Comment by conductr | Hacker News Reader

Comment by conductr | Hacker News Reader

Parent Full thread

They have to predict the weather for the year in a book that has to go through the publishing and distribution process ahead of time.

My local weather news has all the benefits of real time data and weather models yet I think their accuracy rate is just as poor when it comes to producing the 7 day outlook. It’s common to hear a forecast for rain/cold front/etc in 7 day outlook that just never materializes. Also the timing of the event if it does arrive is almost always off by a day or two. Often they have the whole town worried about something that’s definitely happening Friday, they talk about it all week, everyone is preparing, little league games getting rescheduled, etc. then only hours beforehand it’s well looks like maybe Sunday. Then Sunday comes and instead of inches of rain, it’s a sprinkle.

I’m not even trying to be critical of weather reporting, I get that it’s a crapshoot but doing it a year+ ahead of time and getting similar results/accuracy is actually quite impressive.

I didn't study too much meteorology in undergrad, but one thing impressed upon us is that any forecast beyond maybe 3 days is basically guesswork.

I think what might be getting observed here is that when forecasting that many days out, the local data becomes so unimportant to the model's outcome that the model is just reflecting historical climate trends. Which kind of makes both the same kind of model. Ie. when forecasting tomorrow, the current temperature and pressure data really makes a difference. But once pushed to 7 days, those data essentially become a proxy for typical weather at that time of year, possibly down-weighted by a lot.

I just woke up and I feel like I'm doing a very poor job trying to describe this.

That depends on what you care about. Will it rain at a specific date/time - getting that for tomorrow is hard, much less a year. However you can often predict if this will be a wet or dry year with reasonable accuracy and that is important information (farmers plant different seeds). I doubt their model is very good at this, but science can do well enough.

I think you are describing it pretty well, and I've noticed the same thing. The farther into the future the forecast goes, the higher the probability is that it will look like the historical average.

One thing that I've found to help a lot is to go to weather.gov and look at the "forecast discussion". Often it will help to understand what types of uncertainties exist within the forecast.

It isn't unusual to see notes that make it really clear that 24-48 hour variations are expected, or that massive differences will exist based upon hard to predict variables. "Hey we think it will rain heavily as far south as X, but actually it might end up staying north of Y in which case X will stay dry"

It is easy to see how hard it can be even if the forecast itself turns out to be fairly accurate at a high level.

Even if we had perfect information at one instant in time, modeling a chaotic system going into the future becomes increasingly difficult.

We have far from perfect information and very flawed models too.

Interestingly, there seems to be some success with AI models that almost completely skip the science and jump straight to pattern recognition. It's interesting to think of modern 10-day weather forecasting going back to its old almanac roots.

> My local weather news has all the benefits of real time data and weather models yet I think their accuracy rate is just as poor when it comes to producing the 7 day outlook.

Where is your local source getting their forecast from?

> A seven-day forecast can accurately predict the weather about 80 percent of the time and a five-day forecast can accurately predict the weather approximately 90 percent of the time. However, a 10-day—or longer—forecast is only right about half the time.

* https://www.nesdis.noaa.gov/about/k-12-education/weather-for...

There are also 'technicalities': I'm in Toronto, Canada, which is 40km east-to-east and 20km from the lake to the northern border. If rain hits the western half (around 427/Sherway/Etobicoke) but not the eastern half (Scarborough bluffs), is a "it will rain" forecast correct for the city? Some will perceive it as yes and some as no.

They don’t have to predict the weather for a year. They choose to do something which cannot be done. Certainly not with any known technique, and probably not even in theory.

If I predict that the weather in my location on November 7, 2076 will be moderately cool and sunny (as it is today), I have a pretty good chance of being correct. I wouldn’t find it impressive, though.

Reply on news.ycombinator.com