If you have read through posts on my website, you have probably read about how I track my forecast error. While this isn't something that is required of me, I do like to break down trends and use those stats to make myself a better forecaster.
Back in November of last year I moved here to Columbia, SC to start a new job, during the move, I lost my old forecast spreadsheet so this gave me a great opportunity to make a new one from scratch.
On top of formatting data out differently I decided I wanted to also track the precipitation part of the forecast (we will talk about that later). Now, that we are more than 4 months into the year, I feel like there is finally a large enough sample size to look at the numbers and see how things are going.
TEMPERATURES
Looking at things, I have to say, I am pretty impressed with myself. This graph averages the around 40 days I have made a forecast in 2022. With the winter months (Jan & Feb) normally resulting in higher errors compared to other months, the fact that out to day 7 the average error is just around 4 degrees is pretty impressive! As always, the closer you get, the more the error goes down. Highs are under 2 degrees off on average which is fantastic!
Looking forward, with summer weather on the way most of these numbers should go down quite a bit thanks to less variability in temperatures during those months.
PRECIPITATION
New this year, I am now tracking precipitation. I am testing out how to calculate if rain verifies given it is not really based on one specific point. I honestly think there will always be some holes in this calculation but, I want to see what a year's worth of data will show.
Here is how I set things up:
In the forecast, if I put <=20% for the chance of rain, then that equals no rain in the forecast. Of course, that means that anything >=30% means rain is in the forecast. From there, I use the climatology report from KCAE to verify if that forecast is correct or not. I think I will probably tweak things in the future but we will see how things progress with this method.
This graph once again shows around 40 days of forecasting. I think the results are pretty interesting. Using my calculations, the rainfall forecast varies from being very accurate (high 80s on day 1) but only slightly trends downward through day 7. There doesn't seem to be a clear trend with this which means my calculations need to be changed OR it shows the random nature of precipitation. I will definitely update this portion of the spreadsheet throughout the year!
I am still keeping track of things with the Cory Error Index as well as some other ways of sorting through the data. I'll update those likely a little bit later on in the year once those start picking up on trends!