Gordaen’s Blog

Everyone knows that weather predictions are frequently inaccurate, so I decided to do a little investigating last month. I tried to guess what the weather would be like over the next 8 days, then I took screenshots of the webpages from five major weather sites (AccuWeather, NOAA, Weather.com, Weather Bug, and Weather Underground) and plotted the data in a spreadsheet. One page was for temperature and one was for conditions (e.g., rain).

Temperature
For the temperature predictions, I was primarily concerned with daytime temperature (i.e., the high for the day). I did include the lows just for reference, but all the color coding is based on how far off the high predictions were. Notice that only one of the five sites actually correctly “predicted” the high for a single day! The red shading shows which predictions were off by four or more degrees, so you can quickly tell that predictions beyond just a few days are wrong. AccuWeather actually claims to be able to predict up to fifteen days in the future, but the data suggests that the predictions after four days are inaccurate.

At the far right of the chart is the average number of degrees the predictions were off by. The first column is the average for the first four days (the amount of time I had data from NOAA for and, coincidentally, the amount of time most of the sites were semi-accurate for). The second is an average over the entire row. As you can see, the majority of the predictions became quite unreliable.

Rain
The rain predictions were a bit harder to chart. Weather forecasters like to use terms such as “Possibility of showers” in order to never be wrong. Some of the sites actually used percentages, which was a bit nicer, but not a single one even guessed at the amount of rain that would fall. Notice that the sites almost all gave a pretty fair chance of rain for every single day, yet only three of the seven days rained and two of those three were minimal amounts. My predictions were based on “sprinkles” every other day, randomness, and sarcasm. Maybe I’m full of myself, but I think that my predictions were the most accurate.

Conclusion
The results pretty much show what I was expecting: weather predictions are reasonably reliable in the short run, but near useless for the long run. If you’ve lived in a place for a few years, you can probably guess about as well as the weather forecasters. As a bonus, if you’re good at guessing, you won’t have to looking at the hideous websites that I used for this data.