11

The Farmer's Almanac is used by many farmers to time plantings, harvesting, fertilizing, etc., based upon the results of "long-term weather prediction". Is there any evidence that such predictions have any accuracy (more-so than simple educated guessing)?

**Edit: To clear up some confusion, here are some additional examples of long-term weather prediction: Old Farmer's Almanac and Met Office's dynamical climate model forecasts of tropical storms in the Atlantic Sector

  • I haven't read a farmer's almanac, but if the predictions are based on previous years, then that's probably look more at climatology, rather than weather forecasting. Note that there are services that perform actual long-term forecasting, up to 28 days, but as far as I can tell, they're pretty useless after more than a week (ie. the predictions are out by at least 10-20%, which means your dry weekend two weeks from now is not so certain). – naught101 Nov 5 '12 at 7:13
12

The way the majority of the weather forecasters, including the Meteorological Office in the UK do it is by running multiple computer simulations with slightly different parameters (as no model is perfect - or even can be perfect, according to chaos theory)

Where all the simulations agree, there is a high likelihood the forecast is correct, but this confidence reduces as there is more disagreement in results.

As you can imagine, the results are very similar at short times, however the further into the future you look, the more divergence you would expect from the differing parameters, so trying to get long term predictions is unreliable.

Admittedly, in some parts of the world it is much simpler than others - the UK, for example, is notoriously unreliable as it is a meeting point for various weather systems.

But in summary - if a large number of simulations agree, then you have a reasonable likelihood of an accurate prediction, so I think there is an element of truth in this one, but also a large uncertainty.

Edit:

| improve this answer | |
  • For New Zealand, I found the numerical forecasts from this source extraordinarily useful, with acceptable accuracy up to 7 days into the future. – Lagerbaer Jun 24 '11 at 19:00
  • Monte carlo formula – Hairy Jun 27 '11 at 11:05
  • 1
    Chaos theory has nothing to say about how perfect a model is. Besides, what does it even mean for a model to be perfect? – Raskolnikov Jun 27 '11 at 11:51
  • Chaos theory does tell us that for a model to be perfect it must be the thing it is modelling. Any simplification leaves out some info, and chaos theory tells us it is all important! – Rory Alsop Jun 27 '11 at 12:49
6

Rory's answer is good, but I though i'd make it clear that there are really two different sorts of 'long range forecast'.

To meteorologists, a long range forecast is about 7 days. Because the weather systems are chaotic (in fact, Lorenz who founded much of chaos theory, did so as a result of his work on weather forecasting) the longer out you go the bigger the uncertainty gets and meteorologists wouldn't claim even their 7 day forecasts are incredibly accurate.

Then there are forecasts like the Farmer's Almanac who make predictions more than a year our, using in their own words

a top-secret mathematical and astronomical formula that figures in sunspot activity, tidal action, the position of the planet in relation to the Sun, and a number of other factors.

That probably set off your skeptical-spidey-sense, and indeed wikipedia says cites an article (Walsh and Allen (1981) Weatherwise 212-215, not avaliable online as far as I can tell) that says they do no better than chance.

In New Zealand we have someone similar in Ken Ring, who claims to be able to predict the weather using astrology (the moon in particular). He's also been tested and found wanting [disclosure: this is from an article on my own blog]

His prediction was only on the right side of the average 17 times in 48 attempts, about 35% accuracy and significantly worse that you would expect to get from tossing a coin (if you’re one of those p-value fetishists p, in this case, is equal to about 0.03) .

I haven't read the Farmer's Almanac predictions, or how they claim 85% accuracy, but Ken Ring's post hoc justifications will be familiar to skeptical types. Here's a page on his website in which he rates his ability - my favourite among them is this one

Prediction: There's 2 rain days around 19th-20th; possible rain in excess of 100mm each day

Result: Happened further west - Broken Hill got about 20mm on the 22nd

A hit! Just further West, a fifth as much rain and two days later

| improve this answer | |

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .