From Meteorologist Steve Glazier,
One of my pet peeves is a "bustable" forecast. I call it "bustable" in the sense that I can get the weather forecast wrong very easily. People often ask me, "So how accurate are you? How often are you right?"
There are two answers to this. One answer can be, "Never." I'm technically never right. If I say it's going to be 73 today and it turns out to be 72 for the high, I'm wrong. The second answer can be, "Most of the time." I'm sticking with answer number two here, because we won't be able to pinpoint the weather to the hundredth of an inch of rain. However if we're in the ballpark, I say it's a hit.
Here's what I mean by 'most of the time.' I'm going to share some information with you that I just calculated. It's my accuracy verification. Every day I verify my highs and lows for Burlington. I'll list how I did below. The +/- is my error. That being said, when I ran through all 20 of my forecasts in September, the average +/- on any given day is the difference from the forecast to the actual.
Overall in September, on any given date, my error was +/- 3.6° for highs and lows separately. Of course, that varied between days. For instance Days 1-3 were much better than 4-7, because short-term forecasts are more accurate than long-term (usually).
Today's high: +/- 1.4°
Tonight's low: +/- 2.3°
Tomorrow's high: 1.7°
Day 3 low, high +/- 3.3°, +/- 3.0°
Day 4 low, high +/- 3.6°, +/- 3.6°
Day 5 low, high +/- 4.3° +/- 4.3°
Day 6 low, high +/-4.7° +/- 4,7°
Day 7 low, high +/- 5.7° +/- 4.3°
September was a harder month to forecast than previous June, July, and August this year. In fact, here are my month-by-month numbers since May. These numbers are my average weekly departure from what actually happened. So it's the total of 13 lows and highs that I forecast.
The curve bottomed out in August, with the least error for 2012 being that month. September rose by quite a bit. That's because our weather pattern has gotten much more active. There have been more "bustable" forecasts. Generally in these transition months of March-May and September-November, the forecasts can bust more easily. This applies to my data when I look at the numbers from May to September.
My goal is to eliminate these busts. The month of October is looking just as active as September was. Here we go, let's try to nail these numbers down!