Sunday, January 9, 2011

Predicting the Cold Truth

Seems our weather dudes and duderesses need a little work on predicting nighttime cold temperatures. At least that's what we found when we did a "pop quiz" on the forecast for Saturday night's low. The actual low temp was 4 below (recorded at the airport, and yes I know the airport is warm and nobody lives there, but it's a reasonable gauge for towns inside the 494/694 concrete loop and that's what the forecasters had in mind). As of Friday night, these were the predicted lows for Saturday night (as noted here):
WCCO - 11 below, KSTP - 12 below, FOX - 14 below
KARE - 6 below, Star Tribune - 10 below, NWS - 10 below
Most weather outlets touted "coldest night of the season" for Saturday. It didn't happen. We're not going to grade that quiz, but we're here to serve notice to the forecasting clan that we're going to begin tracking the forecasted high for next Sunday. There's just been too much talk of the upcoming arctic blast not to shine the light on it. For the range of forecaster predictions for Monday/Tuesday's snow, click here.


  1. It was -11 when I got up Sunday. So I would say they were right on the money. Concentrate more on grading them accurately and less on trying to prove them wrong. My 2 cents.

  2. I've been thinking for some time that you should be looking at temperature predictions, particularly in the long-term forecasts. I would be easy to quantify the results, although determining a letter grade is somewhat subjective.

  3. I think it may be best to focus on daytime high temperatures as that tends to be the same in the city and suburbs.

  4. Interesting. I just posted a comment on today's Star Tribune/Paul Douglas weather blog because of significant discrepancies in upcoming temperature forecasts - from the Friday blog to today's blog there is a 6 day shift in this "record cold snap" that people are talking about and the posted "15 day forecast" on the same site lists temperatures at or above the freezing mark for the exact same time period that this record cold snap is being predicted in Douglas' blog?!?. I think they NEED to be graded - so they would be more mindful about what they say - people really listen to them and the weather is a highly emotionalized topic and all these different models predicting different things has gotten out of hand in my opinion.
    I for one am grateful to this site for posting all the forecasts in one place - that alone shows how inaccurate it is because there are such differences between stations/sources - not a very exact science, that's for sure. Reviewing the different forecasts here gives a broader perspective about how big of a range it might be. Obviously looking at the Star Tribune material just a few moments ago, there is as much as a FIFTY SEVEN DEGREE DIFFERENCE between what is shown/stated in the blog comments and graphics and in the actual posted 15 day forecast (at this moment, that is specific to what is shown for low temperatures of -33 versus +24 for 1/23)! Now that is ridiculous.

  5. Indeed, half the battle -- maybe more -- is gathering the actual predictions. There are many inconsistencies. The information (text) on the TV station's websites often differ from what was presented on TV. In general, it seems, MPR provides the basic National Weather Forecast, but sometimes Paul Huttner's information varies from that. And while I agree that Paul Douglas provides interesting information, he's inconsistent in how he presents information as well. I guess it's like grading papers when you can't really make out the handwriting!

    Thanks for your nice comments. This blog is what it is -- nothing more, nothing less.

  6. It's not hard to find the temps for Lakeville, Lake Elmo, Crystal, Blaine, and Flying Cloud airports. Why not include all of them instead of the one site you've decided is representative.

    More proof that this site is for entertainment and has zero credibility.


  7. Early information on the upcoming cold wave forecast is showing that Paul Douglas consistently tends toward the extremes. Check it out here: