Every ten minutes, I make an API call to the SMHI's API with the GPS coordinates less than 1 km from here (which cannot possibly matter for accuracy, but does for privacy). SMHI is the Swedish weather forecast "ministry" or however one would translate it. They've been collecting weather stats since the mid-1700s, so you'd think they would be pretty skilled by now. I think they have some of the oldest weather data in the world, but don't quote me on that.

I don't really care about the temperatures, but I do look out for rain. In fact, I automatically clear and update my little "calendar" in my "control panel" with little icons for when this API claims that it will rain in the next 7 days, so that I constantly know how long it's left until the next time it rains. In theory.

I love rain.

Well, yesterday, my calendar was full of those sweet rain icons. I rubbed my hands and started looking forward to a nice week full of wonderful rain, yet now when I wake up, it has updated to be six full days left until the next time it rains again. And this is far from the first time. In fact, it happens all the time. They basically seem to have no idea whatsoever when it's going to rain, until just before it starts.

Are modern weather forecasts this inaccurate? Are they just particularly incompetent over at SMHI? I did integrate another, international API. I guess I should write some mechanism to compare their data with SMHI's. However, I suspect that the international API might well just use SMHI's data for coordinates within Sweden, and I'm not exactly eager to figure out how to make such a comparison.

Is rain forecasting really this inaccurate still? Isn't one week a pretty reasonable window to expect accurate weather data?

This makes me really wonder how accurate those "farmer's calendar" predictions were. It would interest me to see how right they have been compared to modern "scientific" weather forecasting.

  • $\begingroup$ you can take a look here yr.no/?spr=eng it is not perfect but it is the best weather forecast we have for every place on earth. $\endgroup$ Aug 14, 2020 at 5:03
  • 1
    $\begingroup$ Rain forecast usually comes with a probability: en.wikipedia.org/wiki/Probability_of_precipitation If there is a 30% probability of rain, some weather agencies will still put those sweet rain icons (they tend to prefer predicting rain and there being no rain in the end than the opposite, which makes them look even worse). And even if there is a 90% probability of rain, you could still end up in the 10%. Add to the mix that this prediction is calculated over a large area, and could very well be realized somewhere in the box, but not 1 km away where you live... $\endgroup$ Aug 14, 2020 at 8:33
  • $\begingroup$ They do not predict rain for all over Sweden for the coming days. This can, of course, change without notice because weather like anything else in the earth system is non deterministic. Also, the tea pot effect may play a role: as long as you watch it won't boil (i mean rain) :-) $\endgroup$
    – user20217
    Aug 14, 2020 at 8:52
  • $\begingroup$ I think predicting the interaction of of oodlplexes of interacting thermal and pressure streams with days if accuracy is an amazing accomplishment, you are basically talking about simulating all the gas in our atmosphere,all the liquid in the ocean, and how they all interact in with the differential thermal heating of every material on the planet. It is literally one of the most chaotic systems in existence. it gets even more impressive when you realize how little data they actually collect, and that the system they are trying to predict is now changing in a way they have never seen before. $\endgroup$
    – John
    Aug 15, 2020 at 4:03

1 Answer 1


I am going to pause and say I don't know anything about SMHI. Therefore I can't comment on their competency. But I am sure they don't appreciate being called incompetent. If you are so passionate about rain, have you considered looking at weather prediction, and trying to beat them?

It is well known that the limit for predictability of weather is 2 weeks. Past that, climatology is just as accurate as modern forecasts in the grand scheme of things. Individual events may have more or less predictability. Precipitation is very hard to forecast, for various reasons including the lack of quality water vapor observations, the complexity of microphysics, and the interaction with other atmospheric physics.

It really depends on the mechanism for rainfall. Different mechanisms have different amounts of predictability. For example, a 5-day forecast of a hurricane is off by an average of 200 miles (321 km). Thunderstorms, especially individual cells, are hard to predict without probability. Extratropical cyclones are a bit easier predicted.

Farmers almanacs are notoriously inaccurate. Moreover, their "predictions" are fuzzy. You'll find better accuracy with actual forecasts.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.