So, if you have stumbled upon my website in the past you probably know that I keep a forecasting spreadsheet that I update each time I do the weather. In 2022 that total was 118 forecasts! This blog post will explore how my forecasts over the span of the last year played out and how they went against my expectations.
In the past, my forecasts played out pretty similarly: 1. High Temperatures had a lower overall average error 2. The day 1-4 error was pretty low overall With that being said, I went ahead and gave myself an expected grade for what my forecast would be before actually compiling my data. You can also see the actual grades I assigned to each category. We will go over each down below.
THE RESULTS
LOW TEMPERATURES: GRADE B+
Let's start off with the lows. I was honestly a bit surprised by the results! The overall average error was more consistent than the high temperatures and combined for overall less error. This broke the trend I had seen over the last 2 years of keeping forecasts. As with previous years though, the day 1-4 forecast netted much lower error compared to days 5-7.
Looking below we can see how the low-temperature forecast skewed compared to the actual temperatures. According to the information below, we can get 2 main things from my forecasts. The day 1 forecast tended to be too warm while all other days tended to be too cold. This actually helps a lot in trying to correct my own bias going into the future!
HIGH TEMPERATURES: GRADE B-
I am going to be honest here, I thought about giving myself a C+ for this category. I was somewhat disappointed by how high my error was from days 1-4 specifically but also overall. I decided to cut myself some slack given Columbia is an inland location compared to the coastal weather I was forecasting for the past 2 years. Drier conditions will always lead to more temperature variability which inherently have more error. Even with that being said, I think there is plenty of room for improvement. Negatives aside the day 1-3 forecast averaged less than 3 degrees of error, acceptable but most certainly not bad at all!
Trying to look at forecast trends like we did with lows, things get a bit messy with highs. There isn't really a clear trend for how days skew so it will be a bit tougher to identify why I was so off in my forecasts.
RAIN FORECAST: GRADE A-
This is a new category in my forecast error spreadsheet that I began tracking this year. To explain how these numbers were determined, any day forecast with a rain chance of 30% or greater was hard coded as rain being expected in the forecast. To use verification any rain (trace or above) is counted as rain.
The rainfall forecast is what I expected to be the worst part of my day-to-day forecast but as it turns out it was actually really solid! Given rain forecasts take into account precipitation timing and the variability of location when it comes to where it actually rains, I thought accuracy would be pretty low.
Turns out I was pretty wrong! Day 1 and 2 had over 80% accuracy which is great! As you get further in the forecast accuracy went down BUT given day 7 has 67% accuracy is much, much better than I expected going into this year!
That wraps up my forecast error blog post for 2022, let me know if you want to know anything more about how I forecast the weather as we go into 2023. Thanks for reading and keep posted for more content!
留言