Accuweather forecast accuracy: a preview

Here’s a preview of a little something I’ve been working on: An analysis of Accuweather’s 25-day forecast for self-consistency and accuracy.

Here’s what the forecast high temperature looked like for today, December 10, 2012 for zip code 19103, on the 25 days leading up to today, as compared to the historical normal high for today, and today’s actual temperature:

Put differently, let’s look at how much that forecast was off from the actual temperature we ended up with today.  But, since a forecast is only useful if the forecaster sticks to it (a forecast 14 days ago that’s pretty close to today’s high temperature is no good to us, if the same forecast was 20 degrees off just 2 days later), I’ve added in a bit more data here too:

In this particular case, if more than a week ago you’d wanted to know what today’s high temperature would be, you’d have been much better off by just looking at the normal temperature for December 10th (even though it was 13 degrees off!) than you’d have been by starting to follow the Accuweather forecast.  Only the forecast 7 days out and closer got better.

For fun, here’s how Accuweather described the weather forecasts for today, a day that ended up foggy and drizzly:

24 DAYS OUT Snow; rain mixing in
23 DAYS OUT Mostly sunny
22 DAYS OUT Mostly sunny
21 DAYS OUT Mostly sunny
20 DAYS OUT Mostly sunny
19 DAYS OUT Considerable clouds
18 DAYS OUT Considerable clouds
17 DAYS OUT Considerable clouds
16 DAYS OUT Considerable clouds
15 DAYS OUT Mostly sunny
14 DAYS OUT Mostly cloudy
13 DAYS OUT Periods of rain
12 DAYS OUT Rain mixing with snow; colder
11 DAYS OUT Snow and ice, then rain
10 DAYS OUT Colder with periods of rain
9 DAYS OUT A mix of snow, ice and rain
8 DAYS OUT Mostly sunny
7 DAYS OUT Mostly cloudy
6 DAYS OUT A shower possible
5 DAYS OUT A couple of showers possible
4 DAYS OUT A couple of showers possible
3 DAYS OUT A couple of showers possible
2 DAYS OUT Showers, mainly later; warmer
1 DAYS OUT Couple of showers, t-storms
0 DAYS OUT Thundershower

Much more to come in a few weeks or a month or two, after I’ve collected some more data.

One thought on “Accuweather forecast accuracy: a preview

  1. Pingback: Accuweather long-rage forecast accuracy questionable | Josh Rosenberg

Leave a Reply