Thursday, October 20, 2011

A Statistician Weighs In On the Upcoming Twin Cities Winter

One of TMF’s followers, Rigil Kent, is a weather enthusiast and statistician. He provided a prediction of winter snowfall and temperatures based purely on statistical analysis and the historical effect of La Ninas. He has compiled datasets over the years from weather records from a variety of different sources, including NWS, Accuweather, etc

“I thought I would try a systematic approach to this,” says Rigil. “I took the average snowfall of the La Nina winters, plus a random correction taken from the statistical distribution that best approximates winter snowfall in the Twin Cities since records started.” Rigil came up with a result of 62.1 inches of snow.

We asked Rigil to take a shot at predicting temperatures as well. “Temperatures are a bit trickier as the official NWS means are updated every 30 years or so. To keep things simple, I used the current averages and looked at the discrepancies observed during La Nina winters plus some random corrections.”

The results were enough to send chills down the spine of even the most hardy winter Minnesotans. Based on the statistics, Rigil predicts February and March to be far colder than average with December, January and April temps running slightly below normal. Here are his monthly temperature predictions for November to April:

November. Minimum temps .9 above average; maximum temps .4 above average
December: Minimum temps .5 below average; maximum temperatures 1.1 below average
January: Minimum temps .8 below average; maximum temperatures 1.0 below average
February: Minimum temps 3.8 below average; maximum temperatures 2.9 below average
March: Minimum temps 4.4 below average; maximum temperatures 5.3 below average
April: Minimum temps .4 below average; maximum temperatures 2.2 below average

Rigil’s quick to point out that his model is a purely statistical one and does not take into account any meteorological variables for now. He’s currently working on a model to predict snowfall for the Twin Cities year by year that takes into consideration several indices like ENSO (La Nina/El Nino), the Pacific Decadal Oscillation (PDO), the NAO, etc.”

Weather/statistical geeks may be interested in attending an upcoming lecture at the University of Minnesota. The lecture, conducted by Professor Adrian E. Rafferty from the University of Washington, is titled Probabilistic Weather Forecasting Using Ensemble Bayesian Model Averaging and takes place on Thursday, Oct. 27 at Ford Hall 115 at 3 p.m. The lecture will look at ways to post process numerical model ensembles (GFS, etc.) to obtain better calibration. As Rigil points out, “To a Bayesian statistician with a passion for weather like myself, you really can’t ask for more.”

A reminder that you can follow The Minnesota Forecaster on Facebook and Twitter.


  1. Hmmm....I find this very interesting, from the Climate Prediction Center, this seems to agree with Rigil's thinking, along with the forecast I put out Nov-Jan.

    First link is for Nov-Jan, second link is Dec-Feb, the third link is Jan-March.

    It appears that Nov-Jan for that three month period has equal chances of being above or below normal.

  2. According to the Minnesota Climatology Working Group, which btw hasn't updated their site to drop the averages from the 70's (burr) and to included 2000-2010, the average high temps by mid Feb is around freezing, and by mid March it is close to 40°. If the real cold doesn't set in until late Jan-March..I can live with that, temps way below normal for DEC and Jan would not be a good thing.

  3. Yea, I was thinking the same thing. The worst time of year to be way below normal is December and January. Thanks for sharing the links.

  4. BTW the new ENSO forecast is out, scroll down and you can see how much stronger the Climate Forecast System (cfs version 2, the USA model)is on the La Nina as opposed to the other models. Curtsey of the IRI, Thanks to Duane for sharing this link. If that would verify it will be nasty.

  5. A statistical prediction (62.1 inches) without an error bars/confidence intervals/call it what you will is not a statistical prediction.

  6. @John...
    confidence intervals have sense if you are inferring data from a sample and try to project it to the overall population (like polls etc.)
    In this case there is no sampling of data. All data available have been used. That is why the NWS averages don't have confidence interval around them. What you are probably thinking is the model prediction fit, which of course is not 100%... but that's a totally different story...

  7. Paul Douglas' blog this morning, he stated "No major storms (or snow flurries) through the first week of November." From there I checked out the 15-day and found this:

    Friday November 4, 2011
    Colder with snow, some ice early
    High: 39 °F Low: 26 °F

    Which is it, Mr. Douglas? Obviously there's a reason most forecasters don't go out beyond 7-10 days with any sort of confidence. I also know there are more people behind the scenes in the Strib weather department, so I will assume the 15-day forecast is determined by others or a computer and not necessarily associated with his blog. Still, if he's the face of the weather department, shouldn't there be a correlation between his blog and the forecast? In the least, they shouldn't be a 100% contradiction of each other!

    Any models hinting at this weather change? I was hoping for a few more warm weekends...

  8. Bring on the FURY!!! Cold, Wind, Ice, SNOW. WINTER WINS!!!

  9. Just a heads up, Dave Dahl's snowfall forecast will come out on Thursday Oct. 27th during the 10pm news

  10. Just heard Dave Dahl on the radio.
    He said almost a repeat of last year.
    He's going for 75".

  11. Not too big a surprise there. I should accumulate the various winter prognostications ... let's see MPR had one, Paul Douglas had one, and now DD. Any others to capture?

  12. accu weather is here

  13. Intresting. The cfsv.1 and ths cfsv.2 think thal la nina will strenghten drastically in the next couple of months, perhaps stronger than last year's. What is now happening out east could happen to us.

  14. 62.1 inches of snow is a lot. i am moving all the time and it was the 1st time i've seen so much snow