" is 20 different calculations (perturbations). Essentially they are “what if” scenarios - when the ensembles are similar, the model is signalling a higher degree of accuracy. Here is a graphical representation of an ensemble run from earlier today at forecast hour 126 which would be next Sat I believe. The picture (Click to Enlarge) shows 6hr Precip, 850mb Temps, and Pressure.
Sunday Jan 30 18Z GFS Ensemble - Forecast Hr 126*


 *From Penn State's Meteorology E-Wall, found here: http://www.meteo.psu.edu/ewall/ewall.html

Essentially - granted this is 5 days away - the ensembles agree that there is going to be a low pressure system forming in/near the Gulf of Mexico. But as far as timing, critical thickness temperatures (rain vs snow, etc), and amount of precipitation we can expect here is impossible to predict with any certainty. This is the reason behind the National Weather Service's use of percentage probabilities for a given event. A 50% chance of rain doesn't mean maybe yes, maybe no. It is the probability of a given area within one of their forecast zones (or subzones) of rain falling. If you look at the NWS forecast for Salisbury on Saturday the probabilities are somewhere around 30% or thereabouts. All the major models are indicating that a storm will be in the making or occurring around Saturday; but with 120 hrs until potential impact the chances of model correction in the time/speed/strength of the low pressure system that doesn't even exist yet are way too high to place the probability above that. This is especially true when it comes to predicting winter weather events in the NC Piedmont because we are usually the borderline between the cold to the North and warmth from the South associated with low pressure systems moving north from the gulf region. Perhaps one day a computer will spit out a perfect prediction for the weather in your backyard five days from now; we simply need someone to write a flawless algorithm or two...

Tangent aside, do read the discussion - an actual meteorologist analyzing model data and accuracy - and while computer modeling is (in my opinion) the #1 forecasting tool - you will see that is not all he discusses in his extended range discussion.

Jump to Discussion:  http://www.hpc.ncep.noaa.gov/discussions/pmdepd.html 

Last note - since today is another good example of the impact cold air damming has on our weather I am going to try to do a post about it if I can find some info I can link to. Enjoy the gray for a few days.