As all good meteorologists should, verifying your own forecast is essential to improve your skills and forecast accuracy. At the start of every month, I go through my forecast values for the month prior to see how far off/accurate I actual was!
Truth be told, this past month was overall really good, but I did have a few serious busts!
I calculate the overall +/- error for each time period (night 1, day 1, night 2, day 2, etc...), and can also compare to past months performance. Here's the breakdown of my November forecasts... (keep in mind, these numbers are only compared off of the data from the Burlington International Airport. I do keep (looser) tabs on all other regions within the forecast area, but generally don't keep statistics for every point of data.)
Night1/Day1 Night2/Day2 Night3/Day3 Night4/Day4 Night5/Day5 Night6/Day6 Night7/Day7
1.25/1.92 2.17/2.83 2.83/2.83 4.58/3.5 4.92/4.17 5.5/5.67 7.17/5.33
So, what this means in simple terms, on average, I will be within 1-3 degrees of the actual temp for the first 6 periods (night 1, day 1, night 2, day 2, night 3, day 3), aka the first 3 "days" of the forecast, but after the 4th "day", accuracy goes down a little bit with an average error of 3.5-7.5 degrees.
As you might naturally expect, the numbers get larger farther out into the forecast (meaning accuracy goes down). That's certainly what I would expect, but am constantly striving to improve accuracy.
Compared to last month, I definitely improved in the Night/Day 1-3 forecast period. In October, I was +/- about 3 to 4 degrees in the first 6 forecast periods, with 4-8 degrees error past "day" 4.
Just something you might want to know! An honest and open forecaster is one (I hope) you can trust!
One more thing! As you may already know, Fox 44/ABC 22 News now provides a 7 day mountain outlook as well (click to check it out)! This is something we launched on the air in late October of 2012. I am also keeping statistic for the mountain forecast, too! How? Well, I am doing more of a subjective forecast verification. I will cross check our mountain forecast number with 2 sources...the Mt. Mansfield data, and the Mt. Van Hoevenberg data in NY. The 2 mountain points offer a range in temps that cover an elevation of 2000' (Van Hov.) to the tallest peak our forecast area, 4396' (Mt. Mans). It's only been about a month and a half since we've been forecasting and verifying here, so we are still learning and developing a plan to make our forecasts more accurate for this challenging area. Once we get some more data going, I will likely post my findings on this data as well.