NOAA has come up with a metric for measuring the accuracy of Punxsutawney Phil’s prognostications. They compare his forecast to whether the actual temperatures of February and March for that year are above or below normal. This is a defective analysis, guaranteed to make the groundhog fail, for two reasons:
- Phil’s forecast is not “warmer than average” or “cooler than average”. It is “6 more weeks of winter” or “spring is right around the corner.” Since spring is technically 6 weeks away, cooler and/or average weather would be indicated by seeing his shadow, and MUCH warmer than average would be indicated by not seeing it. I would submit that Phil’s record suggests that what he means by “warmer” is in the top 15% of warmest February-March periods. That is, he makes that prediction less than 15% of the time.
- Global warming requires that the comparative data be adjusted on a yearly basis to compensate. NOAA is using the average of 20th century data. Obviously skewing comparison data to the warm side.
Both of these reasons tend to reduce the accuracy of Phil’s prognostication, by giving a “result” of spring, when Phil’s result was more winter. (Of course, when analyzed this way, maybe his record is even worse. )
Maybe they fear the competition. As reported in LiveScience back in 2011:
…we checked Phil’s performance with David Unger, a meteorologist at the National Weather Service. It looks like Phil probably won’t be getting a job at the NWS any time soon, either.
“It’s extremely difficult to give an estimate of how accurate climate predictions are,” Unger told Life’s Little Mysteries. “But compared to the terms with which Groundhog’s Day predictions are made which are if the weather will be mild or not mild then if our forecasts are about 60 percent accurate or higher, then we consider that to be a good estimate.”
This is just unfair to the poor little whistle pig. Vox has even come out with a slanderous piece accusing the woodchuck of evil intentions! What???