Monday, February 18, 2013

Why are weather forecasters "always" wrong?

OK, more science, and this time I think I've outdone my blog nemesis (or "blogesis" if you will), Background Dominated.  I'd like to see his "rational" thinking outsmart me here!

So, one particular family member of mine, who will remain unnamed, always complains "Weather forecasters are the only people to be wrong half the time and still get paid!"  Well, little does he know that weather prediction is an incredibly challenging field, and the fact that weather predictions are sometimes way off has nothing to do with the forecaster being wrong, but rather with the so-called butterfly effect.

The basic idea behind the butterfly effect is that if you have an incredibly complex system (like... well, the Earth) and you add a small perturbation (like the flap of a butterfly's wings), eventually this perturbation will drastically affect the system, causing it to be very different than what it would have been without the flap of the butterfly's wings.  Put more simply, the atmosphere on the Earth is so complex with all of its motions and solar heating that a little poof of air can evolve into something much bigger!  If you really want to know more about this, look up chaos theory.  I won't get into the math here...

So, what does this have to do with weather prediction?  Well, atmospheric scientists run large scale simulations (actually quite similar to what I do) using huge computers to solve a set of equations that describe the weather. They then get a prediction for how a particular storm will develop, as an example. Except the problem is that even with the most advanced simulations out there, no one can account for every single variable.  The Earth is just too darn complicated!  We don't account for all the butterflies!

So, sometimes these models predict something different than what actually happens.  BUT, what is amazing to me is that weather predictions are more often right than they are wrong.  Given the power of the butterfly effect, I think the above statement made by my scientifically-inept family member should be revised to say "Weather forecasters should be paid more because they have to deal with an incredibly complicated system, and most of the time they actually get it right".

Now with that, I want to gloat a little bit about the power of computers.  In his last post, Background Dominated made fun of these precious "adding machines"  Well, these adding machines do a great job at making predictions, even in the most complicated systems.  With the proper software (which scientists develop), these machines are more often correct than incorrect.

Background Dominated will now say that his X-ray telescopes are never incorrect because the collected data comes from real systems, and not simulated systems.  But as he himself points out in his blog, he has to deal with all that noise in the background, which means he has to process his data to get anything meaningful out of it.  Have you seen his annoying spotty background on his own blog page??!

Anyway, I leave you with this because THIS is my tool for research.  This "adding machine" is a 100,000 CPU supercomputer (imagine 100,000 normal computers linked up to act as one!):

Kraken Supercomputer
Courtesy: XSEDE, NICS

21 comments:

  1. From Baltimore, today I watched the weather reporters create mass hysteria predicting several inches of snow for the whole state. Warning us not to go out saying it was already bad etc. Most of what they were talking about was predicted to happen within a few hours but we saw nothing. I can understand a prediction days in advance turning another way but they were saying it like it was happening now and getting worse when nothing was happening and it never got worse. There is really no 8 paragraph article that can explain this kind of inaccuracy and we see this on a constant basis. From now on I will just look out the window everyday and wing it.

    ReplyDelete
    Replies
    1. I guess the real question is how can you be a meteorologist and get paid and be wrong most of the time, I know that would not work at any other job you would lose your job, all I see here is excuses why they are inaccurate, but if a mechanic was working on a car and somehow managed to crack the block he certainly would not be able to blame the machinery or have any excuse. If the weather is based off the predictions of a super fapbox then what is the point of meteorologist anyway? All I'm saying is couldn't the news people just read what the fapbox says and skip paying the middle man? Or will taxpayers continue to pay people to always be wrong? Seems like the only trolls here are the ones defending the fact that they can't give an accurate forecast. But I guess if I was getting paid to be a lazy sob and always be wrong I would look for excuses and blame machines etc. meteorologists should just come on TV and say "there's going to be weather today" I could half swallow that knowing my taxes are paying these halfwits but the fact they say what's going to happen and are wrong most of the time is unacceptable. So when I miss a day of work because a halfwit on TV swears 100% rain and it never rains why the @#$# should he get paid when his lies cost me a day of work and my taxes pay his salary? Save your excuses turdulent scientist and start paying for days people miss based on your bs predictions after all we are the ones making sure you get paid and this is how we are related? With innacurate forecasts and excuses? Most ridiculous job on the planet!

      Delete
  2. Thanks for your comment! I won't give you an 8 paragraph explaining this kind of inaccuracy. You are right in that sometimes, the weather predictors are just wrong. If you're curious, I can talk with you more and explain why. But I think if you keep track of how often predictors are right vs. wrong (maybe just jot down the forecast every night using a pen and paper), you'll find that they are right much more than they are wrong. It is human nature to *remember* the times when things fail and *forget* when things go well.

    There's a lot more I can say, and again we can talk more if you want (through this thread). Also, I sincerely hope you are not just trolling me. I don't think you are, but if you are, don't comment on my posts again. Thanks!

    ReplyDelete
  3. That computer on steroids is just another way of wasting tax payers money. Dont they ever look outside anymore?

    ReplyDelete
  4. I am not sure if you are the same Anonymous who posted previously, but whether you are or not, this sounds very much like a troll post.

