www.accuweather.com has daily weather projections 45 days into the future. It was my impression that weather was chaotic enough that accurately pegging it a week to 10 days into the future was a big accomplishment.
Is this 45 day projection serious of just some kind of place keeper? Has meteorological science improved that much? How the heck can you know what the weather will be a month from now except in the most general terms?
Accuweather often blows it even one day in advance. Often.
Oddly, the Farmer’s Almanac claims to get it right 80% of the time, and they publish even further in advance. Maybe Accuweather is gigging off Farmer’s Almanac.
It’s hard to impossible to model weather systems more than a week in advance. It’s easy to predict the weather. I predict that four months from now, daily high temperatures in my area will be in the upper 80s to mid 90s Fahrenheit and that the chance of rain will be less than 5% each day. Do you really want to bet me that I’m wrong?
Isn’t it implicit that the further in the future a projection goes, the less “accurate” it’s supposed to be. My understanding is that in reality climatologists never “predict” anything. They aren’t trying to “peg it,” or say that “this is how the weather is going to be.” They’re just saying that conditions such as they are now have a such-and-such likelihood of leading to such-and-such outcomes, at such-and-such point in the future.
So what is the point (if any) with showing a 45 day daily projection?
You are correct that ten days is pushing the limit.
This site has some data that you might find useful - see post from March 6, 2009:
http://www.forecastadvisor.com/blog/
I don’t see any problem with giving projections even a year out. There is some value in knowing the historical values of what the weather is like in a certain area. Having it all on one site is fine. The just shouldn’t act like they are actually doing much predicting.
The site I posted is pretty good - I think I originally found it here. You can put in a zip code and it will show you which weather websites have been the most accurate for that area in the last month and the last year.
As pointed out in the blog - the predictions that were given for 10-14 days out were WORSE than using a “climatology” (one based on historical averages) prediction. I was kind of surprised to find the error for a “persistence” model (using today’s temperature to guess tomorrow’s) was only petter than climatology for one day out. Now keep in mind that any decent model would be able to take into account historical data as well as the days passed and such - so that they would heavily weight the models towards historical averages by day eight - and should put it all on historical averages (there may be some exceptions where a predictive model could be better) after that.
Therefore - the fact that they did WORSE than the climatology model leads me to believe that their internal training isn’t giving that much weight/penalty to predictions eight days out. I bet if they did - there could be some improvements, but it isn’t going to be THAT great.
Are you from Southern California? Because if you are, that’s cheating. 
Shit.
Because it looks to the layman like you know what you’re talking about.
I can predict the weather for any date in the future - even years in advance, as long as you don’t mind me keeping open the option to amend my prediction as often as I like between now and the morning of that day.
45 to 50 days is state of the art in the research field. Accuweather is just having fun publishing the results, but the latest weather (and oceanographic which is what I am familiar with) model ensembles make reasonably accurate predictions out that far. There is a lot they can’t describe-but for what they can do, they do quite well. Except of course when they don’t… 
They are research models after all and sometimes the tweek goes way wrong. Good thing for England to. One model I saw recently had arctic ice all the way down to the English coast. That one wasn’t a success…
The Farmer’s Almanac does claim this, though it’s been repeatedly disproven.
I claim the Farmer’s Almanac is about as good as a coin flip.
Nate Silver has a chapter in his book “The Signal and the Noise” about weather prediction. He shows that 5 day predictions are pretty good. The 10 day predictions are about as accurate as using historical averages. So anything past 7 days or so is getting pretty iffy.
Multiple choice:
Accuweather exists
(A) to inform the public
(B) to profit at the expense of a gullible public.
A friend an I actually took a screenshot of one of Accuweather’s extended forecasts (for late February and early March) about a month ago, and we’ve been keeping track of the forecasted highs vs. actual highs (reported daily by Accuweather). The forecast runs out in a few days, so I was planning to make a little spreadsheet and laugh about it then. I’ll post the data here if anyone’s interested!
10 days is a forecast. trends or outlooks for a month or two are made, temperature or precipitation compared to average.
If you move the predictions far enough out, they become climate.
Then you can be OK again w/ being confident abut accuracy.
So does that make astro a climate denier?
I did the same sort of thing for the BBC 5 day forecast - it was only a quick and dirty experiment lasting a week, but I found that across the span of the test
the forecast for that day varied greatly more than the actual weather did for the whole week. I’d post a link to my results, but I’ve just discovered all the images are broken on the page.
post it, please.