13 January 2014
Numerical weather models tend to become very unreliable beyond 5 or 6 days but meteorologists have discovered a trick that helps in long-range forecasting. At least a bit. We still cannot pin down what day a storm will hit at a particular place but we can get a good idea of whether or not the pattern will be warm/cold/dry or stormy.
Here is the trick: Run the weather model multiple times using very slightly different starting points. To use physics and maths to model the atmosphere we have to give the model a starting point, and since we do not know the exact temperature pressure etc over every square kilometer of the planet, we have to do a best guess. Since weather is chaotic and follows chaos theory, a small change at the start will lead to major differences as you go out farther into the future.
For five to 10 days however, you will usually see that most of the model runs tend to be pretty close with a couple of outliers. The differences magnify with time, but taking an average of the models can yield some surprisingly good information. Studies have shown that this method is more accurate on average than using one or two models, and it is now being used in hurricane forecasting, severe weather threats, and rain and snowfall forecasting.
Forecasters call all these model runs the ensemble and we even have short-range and long-range ensembles now.
So, with this in mind, I can say with some confidence that the cold in the Eastern U.S.A is about to return. Initially it does not look as severe as this last batch and that makes sense since we have drained most of the really frigid air out of the High Arctic…for now at least.