I have been tracking it for a month or so, and the five furthest days out end up seeming to be no closer to reality than random chance would dictate. But I do live in a part of the country where the weather changes dramatically and very quickly due to the jetstream and storm fronts and so on. Does anyone else find it works better for them?
My impression has always been that ANY forecast more than about 5 days out is a crap-shoot.
Nate Silver discusses this exact thing in his book The Signal and the Noise. If I recall correctly, forecasts do better than simply pulling up historical averages for the first 7 days or so. After that, they provide no useful information.
One expects this to improve slowly, but maybe it’s just too chaotic.
It’s a marketing thing and nothing more. Unless there are some breakthroughs (and surely there will be someday), weather is way too chaotic to predict 15 days out. Those two week forecasts have been a joke in weather circles since day one.
I live in Southern California, where every day is sunny (and during the summer, hot). When I check the 15 day forecast, every day is labeled hot and sunny (OK, 2 days are actually labeled “partly cloudy” but that’s just their way of trying to make it seem less monotonous when the reality is that it’ll still be pretty sunny). I expect it to be amazingly accurate.
Sunday morning I was bummed out after seeing two weather apps calling for a 40%-60% chance of rain all day. Then I looked at a third app that said no rain. I chose to believe the “no rain” forecast, which turned out to be correct.
Hard to understand how an inaccurate forecast is good marketing.
Agreed. If I predict something that will happen 10 days hence, and have 10 opportunities to change my mind, my prediction today is worthless.
I did a small-sample study on the BBC 5 day weather forecast and I found:
[ul]
[li]The forecast at 5 days was seldom the right one[/li][li]The forecast for any specific day changed more often during the lead-up than the actual weather did from day to day during the period.[/li][/ul]
The choice isn’t between an inaccurate forecast and an accurate one - it’s between an inaccurate one and nothing. When all the other guys have X day forecasts, not having them loses viewers.
All of this is proof that weather.com and the Weather Channel jumped the shark a few years ago and should simply be ignored out of existence. They are not in any sense a useful source of actual weather information.
As opposed to?
Someone will correct me if I’m wrong, but I seem to recall hearing that all weather services get the same information from the national weather service, so you might as well go straight to the source - weather.gov.
n/m
I wouldn’t go that far. I have found their hourly forecast and the first several days of their 10 day forecast to be pretty accurate and useful.
And some of them weather girls are HOT!
I thought weather.com/Weather Channel had its own meteorologists who at the very least riff off the National Weather Service forecasts. Often there are minor and sometimes significant differences in their forecasts.
At “quiet” times of the year, I find forecasts are fairly reliable for about 72 hours out, then reliability diminishes rapidly. In winter through early to mid-spring, things are even more unpredictable. 10-day forecasts at such times are a crapshoot. 30-day or seasonal forecasts might just as well depend on ouija boards and magic 8-balls.
As noted in another thread, I think it would be instructive for the NWS and Weather Channel to keep records readily available to the public on how accurate their forecasts are in the longer term. (the NWS trumpets favorable info from time to time, but doesn’t make it easy to find.).
The ten day forecast works as a marketing tool because they offer it and no one else does, plus they’re going to be right sometimes. The rest of the time, they’re going to update it as time passes so that it will end up being right. What I mean is, they predict it’ll be sunny and warm in ten days on Sept 10th. By Sept 4th, a front is projected to be in the area that may make it cold and rainy Sept 10th depending on if a high pressure zone moves out of the way. The forecast is updated a lot between Sept 5-9 as the data firms up so by the 9th, the forecast is probably right, cold and rainy on the 10th.
In my opinion, the best way to get weather is from NWS forecasts and NWS Area Forecast Discussions. AFDs are released several times a day and are a look at the thinking that went into the forecast. They do take a little weather knowledge to understand, but it’s minimal.
Here’s one of my local AFDs from yesterday:
SHORT TERM…(TONIGHT AND TOMORROW)…UPPER LOW CONTINUES LIFTING NORTH NORTHEAST OVERNIGHT AS IT GETS SWEPT UP BY WEAK INCOMING UPPER TROUGH. BOTH OF THESE FEATURES ARE RATHER WEAK AS THEY MERGE TOGETHER AND SHOULDNT POSE A THREAT FOR ANY WIDESPREAD RAINS. WILL HAVE CHANCE AND SLIGHT CHANCE SHOWERS IN OVERNIGHT. FOG IS TOUGH TO FIGURE OUT TONIGHT DUE TO UNCERTAINTY OF PRECIP COVERAGE AND HOW MUCH CLEARING...IF ANY...CAN TAKE PLACE TOWARD EARLY MORNING. NOT ENOUGH CONFIDENCE TO PUT PATCHY FOG WORDING IN ZONES BUT A FEW ISOLATED AREAS OF FOG CANT BE RULED OUT FOR LOCATIONS THAT SEE RAIN. LOWS WILL BE NEAR NORMAL TONIGHT.
We had a 30% chance of rain and possibly storms but the AFD was telling us that the things that would initiate that weather was weak. The chance was still there, but small. We got sprinkles overnight, enough to wet the ground. Some people would say their 30% forecast was off, but after reading the AFD, I think they were dead on.
That’s probably mostly true. Most weather models are run by the NWS or use the same input as the NWS models (i actually run my own WRF model for fun, it uses NWS NAM data as input). Everyone else has access to them and can interpret them differently, but the NWS has a team of meteorologists in every area interpreting them. Most TV stations have one meteorologist and some “weather people,” good luck being more accurate than a team of professionals.
The Weather Channel does employ a lot of meteorologists so, in theory, they should do ok. But they still don’t have the resources to compile massive amounts of data to put into a model, not like the NWS anyway. They also have to cover the entire country. They’re going to spend a lot of time working a hurricane or other extreme event and not so much on whether there’s a 50% or 30% chance of rain in Boise tomorrow.
Time to watch L.A. Story again.