Hacker News Reader

Comment by rikkus | Hacker News Reader

I half-jokingly tell people that if they get their forecast from the BBC (via the Met), they should expect the inverse weather.

I wonder if the BBC has checked the quality of the forecasts based on what actually happens? I use AccuWeather and the Aix Weather Widget (for Android) because they're nearly always right (where I live, at least) - and gives me much more useful information.

For example, I can't translate the Met's '50% chance of rain' - for a whole day - to real life. To me, that means it could not rain, or rain heavily all day, or anything in between.

View on HN

The accuracy?

http://www.metoffice.gov.uk/about-us/who/accuracy/forecasts

There are local variations which mean for example, that if you live in the welsh valleys, someone on one side can be having sun, while on the other side (and in view) there is rain.

The met office does forecasts for a broad area, if you want hyperlocal, wunderground is where I go. The met office could also run forecasts for each london borough, but I doubt you or anyone else would sit through a hyperlocal forecast on the news.

I actually learned that the rain percentage is not the percentage chance that it will rain but is in fact the percentage of the coverage area that they will expect to get rain. So 50% rain meaning no rain at all or rain for the whole day is entirely reasonable.

That's funny. I had long suspected that was the case but never knew for sure. It's also important to know where the weather station is for the forecast area too -- here they can cover relatively large areas with significant elevation change, which can greatly affect things like temperature and wind speed.

Reasonable unless you want your forecast to be at all useful!

How does that work when the percentage is supposed to cover a small area, like a town?

FWIW I've found that combining the iPhone weather app with a rain radar app gives me usefully accurate predictions of rain.

The Met Office is useful for coarser extreme weather warnings, but I don't pay much attention to its daily forecasts.

I've always interpreted it as the percentage of time it will rain, i.e. if it says 50% chance from 7 to 8, it will rain during about 30 minutes of that interval and not rain during the other 30 minutes.

(equivalently, you could say it's the % chance that it will be raining if you choose a given instant X in the interval).

I've never read an official interpretation but that seems to fit reality (where I live) quite well.

50% means that they ran the simulation say 10 times and in 5 of those simulations there was rain in the area. How to translate that to real life - you should probably carry an umbrella if you don't like getting wet. Also interesting to note that weather presenters usually bump up the probability of rain by 10-20%, because people complain about false negatives (rained when it was predicted not to) rather than false positives (didn't rain when it was predicted to).

If you're interested in a good intro to this topic I recommend the chapter on weather prediction from The Signal and the Noise by Nate Silver. Its a fascinating read.

Yes, it's the probability of rain.

Not quite:

http://www.srh.noaa.gov/ffc/?n=pop

Interesting that it only measures the chance of "measurable rainfall" and does not attempt to predict the duration of the rain.

I've found the opposite. Android/Yahoo/Google weather are generally telling me the wrong thing. I've come to rely BBC weather.

http://www.metoffice.gov.uk/news/in-depth/science-behind-pro...

First ghit. First duckduckgo hit.

And this matches what I thought it meant. It doesn't make it useful to me in real life, though.

Reply on news.ycombinator.com