10 Day Forecasts: Partly Bullshit
To me, weather prognostication is like stacking pillows: The first two or three may find themselves accurately above the ones below, but once you’re ten pillows high the tower will come crumbling down in a heap of inaccuracy.
That is to say, tomorrow or the day after’s weather prediction may be accurate, but I’ve always operated under the assumption that 10-day weather forecasts are bullshit (and judging by conversations I’ve had with my friends this is not an unpopular opinion.)
The problem is, who is going to call out weather forecasters on their faulty 10-day predictions? Nobody remembers what tomorrow’s weather forecast was nine days ago. People only look at tomorrow’s forecast, because frankly anything beyond that is fairly inaccurate.
But just how inaccurate? How wrong is that 10th day prediction? And does it get more accurate the closer that day gets to today? That’s what I wanted to find out.
First I poked around the Internet looking for old 10-day weather forecasts. When I couldn’t find any, I started recording my own data by taking screencaps of weather.com’s 10-day weather forecast for my old Manhattan zipcode.
I assembled my first ten days worth of data into the info-graphic below:
Superficially this looks like a regular 10-day forecast, though take a closer look and you’ll notice one large exception: These are all predictions for the same day: Saturday, May 01. (GROUNDHOGS DAY!!! No, just joking.)
It’s organized in chronologically descending order with May 1st forecast at the top, and April 22nd prediction at the bottom — I should note now that 9 days away is the last day in a 10-day weather forecast, with the first day being todays prediction. You fucking idiot, did you really not know that?
It hardly takes a statistician to notice accuracy and time are inversely proportional. Or perhaps more simply put: The farther you are away from the day you are predicting, the less accurate your prediction becomes. Again, this is common sense, lets look at the farts. Just kidding, the facts.
Nine days away, on April 22nd, Weather.com had predicted a rainy day, with a high of 66 degrees. Well I hope people checked the weather for May 1st since that day, because they would have looked awfully dumb in a rain jacket and umbrella when May 1st approached and it was in fact 86 degrees and sunny! “At least take off that jacket you dumb fuck!”
Some other interesting things: you’ll notice a great leap in accuracy between 4 and 5 days away from “today’s date.” Meaning by April 27th, four days away from May 1st, weather.com had finally zeroed in on an accurate prediction, getting the high wrong by only 3 degrees Fahrenheit and the lows completely accurate!
Certainly one set of data isn’t very telling, which is why I kept recording forecasts until we landed on a rainy day. I was eager to find out if weather.com, which wasn’t very accurate about temperature, was perhaps a little better at predicting precipitation. I’m crying right now. It won’t stop.
Below is a similar info-graphic compiled from 10-day forecasts from May 3rd thru May 12th.
So for this set of data, these are all predictions for May 12. Based on temperature alone, weather.com was even less accurate then our first info-graphic. Nine days away, on May 3, they were not only off by 22 degrees (predicting a high of 73, vs the actual 51) but they were incorrect about rain (predicting mostly cloudy, but no precipitation). And they say banks need to be bailed out? …What?
You also may notice that 4 days away, May 8th, weather.com is still very off with their prediction, estimating highs and lows of 69/54 against the real 51/44 — a very chilly day for May! Yo quiero taco HELL!
However, on a purely precipitation level, by May 4th, weather.com pretty much nailed it, and never relented: it was going to rain on May 12th, and you know what? They were right!
Obviously this is just twenty dates worth of information (out of a possible two-million-eighty-thousand-five-hundred — Creationists do it from behind!) but it should hopefully help shed a little bit of light on just how much these so-called “professional” weather predictors get away with when it comes to 10-day weather forecasts.
;)
July 22nd, 2010 @ 6:31 am
you dont need a weatherman to know which way the wind blows
July 22nd, 2010 @ 9:03 am
too long did not read.
July 22nd, 2010 @ 9:55 am
hmm pretty interesting. I love how you rant about weather for like 6 paragraphs and then you put a smiley face at the end. cute!
July 22nd, 2010 @ 10:08 am
That is amazing that you did that. You’re officially my hero.
July 22nd, 2010 @ 11:08 am
it’s 12:08am right now and i could not process any of that. you’re so awesome though for doing this
July 22nd, 2010 @ 11:42 am
i read the first two lines before i realized how long it was, but i appreciated that suggestive smiley in the end;)
July 22nd, 2010 @ 11:46 am
Interesting. I wonder if location has any part to do with it. Are they more accurate for New Mexico or Nevada where there isn’t much rain?
July 22nd, 2010 @ 2:19 pm
Here’s a fun fact from the nerdiest people I know (i.e. physics grad students who also dabble in meteorology): they can confidently predict the weather up to 2 days ahead; anything after that, they may as well use a dartboard.
