If I look up an on-line weather forecast (in the UK I tend to use the government Met Office site, but others are similar) I am presented with today’s forecast in 1 hour slots. One of the figures is the probability for rain as a percentage. Obviously as a numerate person I have a pretty good feeling for what that number is telling me, but numbers expressed as a percentage should have a precise meaning and I don’t know what that is. An x% chance could mean that, within that hour, there is an x% chance that at least one raindrop will fall. Alternatively it could mean that should I poke my head out the door any time within that hour there is an x% chance it will be raining right then. Another thought is that within the forecast area, on average, it will be raining over x% of the area at any time during that hour, or that x% of the area will experience some rain within that period. Alternatively it could all be BS and they somewhat arbitrarily assign probabilities to much vaguer notions of likelihood such that my “pretty good feeling” is about as good as it gets. Fight my ignorance.
My understanding is that what it means is that when you look at the historical record of instances in which the parameters matched the current readings, it was raining on X percent of those days.
Sent from my iPhone using Tapatalk
Cecil answered this question some time back, you can find the link here
My understanding of it lines up with both Cecil and Acsenray.
Here’s Tom Skilling from WGN on the question:
Important to note:
“So, when there is a 70% chance of rainfall as a certain location, there is also a 30% chance rainfall will not occur.”
Probability of Precipitation from NOAA’s Jacksonville NWS office.
That makes no sense. If there’s a large, heavy rainstorm coming, for instance, the chance of rain is 100% regardless of how often it has rained on that day in the past. The weather prediction is based on more than past records, it’s based on current conditions.
Of course, it`s based on current conditions.
If the current conditions are sunny and calm, it`s not going to match previous conditions that were cloudy and windy.
OKay, I see now that I should have read Ascenray’s comment more carefully. I’m still not sure he’s right, but I see he wasn’t just talking about past conditions.
Note that it gets a lot more complicated than what’s been answered so far, especially once a forecast is made by a less analytic entity.
E.g., the forecast for my area tonight is a 50% chance of rain. But for one four hour block alone, weather.com* predicts at least a 40% chance of rain for each of those four hours.
Note that just for the first two hours alone, if there really were a 40% (2/5) chance of rain in each hour, then the odds of rain during those two hours are 2/5 + 2/5 - 4/25 or .64 – a 64% chance of rain, which is higher than the 50% predicted for the entire night. A sequence of 4 hours with each a prob. of 40% chance of rain would then have an 87% chance of rain occurring during that block.
So if you see an hourly forecast for a whole day where each hour has a 10% chance of rain, it doesn’t mean that there is a nearly 100% chance of rain sometime during the day, even though that is the way probability is supposed to work.
(It also goes the other way when the overall odds of rain for a 3 or 5 day forecast don’t align with the odds given for for each individual day.)
You don’t really know when seeing a percentage whether it is based on a precise definition or something ginned up to express a rough idea of rainy-ness.
There’s probability as defined by Mathematicians and probability as defined by certain other less precise folk.
- Until recently owned by the Weather Channel but this forecast style goes way back for them.
I’m not sure The Weather Channel does anything more than read NWS copy. Perhaps they take the time to run the numerical models on the hour for a few hours, and then rely on the NWS model runs for anything longer.
They take the current weather conditions and punch them into a computer. The computer searches through it’s history and finds all the times the current conditions were the same in the past. If the computer returns that it rained 60% on the time in the past, then they predict 60% chance of rain in the future. NWS gives this data in 6 hours blocks and run the computers for next 72 hours.
I read an explanation of this once that I thought made sense, I think it might have been in Nate Silver’s book the Signal and the Noise, but I might be wrong.
It was to do with chaos theory. They take all the current information from weather stations and feed the data into the computer models. The numbers are flexed to allow for slight uncertainty in measurements. e.g. a pressure of (say) 105.567 could be between 105.5665 and 105.5674
Thousands of numbers are flexed with dozens of slightly uncertain values, and the models run as many million times as possible. In a certain percentage of the results of these it will be raining at X O’clock the next day, and this is where the probability of rain comes from.
Or something like that.
I live near Seattle and have learned to interpret it differently. There is a 100% chance of rain, let’s just take that as given around here. If it says 50% in the forecast, that’s the probability that rain falls on me during the fifteen minutes that I walk to work.
I think Seattle has a base chance of rain of 40% during wet season … on any given day, the probability of rain is 40% PLUS what ever the weather conditions are … 1040 mb high pressure ridge and it’ll rain … light, steady, unrelenting, 24/7, liquid water falling from the sky in all it’s wide variety of phases.
Nm
I’ve heard a different explanation. According to the weather forecaster on one of my local TV stations, when they say “30% chance of rain”, what they mean is that they expect 30% of the area they forecast for will have rain.
You wanna show your work there? (Howmany hours are uou calling “a whole day”?)
On my planet a “whole day” has 24 hours. What planet are you from?
Oh, the actual prob. is 94.1186311…%
1-0.9^24 = 92%. And that’s only correct if each of those hour’s propability of rain is independant of all of the other hours. And I don’t see how they could be.