April 28, 2012

Iconic Confusion

An interesting aspect of the communication of weather forecasts isthe use of weather icons, those little picture with clouds, raindrops, snowflakes and the like.   You have seen them often:

Weather icons are an attempt to quickly and intuitively communicate the type of weather expected, but as we will see, there are some issues.

The National Weather Service is a big user of icons and you will see them strewn across the top of the pages when you get a forecast at a location:

I asked my colleagues in the NWS where these icons came and I was told one of the offices developed them, without the aid of any social scientists or objective evaluations of efficacy.  Now, I am not saying I could do better, but let's say the distinctions between some of them have me scratching my head.    Now what IS the difference between mostly and partly sunny?   Even in graduate school they don't teach this.

Now the difference between showers likely and chance of showers is perhaps a bit subtle
 As are the rain icons, where 20% to 100% get the same treatment.
And the icon for a slight chance of freezing drizzle is enough to send icicle-size shivers down any spine! 
I can imagine what a good chance of heavy freezing rain would look like:

Now the private sector does not have bragging rights about better weather icons.  Here are the state-of-the-art icons used by the Weather Channel:

Not much of a distinction there!  And my friends on KING-5 weather are not exactly consistent in the use of  weather icons, with showers getting raindrops some times, but not during others.

Some media outlets try to communicate the probability of rain by how many raindrops are falling out of the icon clouds...but people confused that with intensity.

So is anyone examining in a scientific way the best approach to developing and testing weather icons?   I am proud to say that such work has been taking place at the University of Washington, under the leadership of psychology professor Susan Joslyn.  Many of you know about our probcast web page, where we present probabilistic weather predictions in an accessible way.   We wanted to have icons for precipitation and Dr. Joslyn created and tested a large number of  possibilities of what she termed Precipicons (a few shown below)

The winner was a pie chart type of presentation:

The probability of precipitation is shown by the raindrop portion, but the probability of not raining is shown explicitly by the solid color.    More work is necessary in such icon research.   The National Weather Service is also very much aware they need to entrain social scientists into their communication efforts and are putting resources in this direction.

Judge Judy and My Lost Dog

My family are still looking for our lost dog (and we continue to get some reports)--we so much appreciate your expressions of support and suggestions.  Recently, it all took an unexpected and somewhat humorous turn.   We got a letter from the producer of the Judge Judy cable show.  They had heard about our filing for damages in small claims court and wanted to fly us down to LA to have Judge Judy adjudicate the issue between us the the woman who lost our dog (Dede Harris, http://www.pet-nanny.biz/, located in North Seattle). Well, I am not going to miss class to fly down to LA for such a show, and do you know how much they would pay for the effort?  $ 250.00  You will not get rich going on Judge Judy.  But you might get on her DVD.

Yelp has again taken off the complaint I put on its web site about her.   You really can't trust Yelp--it appears that negative information is shifted into their "filtered" list---is this due to complaints from advertisers?   Lots of stories about that.  Example.


  1. Displaying quantitative data graphically is a fascinating subject -- thanks for the peek behind your department's investigations into this area.

    I suspect you're already familiar with Edward Tufte's books on this subject; I recommend them (or attending one of his 1-day seminars, which he brings to Seattle every summer) to your readers. Great stuff!


  2. Cliff, never mind the icons. I don't even understand the distinction in some of the forecasts! The one that gets me is "rain changing to showers". What's the difference? Constant versus intermittent? Who knows? Then there's "partly sunny" vs. "partly cloudy".

  3. Speaking of Probcast, what is wrong with it? For the past few days it has been down with just question marks for everything. Usually there is a message at the top saying that repairs are occurring but there aren't now.

  4. Actually, the 10% or 50% chance of rain can mean any of three things. It can mean 50% of the area will get rain tomorrow - for instance if there will be showers from unstable air. Or there is a 50% chance the entire area will get rain tomorrow - if a warm front is moving in and the forecasters are not sure when it will start raining. Or it can mean it will rain for half the day tomorrow everywhere - if the forecaster knows it will rain all morning and clear for the afternoon. So 50% chance of rain tells us only a little. And not enough to plan outdoor activities. The forecasters know - but they seem too afraid of overloading us with details. Wish they would.

  5. Still, with all the ambiguity associated with these icons, even a nice pie chart with 10% chance of rain doesn't help much except to say that if you were rained on that day there would have been just a 10% chance it would happen to you, but then a 100% at the moment that rain drop hits your slicker. I guess it's still better than putting your hand out the window and forecasting from that information.

    Really enjoy your presentations on the website. Please keep up the great work and do keep stressing the importance of basic science learning in k-12.

  6. As a college graduate in graphic design and someone who has seen a lot of icon design, and recently begun learning about user experience, the dilemma you illustrate about weather iconography is a classic UX problem. I'm delighted that people are taking this on, but the designer in me is pained to see the graphics look so...well, so circa-1997. The result you've shown is a great wireframe (UX term), but they shouldn't think to stop before putting some window dressing on it. Or some spackle, for that matter. We have the ability to make things both beautiful and functional. I hope that if the UW's research is adopted by media and news outlets, they will employ some graphic designers to dress up the iconography. I'll be the first to offer my services.

