Confidence levels in forecasting
This weekend’s weather will be unsettled – meaning we’re in an active pattern that will bring clouds, showers and swings in temperature. That much I’m sure of. Sunday’s forecast carries a little less certainty, though, which is why I have low confidence in the forecast.
I’ve written before about what the word “confidence” means when it comes to forecasting, but the idea is worth revisiting since we have such a great example this weekend.
When the computer models meteorologists use to forecast the weather are in total agreement, it’s easy to have high confidence in a forecast. Generally, that scenario happens in a quiet, relatively simple weather pattern. The more complicated the pattern and the more dynamic the atmosphere, the less the models tend to agree, especially when looking more than 48 to 72 hours out.
Sometimes, the difference between the models is only in how high or low the temperature will go. Other times, it’s how much or little precipitation will fall. One model may call for sunny skies and the other shows 90% cloud coverage most of the day. When these differences occur, the forecaster has some decisions to make. Which model seems to be the most believable based on how it has handled recent similar situations? Which one seems to be seeing everything that is currently happening at the moment the forecaster is viewing them?
A skilled, experienced forecaster has a better chance of getting the forecast right in these situations because experience is often the key differentiator. For example, someone who has lived in the Triangle area for a long time will be more likely to spot the kind of forecast-busting, cold air damming-like situation that could happen on Sunday.
As of this writing on Wednesday morning, the Global Forecasting System (GFS) model is showing a cold, cloudy, rainy Sunday, at least through the afternoon. With light breezes from the east and overcast skies, showers are possible. The high temperature may not even make it to the low 40s by afternoon. It’s also showing that the temperature may jump quickly from around 40 degrees into the lower 50s between 7:00 PM and 10:00 PM as a warm front passes and the winds turn from easterly to southerly. If that scenario plays out, our high temperature for the day will happen during the late evening hours.
On the other hand, the European model is showing slower progression of the warm front moving up from South Carolina through the day. By midnight, Wake County may still only be in the mid 40s, and the area from Fayetteville south could be in the low 50s.
The difference between these two models may not seem that great, but when forecasting a high temperature for the 24-hour period of midnight Sunday morning to midnight Monday morning, the timing of that front can make a huge difference. With the GFS solution, the high temperature for Sunday could be about 55 degrees and occur close to midnight Sunday night. With the European model’s solution, the high may only be about 46 degrees.
After so many of us experienced forecasters were burned on Tuesday with that day’s warm front stalling out in the southern part of the state, it’s hard to buy into the warmer solution. Because I know from more distant past experience that when moist wind from the east butts up against our mountains to the west, it’s hard to break down the cold air that gets trapped at the surface, I went with the cooler solution in my weekend forecast this morning.
How confident am I? If I had to give it a number on a scale of one to ten with ten being absolution sure, I’d say I’m at a six. For me, that’s low confidence. I prefer to be at an eight or higher when putting a forecast in print.
As Sunday gets closer, the models should start to agree more. This morning’s model runs were already getting closer than yesterday morning’s. It will be interesting to see how Sunday plays out. One thing I can say with certainty is that if the day turns out warmer than I predicted, that’s one forecast I don’t mind busting.