Winter is a very tough time to forecast weather, especially compared to the summer months. I say this in regards to temperature forecasts. On the other hand, severe summer thunderstorms weather can be much more challenging compared to winter storms because of how quick storms can get severe and the variability of where they may form. However for today's blog I will focus on difficult winter temperature forecasts.
On the left-hand-side you'll see three images. I know they're not the prettiest to look at, I whipped them up in Microsoft Excel (nothing against that program, more of a hit to my graph creativity!).
The first is the raw data of January 2013 for my forecasts. I forecast 13 time periods, 7 high temperatures and 6 low temperatures. I make my forecast between 3 and 5 a.m. each day when I get into work. It's valid for the next week as I update it and verify what actually happened. For the sake of simplicity, I forecast exact numbers for the Burlington Int'l Airport and verify high and low temperatures at that location. When I go on TV I adjust the numbers on the seven day accordingly if the rest of the region is going to be much colder than BTV then I'll drop it a few, or vice versa for warmth. Plus I do forecast town by town (dozens) for day of high temperature and day of upcoming low temperature.
If you can read some of the small numbers on the first image there are a lot of single digits but also a lot of double digits. Those double digits are what I consider busts, meaning I was off by ten degrees or more for that forecast period. You'll see more double digit numbers toward the right hand side of the image compared to the left, as weather forecasts are much more inaccurate between days five and seven.
Day 1 HI is the high temperature today. Day 2 LO is tonight, going into Day 2 HI which is tomorrow's high, then so on and so forth through Day 7 HI.
The second two images show a breakdown of high temperature forecasting by day versus low temperature forecasting by day. Generally low temperatures are harder to forecast than high temperatures, given the data I provided here for Jan. 2013. Both graphs follow a fairly linear pattern of more inaccuracy with time, however it's not a constant rise. Notice for my low temperature forecasts, the inaccuracy peaks on day four while days five and six show lower numbers. The high temperature line follows that pattern, but jumps to six degrees inaccuracy for Day 7 HI. I'm not too sure by days five and six are lower than four.
So what does this mean to you, the viewer and consumer of my forecast? Take my forecast numbers with a degree of 'margin of error' if you will. If I forecast a high of 30 F today, it will likely fall between about 27-32F or 28-33F. However if you're wondering about the weekend's weather on the preceding Monday (six-seven days away) and the forecast high is 25F, it may end up being 19-31F, sometimes closer sometimes farther.
Our 7-day Mountain Outlook (which you can find exclusively from our TV station on Fox44/Abc22) is doing pretty well! I've heard very good feedback regarding it. I just verified February 7-11, 2013. Out of the five day period, four were hits and one was a miss. However that's just for high temperature forecasting, and verifying our single number between Mt. Van Hoevenberg, NY (2,000 feet elevation) and Mt. Mansfield, VT (~4,300'). Previous verifications show that we're hitting our numbers between five to six times in a seven day period. We haven't verified precipitation or sky cover yet.
I do have good news though. I started this verification process last year with the new shift of morning meteorologist. January 2012 showed an average inaccuracy of 63.4F for the whole forecast period, while January 2013 was an error of 54.2F. Maybe that's because weather patterns were easier to forecast? I hope it's because I'm getting better :) Thanks for reading -
Meteorologist Steve Glazier