The blog & portfolio of Matthew J. Rogers

Partly idiotic with a chance of correctness

June 8, 2007

Not a big chance of correctness, though. Every time I look at a weather report lately, I am reminded of Lewis Black’s words:

“You know what the word ‘meteorologist’ means in English? It means liar.”

For lack of a better place, I usually get my forecasts from Weather.com. Over the years though I’ve developed an innate sense of what I can partially trust and what I can’t trust at all. For example, I can usually trust anything they say will happen within the next 12 hours. Usually. The 10-day forecast, though, is a complete joke, because over the course of 10 days each day’s prediction will change almost 10 times.

Then you have gems like today. I’m sitting in my office in downtown Indianapolis, and wondering if it’s going to rain soon. I go to Weather.com, which reports that today’s high is 81 F (it’s currently 83 F…interesting), and there’s a “slight” chance of rain at only 30%. The radar, meanwhile, is this:

Is rain coming?

…which I’m fairly certain a monkey could interpret to mean there’s a greater chance than 30% for rain.

Can someone recommend a place to get my weather from that doesn’t suck? The only thing I like about Weather.com is their interactive radar map. Other than that, I’ll be happy to leave the totally inaccurate predictions behind…not to mention the rest of their web site which is so chock-full of animated flash advertisements that I can feel my computer grinding to a halt whenever it loads the page.

(P.S. This post is a true test to see if Kim ever reads this site anymore ;-) )

6 comments ›
Posted in Business & Legal, Web & Design

Similar posts

6 comments

  • Sean Iams June 10th, 2007 at 10:27 pm

    http://www.noaa.gov…it doesn’t have a pretty interface, but its more accurate than weather.com

  • Steve C. June 11th, 2007 at 6:22 pm

    I use http://www.wunderground.com when I need weather updates when I’m on the air. Rick Mecklenburg and his “Storm Tracker” updates are never up to date…

  • Kim June 28th, 2007 at 7:18 pm

    I love teh intarweb tubes!!!

    Mmmkay so… I am here to rescue you.

    Hang on a second… *BANGBANGBANG*… mmmkay, just had to get the extra varmits away from the Arc I’ve been building here in The Norman Rainforest Jungle.

    So first of all, use Your National Weather Service office (www.nws.noaa.gov and type in your area code). It has a great radar tool… you can zoom in if you loop the radar, and you can look at reflectivities to see how strong any rotation might be for yourself if something significant is actually occurring. Also, they have public weather statements at the top of the page. Don’t wait to hear it from the TV guy… read what the Weather Service (who you are paying for!) has to say about hazards/threats.

    All other companies really copy the NWS and try to tweak it a little anyway. TWC sucks because the local responsibilities are doled out to undergrad interns (who largely look at model output statistics (MOS) and average anyway).

    About the 30% schtick.

    1 – That is a forecast made ahead of time, very likely hours ahead of that mesoscale complex. Obviously, forecasting 30-45 minutes ahead of time is going to be much easier. That forecast is valid over generally an 8-12 hour period.

    2 – It’s really on *you* to be able to interpret the information that is being given to you. The percentage you see is the probability of rain falling within the forecast domain. So if it rains over 30% of the domain, the forecast is valid! (Forecast validation is another thing I could write an essay on, but you really don’t care so I’ll just say that’s “true”).

    Each NWS office sort of uses different guidelines for what percentages to use in certain situations. If you pay enough attention, you’ll notice if they tend to go high or low for your particular location (oftentimes, local topography or even urban heat island effects will confound models, and occasionally green forecasters). In all reality though, if they forecast 30% every day, and you get rained on 30% of the time, they win (the stats work out that way, too).

    That’s why, even though I live in The Norman Rainforest Jungle, where it has rained every day of the last 16 (we’re much rainier than Seattle, this year… rainier than Mt. Rainier!)… the daily forecast is generally 30-50%, because 30-50% of the domain gets rained on. Think of it as shooting at a target… I don’t know why, except that shooting is very very fun :D

  • Kim June 28th, 2007 at 7:23 pm

    PS – that’s why it’s called a “forecast”… looking at it like you are, that’s a “nowcast”. If you would like to do better, then feel free to make an attempt!

  • Kim June 28th, 2007 at 7:27 pm

    Gah. I swear this is my final note.

    Forecasting ability right now is pretty fantastic over the 1-3 day domain. After 5 days, forecasts for local areas are incredibly useless. You can get a general 1000 km-scale feel for the relative locations of things, but on that large of a scale, given compounding errors due to (a) sampling error, (b) underdetermined systems, (c) approximations, shifts in the forecasted large-scale pattern of several hundred miles are common.

  • Clint October 11th, 2007 at 9:31 am

    Over a 10 day period I can give you a rain chance forecast right now. Within the next 10 days I predict a 60% chance of precipitation for you. Uh, for almost anyone else too.

    A percent is the likely area total that will get rain. Not the chance of rain happening in the area at all. If that were true, I could predict a 100% chance of rain for North America today. But actually it would be more like a 15-20% of rain for the entire continent today.

    Heck, if it rains on 80% of the area and I predicted 30%, that would make me right for 80% of the people. Of course, I was right for 20% of the people too because I also predicted a 70% chance of no rain. Any way you look at it, using percentages, I am always right.

Leave a comment