Do certain areas have less predictable temperatures? I started out by calculating difference in the forecast and observed temperature for 236 cities across US states and territories across 167 temperature recordings. I figured this would be a fine baseline as a measure of forecasting accuracy. A forecast of 2 degrees Fahrenheit isn’t twice as hot as a forecast of 1 degree Fahrenheit, but a difference of 2 degrees IS twice as wrong as 1 degree.
After averaging down by city and trimming down to focus on the continental US, I mapped out a basic plot on a map. My eyes were immediately drawn to the Rocky Mountains so I quickly added elevation as a factor into the mapping (early map omitted to reduce clutter. Plus this is the important one.)
It seemed like there was some correlation between forecasting error and the elevation of a city, but since it felt a little difficult to truly discern any patterns on a map I took to a scatterplot and graphed elevation against forecasting error directly.
There seemed to be some positive association between elevation and inaccurate weather forecasting where cities with the least error in forecasting also happened to be some of the lowest elevation cities in the continental US. While I wouldn’t exactly describe the association as ‘strong’, it’s not exactly inconclusive either.
This short reporting may not have much to say on the variability of day to day weather or any quality of life that comes with that, but at the very least if you want to be able to trust your local forecasting, stay away from any mountains!