    If you want a legitimate debate, I will be more than happy to engage you, and please let me know if this is the case. But your language does suggest otherwise.

    Let me be perfectly clear here: I DO NOT debate with trollers. Trollers are a waste of everyone's time and are obviously not mature enough to engage in an intelligent conversation where maybe both sides could learn something.

    So, other than maybe a future dedicated blog post re-emphasizing this point, this is the last you will hear from me on this.

    ReplyDelete
  5. " ...if you keep track of how often predictors are right vs. wrong..."

    That's precisely the point. There's no place online where we can compare forecast and actual weather over time and geography. Anecdotally I can say that in my area weekend predictions of rain and cloud cover appear to be wrong more than 50% of the time.

    Give the amount of weather than these forecasters collect, you would think that they would want to publicize this data to prove how good they are - if the data actually showed them as being correct more than half the time.

    As it stands, aside from major events like hurricanes, for which forecasts seem pretty accurate at least within a 48 hour window, all other forecasts for things like cloudiness and rain appear to be about as scientific and accurate as astrology.

    ReplyDelete
  6. Hello,

    I'm Mrs. turbulent scientist, who is a trained weather forecaster.

    In the U.S. at least, the various government-funded weather forecasting centers do keep track of their accuracy record. I am not sure where or even if this is available online, but being a professional atmospheric scientist I have seen these records. There is one unfortunate problem however, in that the people who are making the forecasts are also grading themselves. I'm not saying they are lying, but they make up metrics that make it so their predictions look the most accurate.

    For what it's worth, weather forecasts are solid within 48 hours; at 3-5 days they are ok. Past 5 days you will be equal or worse than just predicting the climate "normals" for that day. So keep that in mind when you look at internet or TV weather forecasts...producers (i.e., non scientists) want to brag about how far out their website or station can predict. The scientists have to go along with it, but their forecasts are utter b.s. and they know it. They usually just put partly sunny and the climatological average for that day.

    I will say one other thing about weather forecasting. The complicated computer programs (we call them models) that take the equations of physics and current weather conditions to produce a forecast, are generally wrong for a specific location. This is in part due to the fact that these models have spatial resolutions of a few miles, and the fact that we can practically input only land stations (and only some land stations, usually major airports, launch weather balloons to observe the atmosphere above the ground). This is where a statistical tool called the "Model Output Statistics" (MOS) comes in. The MOS corrects for biases in the model output. For example, say there is a lake next to the west of the weather station, that is too small to be picked up by the model, and so whenever the wind blows from the west, the observed temperature is colder than the predicted temperature. The MOS will pick up on this bias and know to predict a lower temperature whenever the wind is out of the west. Lastly, the longer statistical base the MOS has, the better it will be. Real life example: there was a model called the Nested Grid Model (NGM) that was from the 1980's (1970's?). The actual model output was totally useless, but by the early 2000's it had over 20 years of baseline for its MOS, and could compete with more modern models. Unfortunately, they discontinued the NGM in the late 2000's.

    ReplyDelete
  7. So if what you all say is right, then the things we feel are wrong, personally I know when I am cold and hot, I have found that the weather people are right, give or take ten degrees, or two to three inches of rain, or a hurricane or five, I think my grandmother was better at it than anyone, she used the toad and bass techno model, never failed.

    ReplyDelete
  8. It's absurd to try and justify weather forecaster's accuracy by trying to average in all days. Any idiot without any training or sophisticated equipment could forecast every day will be 75 and sunny and be MUCH more accurate than what we have now, which is driven by sensationalism and not the weather. Sensationalism sells commercials and puts a sense of dread in peoples minds so they think they have to keep watching.
    All your models and projections are HEAVILY ADJUSTED to produce sensationalism.
    Trying to defend this is ridiculous.

    ReplyDelete
    Replies
    1. What you say about model accuracy is true beyond about 5 days. I said that in my post. You are right to say "climatology" is more accurate than models weeks in advance. Tell your senator/congesspeson to devote more money to science.

      If you are referring to my comment about the NGM, it has some limitations. I worked quite well when it was 75 and sunny. It was awful at doing rain or snow or anything uncommon, because it didn't have enough info (i.e. wasn't trained) to handle such events.

      Delete
  9. To Mrs. turbulent scientist, who is a trained weather forecaster.

    You say you have seen the data that compares the forecasts to what actually happened. Where can I find this data? I am doing a school project and this would be very helpful.
    Thanks!

    ReplyDelete
    Replies
    1. Finding weather observations online is quite easy: try classic.wunderground.com, type on your city, and somewhere there should be a link to that day's (or a drop down menu) weather conditions. You can select the year you want and output it as a table. (warning: last time I checked the records didn't go back as far as NWS records, which you can find on the pages of the individual offices usually in an inconvenient format.)