  7. What about the icons posted at thetop of the blog - they seem to work quite well on the surface, anyway.

  8. I'd like to see a distinction, say of a 40% chance of rain, of 0.01 versus 0.5 inches or all day versus a few minutes of rain. Makes a world of differnce when I plan an activity outside.

  9. I was under the impression that 50% chance of rain means that under these (approximate) atmospheric conditions, historically it has rained half the time.

    Or half the numerical runs, with slightly different starting conditions or random factors (pr even different software packages), called rain and half the runs didn't.

  10. I would have loved to see Judge Judy adjudicate the author's grievances with KUOW.

  11. Probcast actually specifies in the help text (just hover over the forecast value) exactly what the probability of precipitation means - in this case "Chance of any amount of precipitation occurring [at this point in space] during the corresponding time period"

    Hope this helps!

  12. Good choice in not taking Judge Judy up on her offer. As I see it, you'd have about a 50% chance of a sunny outcome and a 100% waste of time.

  13. Its highly unlikely that this lady is giving yelp the advertising dollars to have them filter her reviews. What is more likely happening is that yelp is "de-prioritizing" your reviews because you are not a frequent yelper. Yelp displays and gives the most weight to reviews that are left by people who use their site a lot under the theory that they are going to give the most reliable opinions. Quite honestly, what they're trying to prevent is someone who opens a business just telling all their friends to sign up for a yelp account once, and then leaving glowing reviews.

    To have your review get more impact on her page, try leaving a few more yelp reviews for other businesses.

  14. If you develop an icon for every possible type of weather, you would have so many that they would be even harder to read. You need numbers and text in addition to a few simple icons.

  15. Edward Tufte is the finest resource for Visual Display of Quantitative Information I've discovered.

    Did you try a cloud with ',',',',', ten drops?

    Twenty? ',',',',',

    Too small & busy. just a thought.

  16. OMG, I am so glad someone with credentials finally pointed out the issue with the NOAA symbols!!!

  17. And, what's with the 100% chance of rain? I assume that means >95.0% chance of rain. In any case, I have a hard time believing that one when it is given more than an hour in advance. Even if there is a 96% chance of rain, the timing is probably off if it is more than a few hours away.

  18. Cliff, I'm glad someone is looking at this topic. Those NWS are icons horrible. I have told my wife not to look at that site any more, it just makes our weather look so bad. I already have to defend the climate to her enough as it is.

    I remember in the late 90s one of the local newspaper web sites, I think it was the Times, had a graphic which I really liked. It was a vector image of a landscape, Mt. Rainier with the city, I believe. Superimposed on this were 5 columns. The weather was actually shown across the five columns in one continuous graphic.

    For example, on the left of one column there might be some low clouds shown, in the middle of the same column there might be no clouds and then clouds again on the right side of the column. This very typical summer pattern is captured in time by the column. The other nice thing about it was that high clouds could be shown at the top of the column above the low stratus.

    Although not 'iconic', this treatment allowed for variation in both time and space which I found very useful. It was easy to see at a glance what the weather for the week would be. It side stepped the whole 'partly sunny/cloudy' issue by showing the weather as a dynamic entity.

    It was obviously put together by a graphic designer and not an academic. Having had some experience with universities, I would highly encourage you to make use of folks like Ms. Pfeifer when looking for some new ideas. Tufte is great and all, but he is not the last word on such matters.

  19. While icons could be improved, there's just more information to communicate than can fit on a postage stamp. Since weather information is by nature time oriented, meteograms make a lot of sense but don't seem very common. For a beautiful implementation see Michal Frankowski's app: http://meteogram.se/ I'd live to see what he could do with the probcast data feed.

  20. I just read an interesting take on the Yelp review problem. A gentleman got ripped off for 2 thousand dollars by a moving company. The company had many bad reviews, but they were all hidden. His critique of their algorithms was very intersting. http://justinvincent.com/page/1874/yelp-you-cost-me-2000-by-suppressing-genuine-reviews-heres-how-you-fix-it

  21. You are too kind to say you couldn't do any better. I do love the heavy freezing rain icon you created, though!

    I had a weather site back in the late 90s and I think the weather icons my digital designer sister made are still more detailed (and more graphically pleasing) than what is out there today.

  22. Yelp has gotten pretty smarmy in the last few years. I sent a letter of complaint to them regarding the pulling of the negative reviews for the pet nanny.
    Probably on deaf ears.


Please make sure your comments are civil. Name calling and personal attacks are not appropriate.

More Rain for the Northwest is Good News for Wildfires

After a very pleasant dry spell, another rainy period is ahead for the western side of the region and the Cascades on Friday and Saturday.  ...