Winter forecast, part III: the Old Farmer's Almanac

Since 1792, the Old Farmer's Almanac has been issuing long-range seasonal weather forecasts. This year, the Almanac is predicting that winter will be colder than average for 3/4 of the U.S., and above average over just 1/8 of the country. Only the Pacific Northwest and the upper Midwest near Minnesota are predicted to be warmer than average. For the Appalachian region that includes the three woolly bear forecasts I discussed last week, the Old Farmer's Almanac is siding with Oil Valley Vick and Kelly the woolly bear, forecasting colder than average temperatures. The Hagerstown Woolly Bears and NOAA disagree, predicting warmer than average temperatures are more likely.

How accurate is the Old Farmer's Almanac?
The Old Farmer's Almanac claims to have a secret formula developed in 1792 based on sunspots and climatology, which gives their long-range predictions 80% accuracy. I've heard a number of anecdotal stories about how uncannily accurate their forecasts are, and have always felt a vague sort of anxiety that maybe I should be checking them out when someone asks me what the upcoming winter will be like. However, the Almanac does not post any verification statistics of their forecasts. It is not hard to do a simple check of their forecast accuracy, though. Unfortunately, the results of my check and those done by several others show that there is little reason to believe that the Old Farmer's Almanac forecasts are any better than flipping a coin.


Figure 1. Observed departure of temperature from average for the period Nov. 2004-Mar. 2005. Superimposed in bold text is the winter forecast made in the 2004 Old Farmer's Almanac for the same period. The Almanac got four regions correct and eight incorrect, with two too close to call.

For example, for the winter of 2004-2005 (Figure 1), the November 2004 version of the Old Farmer's Almanac made a simple prediction of "cold" or "mild" for sixteen separate regions of the U.S. The original forecast map they presented only labels the U.S. in fourteen places, and I've overlaid these predictions on a temperature anomaly map showing what actually happened during the winter of 2004-2005. If we assume that "mild" refers to an above average temperature forecast and "cold" refers to a below average temperature forecast, then the Almanac got four regions correct, eight wrong, with two too close to call. Admittedly, I've "eyeballed" this, and it is a subjective verification. Still, I don't see any way that this forecast could approach even 50% (chance) accuracy. Their precipitation forecast fared better, with seven correct regions, five incorrect, and two too close to call. I also looked at the Farmer's Almanac forecasts for the winter of 2006-2007. They did much worse that winter, with only three of sixteen temperature forecasts verifying, and five out of twelve precipitation forecasts verifying (four were too close to call). For these two winters, the Old Farmer's Almanac made a successful forecast just 37% of the time.

Studies by Jan Null
Jan Null, a meteorologist who founded the private weather consulting firm, Golden Gate Weather in California, has evaluated the Old Farmer's Almanac predictions for San Francisco for three separate years. His first study looked at the forecasts for 1999-2000. His conclusion: "Even trying to be objective and giving the benefit of the doubt to cases that were close, I found last year's forecast from the Old Farmer's 2000 Almanac for San Francisco to be laughable at best and abysmal at worst. The Old Farmer's Almanac was wrong on their monthly temperature forecast 8 out of the 12 months (67%) and wrong on their rainfall forecast 5 of the 8 months evaluated (63%)". His grade for the Old Farmer's Almanac winter forecast for San Francisco during 2006-2007 was a D+. He also evaluated the Old Farmer's Almanac for two separate summers and winters for all sixteen regions of the U.S., and found mostly poor results. For the summer of 2005, just one of the sixteen Old Farmer's Almanac regional forecasts got both the temperature and the precipitation correct. He plans to post a verification of their 2008 summer forecast sometime in the next week.

Weatherwise magazine study
In the October 1981 issue of Weatherwise magazine, pages 212-215, John E. Walsh and David Allen performed a check on the accuracy of 60 monthly forecasts of temperature and precipitation from the Old Farmer's Almanac at 32 stations in the U.S. They found that 50.7% of the monthly temperature forecasts and 51.9% of the precipitation forecasts verified with the correct sign. This compares with the 50% success rate expected by chance.

Old Farmer's Almanac climate forecast
It's also of interest to note that the Old Farmer's Almanac believes that sunspot cycles and other factors suggest that "a cold, not warm climate may be in our future". Their climate forecaster is Joeseph D'Aleo, who was the first Director of Meteorology at the Weather Channel. Mr. D'Aleo is now retired, and is often quoted for his skeptical opinions about climate change.

Conclusion
The results of my forecast verifications and those done by several others indicate that there is little reason to believe the Old Farmer's Almanac claim of 80% accuracy. These verifications attempted to be fair, but one can justifiably argue they were not objective nor complete. However, unless the Almanac posts some scientific evidence to the contrary, I won't believe their forecasts are any better than flipping a coin. One's best bet for the upcoming winter forecast is to use NOAA's prediction, which calls for an an above-average chance of a warm winter across the center portion of the U.S. If you live in Banner Elk, North Carolina, it might be wise to go with Kelly the Woolly worm's forecast of a cold winter, though, given the success of her predecessors!

Tropical disturbance near Costa Rica
An area of disturbed weather (96L) has developed in the extreme southern Caribbean, near the coast of Costa Rica. Wind shear is a hefty 20-30 knots over the disturbance, which will keep any development slow. The disturbance will bring heavy rains to Costa Rica and Nicaragua through Wednesday. If the center can stay off shore, this disturbance has the potential to develop into a tropical depression. NHC is giving a moderate (20-50% chance) that 96L will develop into a tropical depression by Wednesday morning. The GFDL model does develop 96L, but none of the other models do. I'll have an update on the system this afternoon if it gets more organized.

Jeff Masters