      Finding what the model predicted unfortunately requires a great deal of effort. I have a former grad school classmate who kept track of it "by hand" (not really, he used an automated computer program that he wrote). His weather forecasting page is here:
      http://www.atmos.albany.edu/facstaff/tang/forecast/
      Unfortunately, you are not allowed to dial in the station that you want, and right now it's set to KALB (Albany, NY). You can look at the bar graphs for climatology, but the verification, which is what I was referring to in my post, is restricted (and I can't give out the password on this forum). I'm sorry I misled you into thinking there was a giant dataset somewhere available to the public, I was speaking from years of weather forecasting using his page.

      Delete
  10. Well you really only confirmed what I was thinking originally, that is there is very little forecast verification available to the public. There is a site called http://www.forecastadvisor.com/
    which tracks a few of the major forecasters and reports their accuracy up to three days. Which ranges between 58% to 80% accuracy, the weather network being the best.

    I think the public would be interested to hear how accurate their local forecasters are in general and compared to their peers. I don't thinks its right that the weather man on the news comes on telling us about the 5 to 7 day forecast and that we should be planning our schedules around it. When in fact the accuracy of those forecasts is no better than 50%. People would be better off not listening at all, or simply guessing. If anything the weather forecasts greater than 3 days should have disclaimers describing the accuracy and that they shouldn't be taken seriously.
    Just my two cents anyhow

    ReplyDelete
  11. All I'm saying is that when I read a prediction of 50% chance of "Thunderstorms" in less than an hour I'm at least expecting more than 5 minutes of light sprinkling.

    ReplyDelete
  12. In the morning, open the door, step outside, look around, and you can forecast the weather more accurately than the "experts" on TV. Rain turning to snow has been predicted for today, but what do we have--a huge NOTHING. And this happens not occasionally, but most of the time.

    ReplyDelete
  13. more silly accuses. they like the attention so they create the hype. they are WRONG
    more often than not. they never apologize only mince words sometimes
    actually changing what was original stated or simply changing the subject. seems like they have the greatest occupation ever. create the chaos while leaving those great politicians trying to do the right thing with egg on their faces only because of info given to them. there are no repercussions and they still get a good paycheck. they talk endlessly on their daily forecasts yet you rarely get exact info of what's to come. it's gotten worse as they are now costing us money and worrying people needlessly. they along with their channels should reimburse us every time this occurs. they should focus on getting their act together, period. WHO NEEDS THEM!!

    ReplyDelete
  14. Ha... I'm in Dallas and our weather fools... er, forecasters have been bleating about severe storms for Wednesday, Thursday and today (Friday). They've been on this hype for over a week. But what actually happened here in Dallas-Ft. Worth? There was a smallish tornado "in" North Texas and some large hail to the west, but nothing truly severe for the Metroplex on either Wednesday or Thursday. It's still early as I post this, but right now it's raining a little. All the heavier rain is to our north, south, east and west while a 'rain free slot' or donut hole is going right over DFW. Anything that approaches us is diverted around. But it's enough to queer our severe chances, except for MAYBE some elevated stuff above the 're-stabilized' lower atmosphere in our area. I don't hold out much hope on even that happening. This would make the third bust day in a row. 'Three strikes', ya know?

    Now, imagine being a communications volunteer. You are one of the folks who gets the late-night phone call or page for a severe weather activation. You get up, make your coffee, and man your radio. Or you are a field spotter and you drive, groggy perhaps, to your spotting location. And you wait, and wait... for storms that peter out or change course just as they get to your county. Nothing happens. And this scenario happens maybe half a dozen times during the PEAK season of April through June-- or even more some years.

    Would YOU be so willing to lend your services in the future if you experienced bust, after bust, after bust? I think not. I only stick with it to support my ham radio friends. If not for that, I'd have bailed as a Skywarn Spotter long ago-- at least in this part of the world where it seems that our forecasters couldn't forecast a sunrise if they saw it happening.


    If I f**ked up that badly for 3 straight days on my job, I'd be fired.

    ReplyDelete
  15. When it rains I get the day off work so it seems like they are wrong all the time to me!

    ReplyDelete
  16. There is more people that are liked minded. I have at a fews time in my life recorded the predicted weather (not weather at the moment, or what anyone can tell may happen, within the next few hours). Anyone can also project the weather a little more into the future with global pictures/radar. .... Anyway... they are wrong So, So much!
    Pay them only when they are right for their small prediction area. If we take a very large area and time, they may always be right. Come on science, you are better that That! jess

    ReplyDelete
  17. I've worked in the weather business, and based on what I've learned about how truly accurate forecasters are, which is exactly what Mrs. Turbulent Scientist said. So, with that in mind, here's what you people who think weather forecasts are always wrong can do to prove that your perception is right:

    Next time a hurricane is predicted to hit a particular beach, go down to that beach before it hits and tie yourself to a pier post. Because, if your ideas are correct, then you'll be just fine and won't drown in storm surge. Right?

    And all I can say based on my knowledge of improving accuracy of hurricane tracks, my guess is by doing this, after the storm hits and your under water for hours, the world will be a better place.

    ReplyDelete
  18. So the only weather forecast that will be accurate is a hurricane because it's 200 miles wide and almost impossible to not be accurate? Pffft,well then, .save next time if you're trying to defend being wrong do yourself a favor and blame the super computer, that was a ridiculous analogy.

    ReplyDelete