Use the hashtag ‘valleywx’ to talk weather with us on Twitter.
Ensemble Signal of Cooler Weather Ahead
Disclaimer: This post gets a little bogged down in charts and technical stuff. Proceed with caution 
Preamble: Ensemble Forecasting (skip down for a discussion of the seasonably cooler weather possible in the coming weeks)
You’ve probably heard us talk about ensemble output from weather models, usually in the context of tropical forecasting, but their utility extends well beyond the tropics. In fact, ensembles are most often used in long range forecasting, when the best you can hope for is an accurate diagnosis of uncertainty and a likely range of temperatures.
Briefly, here’s how ensembles work: First, you need to know that the accuracy of forecast models depends on the initial data fed into them and the equations that govern their forecasts. Both the initial data and equations are imperfect and contain errors for several reasons that include, but are not limited to, gaps between measurements, inaccuracies in the instruments themselves, and the inability in mathematical equations to perfectly represent the complexity of the atmosphere. Ensembles seek to diagnose these errors by “perturbing” the initial data input into the model and seeing the how the model responds. The different solutions from the various perturbations are known as ensemble “members.” Comparing how these members differ from each other with time gives the forecaster information on the predictability of the atmosphere and a range of likely scenarios (e.g., a range of likely high temperatures).
The spread between the members tends to grow with time, indicating the model forecast errors are growing and that the forecast certainty is decreasing. How fast the errors grow depends largely on the weather pattern, with some patterns being more predictable than others. This error growth can be seen visually by use of a spaghetti plot, which displays the forecasts from all the members on one graphic.
Example of a spaghetti plot ensemble forecast of upper-level winds from the Global Forecast System model.
Notice how the individual lines (forecasts) start out tightly grouped and become more spread out as the forecast period gets longer. As you probably expect, the uncertainty in the forecast for 15 days is much higher than the 4 day forecast.
Of course, ensemble forecasting does have limitations and usually does not add substantial value to the forecast until around 4 days out.
Ok, now let’s talk about what the ensembles are saying about the weather over the two weeks…
Cooler Weather Signaled by the Ensembles
We’ve been talking about a strong cold-front set to arrive this weekend. The Global Forecast System (GFS) ensembles are showing a strong signal that after the front passes, we won’t be seeing a significant warm-up between Saturday and the official beginning of fall.
The graph below shows the ensemble probability of the high temperature in Huntsville being above 90 degrees through September 22nd.
GFS ensemble forecast of probability of the afternoon high temperature exceeding 90 degrees.
Notice, after tomorrow, the odds that we eclipse the 90 degree mark during the next two weeks are slim to none.
How about another threshold? Let’s look at the odds of us getting above 85 degrees during the same period.
GFS ensemble forecast of probability of the afternoon high temperature exceeding 85 degrees.
We definitely have a better shot at getting above 85, especially toward the end of next week, but its still only a 50/50 shot at the highest, according to the ensembles.
For perspective, here’s a graph of the number of days we’ve seen temperatures at or above 90 since 1959 during this period. It’s certainly not extremely uncommon. After all, the 30 year average high for the next few days is still around 87.
Bottom line, it looks like temperatures should remain at or below the seasonal averages for the next couple of weeks.
