Weather (inspired by Sharky)

On Friday, weather.com reported that the chance of rain on this coming Saturday — a day of no small importance to me — is 0%.

Now that’s a bold statement, of course, but a reasonable interpretation of it is: “There is no way it is going to rain on Saturday”.

When I checked the weather again today, weather.com now reports a 20% chance of precipitation.

How is this possible? My not unreasonable interpretation of the precipitation chance number N on day D is: “There is an N percent chance that it will rain on day D: given the initial conditions present today, if day D were to occur 100 times, it would rain on N of the times.”

By this logic, 0% and 100% are very dangerous numbers. If a percent ever falls to 0% on a given day in advance of D, it can never leave 0% on subsequent days before D. Same with 100%. There is no occurrence that could induce a change. (Assuming arbitrary precision, which we’ll discuss later.)

I think what weather.com really means is: “Given what we know now, there’s an N% chance of rain on day D. But we may know more later.”

Also reasonable, but a really irritating interpretation for those of us who just want to know what the weather will be like. We don’t want to have to worry about what weather.com does or doesn’t know. And I think weather.com could account for this uncertainty by an empirical analysis of past 0% and 100% claims. How often did they have to change that number before day D? Add in a little factor that accounts for this uncertainty.

I don’t know enough about weather prediction to make any better suggestion at the moment. Also, I suspect that their desire to not appear overly precise, manifested in their use of only a single digit of precision in their percentages, is a real problem here. Anything under 5% might appear as 0%.

So that’s another good excuse. But the fact of the matter is that when I see 0%, I want to be able to plan my beach party with confidence. Don’t destroy my confidence, weather.com.

This entry was posted in Uncategorized. Bookmark the permalink.

3 Responses to Weather (inspired by Sharky)

  1. wingerz says:

    you know what else kills me? when it says “20% precipitation” for every hour of the day when they mean it for the entire day.

  2. Even under the “Given what we know now, there’s an N% chance of rain on day D. But we may know more later” interpretation, your observation holds up — if the chance of rain ever drops to 0%, it should never again rise (unless the 0% is just rounding error, and they really mean 0.01% or whatever).

    Of course, this is attaching way too much mathematical credibility to weather.com.

    • aj says:

      Well, I imagine weather to be a function of a few thousand dimensions. They may just be missing a few. But yeah, the whole thing is suspect.