July 22nd, 2010 @ 3:57 pm
Very in depth Amir, pretty thorough journalism. One note though- wasn’t this a crazy May for weather in New York. I went to NY for work in mid NY and it was like 45 degrees. I remember everyone freaking out about how erratic it was.
More essays please. I enjoyed.
July 22nd, 2010 @ 10:00 pm
I feel like somebody saw this on your computer while you were in the bathroom and added a sentence to every paragraph.
i.e. “I’m cyring right now. It won’t stop”
July 22nd, 2010 @ 10:41 pm
lol!! you r so right amir
lov the smiley face at the end!!!!
July 23rd, 2010 @ 12:47 am
Wow, this is something I have always strongly believed in and voiced: that weather-prediction is a farce.
So I am a touch ashamed that me, a “scientist”, has never collected data like this before, yet one of my favorite “comedians” has.*
(* The quotes are to avoid being presumptuous about what Amir refers to himself as - not to imply a lack of humor.)
Meteorologists aim to operate at about a 40% accuracy. Ridiculous.
Would they take as much pride in their work if it was phrased the opposite way: that they aim for a 60% inaccuracy?
Now I know technically it is not a straight 50/50 call. There are nuances like just how hot/cold, how much precipitation, and how partially cloudy, etc.
But the common person pretty much is only interested in one thing, will it be sunny or rainy?
This implies I could consistently be more accurate with what the weather will be just by GUESSING.
July 23rd, 2010 @ 1:19 am
Genius and funny…tis a dream.
July 23rd, 2010 @ 1:29 am
I want to know what the circumstances were that compelled you to write this. Anyway, it was a fascinating viewpoint.
Here’s an extended limerick.
“Whether the weather be fine,
Whether the weather be not,
Whether the weather be cold,
Whether the weather be hot,
We’ll weather the weather,
Whatever the whether,
Whether we like it or not.”
July 23rd, 2010 @ 1:38 pm
I intern at a paper where I write a sentence or two about What Happened Today? 50 and 25 years ago. To do this, I comb through old newspapers. On the cover of the August 1, 1960 paper is a blurb predicting in a vague way the weather for the entire month. It’s easier to accept that 50 years ago The Weather Bureau said that the “30-day outlook for August calls for cooler and wetter weather than normal in the Great Lakes region… in the remainder of the nation, near normal conditions are expected,” because these are blanket statements. Still, fairly presumptuous of you, The Weather Bureau of 1960, to assume something we can’t with any certainty predict in The Future.
July 23rd, 2010 @ 5:14 pm
I found this off of Streeter’s and Ricky’s blogs.
It’s funny how they think it’s funny, and how you take the weather seriously.
Kudos for all the research you did! (: && Making it humorous and fun to read!
July 23rd, 2010 @ 10:19 pm
Accuweather is where it’s at.
July 26th, 2010 @ 1:22 pm
and the award for the 5th grade science fair goes to amir blumenfeld!!!!
but i enjoyed this
July 26th, 2010 @ 10:03 pm
I don’t fault the weathermen for not being able to predict the weather. Our entire atmosphere is constant flux. There is zero order to it at such a fine level. The very definition of forecast tells you it is not a definite. And besides, I am sure it is the common man, the viewers that are demanding longer and longer forecasts, causing weather bureaus to invest in doppler radar and blah blee blah, which, is no more accurate that anything else, because the landscape changes by the minute.
The more interesting experiment would be to use your own predictions for a 10 day forecast, and see how accurate the untrained and unequipped is at predicting the weather.
My personal peeve: a 40% chance of rain. What the fuck is that. Either it rains and you were right, or it doesn’t and you are a hero. There is no risk. I say, it should be 20% 80% or 100% chances only.
July 27th, 2010 @ 8:33 am
There was an episode of NOVA on chaos mathematics that dealt with this matter and others. The weather can be predicted within three days to 85% accuracy. After that the calculations involved are too complex for any supercomputer to handle.
July 28th, 2010 @ 3:11 am
43 lines of incomprehensible rantings about weather forecasts.
That’s Amir all right.
July 29th, 2010 @ 1:01 pm
I’ve discovered that the trick is actually really simple. If you bring an umbrella, it will not rain. If you wear a sweater, it will be hot out.
August 1st, 2010 @ 5:22 am
Haha I think Raynie’s nailed it! Great data analysis, Amir!
August 19th, 2010 @ 11:07 pm
this is brilliant. you’ve uncovered a conspiracy. the illuminati aren’t masons, they’re meteorologists!