A forecast that says 30 % chance of rain is telling you something quite precise. Most people read it as something else entirely, and then conclude that forecasts are unreliable.
The figure is a probability of precipitation. It answers one narrow question: how likely is it that measurable rain, usually at least a tenth of a millimetre, falls at a given point in the area during the stated period?
The cleanest way to read it is as a long-run frequency. If a hundred days with exactly this weather situation came along, it would rain where you are standing on about thirty of them. On the other seventy it would stay dry.
Which means a 30 % forecast followed by a dry day is not a mistake. It is the expected outcome, and it should happen roughly seven times out of ten.
It is not the share of the area. A great many people read 30 % as "rain over 30 % of the region". That is a different quantity, and although the two are related in showery weather, they are not the same thing.
It is not the share of the time. Nor does it mean rain for 30 % of the day. A brief shower and an afternoon of drizzle can carry the same probability.
Some forecasters do combine confidence with areal coverage when computing the number, which is part of why definitions differ slightly between services. The practical reading stays the same.
Modern forecasting does not run one calculation. It runs the same model many times over, each with the starting conditions nudged very slightly, because we never know the current state of the atmosphere exactly. This is called ensemble forecasting.
If nearly every run produces rain over your town, the forecast comes out at 90 %. If a third of them do, you get 30 %. The percentage is literally a count of how many versions of tomorrow ended up wet.
That is also why the spread itself is informative. A tight ensemble means a stable, predictable situation. A wide one means the atmosphere is poised somewhere that small differences will amplify.
In the early 1960s the meteorologist Edward Lorenz restarted a simulation from rounded-off numbers and watched it diverge completely from the original run. The difference was in the fourth decimal place. The result became known as the butterfly effect, and it is the fundamental reason forecasts have a horizon.
Tiny errors in the starting state grow, roughly doubling every few days. However good the model, and however dense the observations, the growth is unavoidable. Better data pushes the limit outwards but never removes it.
In practice that gives temperature forecasts useful skill for around a week, and precipitation rather less, since rain depends on small-scale detail that grows unpredictable soonest. Beyond about ten days a forecast carries little more information than knowing the season.
Amount is a separate number from probability, and it is the one that decides how your day goes. One millimetre means one litre of water per square metre.
Snow is measured the same way, as melted water, which is why 1 mm of precipitation as snow looks like far more than 1 mm of rain. Depending on how dry the snow is, it can pile up around a centimetre deep.
The number is not a yes or no in disguise. It is an input to a decision that depends on what getting wet would cost you.
At 20 %, most people carry on as planned. At 50 %, take the jacket. But the threshold should move with the stakes: 20 % is worth planning around if you are about to leave a laptop on a balcony, and 60 % might be perfectly acceptable for a walk you can cut short.
The other habit worth having is checking again. A forecast for this afternoon is far better than the one issued three days ago, and updates are the cheapest improvement available.
Nollan.se shows the weather for whichever town you choose, right next to the clock, the name day and the days until payday.
No, that is the most common misreading. It is the probability that measurable rain falls at any given point in the forecast area during the period. If the same situation occurred a hundred times, it would rain at your location in about thirty of them.
No. The percentage says how likely rain is, not how much. A 90 % chance can mean persistent drizzle, and a 20 % chance can mean a violent thundershower if it does arrive. Look at the millimetre figure for amount.
One millimetre is one litre of water per square metre. Under 1 mm an hour is light rain you can walk in, around 5 mm an hour is a proper downpour, and anything above 10 mm an hour is heavy enough to overwhelm drains.
Temperature holds up reasonably for about a week. Precipitation is much harder, and beyond four or five days a rain forecast is more of a tendency than a plan. Past ten days or so, forecasts are barely better than simply knowing what the season is usually like.
Because they use different models, or different post-processing of the same model. Disagreement is genuine information: when several sources agree the situation is stable, and when they diverge the atmosphere is in a state where small differences grow quickly.