How high the temperature?February 25, 2011 12:26 PM   Subscribe

I can't believe I'm asking this, but I couldn't find an answer anywhere: what do the daily high and low temperatures provided with weather forecasts mean? I always thought they put a boundary on the range within which the temperature is predicted to fluctuate, with no particular assumption that those temperatures would be reached; however, I've recently had a few friends tell me that those numbers represent the predicted maximum and minimum temperatures that will actually be reached in the course of the day. Which is it?
posted by invitapriore to Society & Culture (20 answers total) 6 users marked this as a favorite

What's the difference between those two things? I would say your friends are more right in that that phrasing makes more sense (it's predicted to go to that maximum temperature in the afternoon, and that minimum temperature overnight), but that's the same thing as saying it's the boundaries of that 24-hour-period's temperature fluctuation, isn't it? What would those numbers indicate in what you mean if not the... well... predicted high and low temperatures?
posted by brainmouse at 12:29 PM on February 25, 2011

I think your friends are right.

posted by jindc at 12:30 PM on February 25, 2011

yes what (above) said
posted by joegardy at 12:35 PM on February 25, 2011

Response by poster: brainmouse: The second is a stronger assertion. I can't think of any way to put this except mathematically, so I hope this is understandable, but imagine a function f(t) that gives you the predicted temperature for any time of day you provide. The way I understood it was:

For all times t in a given day, we expect LOW <= f(t) <= HIGH.

The way my friends put it is:

For all times t in a given day, we expect LOW <= f(t) <= HIGH, AND we expect there is some time t1 for which f(t) = LOW and some time t2 for which f(t) = HIGH.

Does that make sense?

One other thing: if possible, could you provide cites?
posted by invitapriore at 12:37 PM on February 25, 2011

Response by poster: Ah, I should have previewed. So, zamboni, does that mean that the highs and lows are temperatures that have already been observed in the previous six, twelve or twenty-four hours prior to the time when you get your forecast?
posted by invitapriore at 12:39 PM on February 25, 2011

i think zamboni's glossary link and your question are two different things. the glossary definition is for recorded maximum temperature whereas your question is regarding temperatures provided with forecasts.

that said, i've always understood it to be the way your friends describe it. the weather forecast predicts the high/low temperature to equal the high/low given.
posted by karmaportrait at 12:50 PM on February 25, 2011 [1 favorite]

Best answer: Your friends are right. You had "no particular assumption that those temperatures would be reached", but for a weather forecast there must be some basis on which to choose the predicted high and low. The chosen temps are based on past measurements for that air mass, computer models, and knowledge of local conditions.

Zamboni's citation seems to apply more for recorded highs and lows, and may or may not apply to the forecast.
posted by TDIpod at 12:51 PM on February 25, 2011

Best answer: Interesting question and not as trivial as it first sounds.
I can't find a definitive answer, but I'll add my observation that in order for high and low to be meaningful and useful, it should follow the second definition. I could set LOW to be -50 and HIGH to be 100 and that forecast would be correct by the first definition but not the second. The precision (one degree) and daily fluctuation of these forecasts in practice strongly suggests they are intended to be the tightest bounds.

I think zamboni's link does not apply to the forecast, but rather to previous observations. You could think of the forecast as "predicted daily maximum" and "predicted daily minimum", using those definitions of daily maximum and daily minimum.
posted by PercussivePaul at 12:52 PM on February 25, 2011 [1 favorite]

My understanding is and always has been that the forecasted high & low temperatures are indeed expected to be reached at some point during the day. Sometimes they are wrong, and the actual high or low exceeds (or falls short of) the forecast.

If your interpretation was correct, couldn't the forecasters just always say "the low today will be -100F and the high will be 150F" and be technically correct? This is an exaggerated example, of course, but it gets my point across.
posted by Nothlit at 12:53 PM on February 25, 2011 [1 favorite]

Presumably "forecast" highs and lows are predictions of future temperatures whereas "observed" highs and lows are actual temperatures observed for a given time-period that has already transpired.

It seems that your friends' interpretation is obvious if you've ever heard a TV weatherman talk about "overnight lows in the 30s but comfortable daytime highs in the 60s" or anything similar to that.

Reading from weather.com's description of the forecast in my current location for "tonight":

"Rain showers this evening mixing with snow showers overnight. Low 34F. Winds W at 10 to 15 mph. Chance of precip 50%."

And for "tomorrow":
Sprinkles or flurries possible early. Partly cloudy skies. High around 50F. Winds NNE at 5 to 10 mph.

I don't know how that can be interpreted in any other way then "tonight, it will get as cold as 34F, and tomorrow during the day it should warm up to about 50F."

Your interpretation of that meaning "the temperature will probably be between 34-50 for the next 24 hours" seems bizarre to me. It implies that the more precise a forecast is, the tighter the range of numbers would be, i.e., a forecast for 42-42 degrees would imply that the temperature would range anywhere between those two values, not necessarily reaching either one. Since they're so close. this implies a ridiculous level of precision given that it's obvious the temperature fluctuates by more than that during a day in most locations around the world.
posted by tylerkaraszewski at 12:53 PM on February 25, 2011

Best answer: In other words, invitapriore's view is that the given high and low temperatures are outside of the bounds of another number, the actually known (predicted) high and low. I believe the common interpretation is that shared by the others here, the friends are correct.

But, that's actually "more or less correct". The prediction is an estimate, so it of course has some degree of error involved. Is saying "the low is 42 degrees" somehow a guarantee that it will be exactly 42 degrees? Well, no. Similarly, a bound can't be guaranteed, either--if for no other reason that there is always going to be a possibility that the temperature could go lower/higher.

Also, saying that the temperature is going to be between two extremes is not all that useful; you could just pick +-300 degrees F and be pretty sure you're right, but no more enlightened on whether to wear a jacket. You want to know how cold it's expected to be.
posted by RikiTikiTavi at 12:54 PM on February 25, 2011

You might think about it this way: if your interpretation were correct, those upper and lower boundaries could be arrived at by only one of two processes:

1. Picking ridiculously extreme limits that would be almost certain not to be exceeded whatever the weather;

2. Using your meteorological knowledge and forecasting skills to predict the likely highest and lowest real temperatures.

1. would be pointless; 2. is the same as your friends' interpretation.
posted by oliverburkeman at 1:08 PM on February 25, 2011 [1 favorite]

Response by poster: In retrospect, after y'all's explanations, my initial belief does seem a little ridiculous, given that a weather forecasting model capable of providing a predicted temperature range probably can't provide that range without having already predicted a maximum and minimum temperature for the day.

I guess I always just thought of the high and low as being expressions of the fact that the temperature would not go higher/lower than the given number with 99 (or whatever high number of your choice) % certainty, but as some of you have pointed out that's not a very useful statement.

I work in a science lab, so I've kind of been conditioned into assuming that the weakest version of an assertion is what was meant, but that's probably not always the best tactic for everyday life. Thanks!
posted by invitapriore at 1:22 PM on February 25, 2011

Right, it would make some sense if the high and low were basically error bars, say the 50% confidence interval or something. But I think your friends are right, and they are the predicted maximum and minimum temperatures over the course of the day (or nychthemeron :) ).

From reading weather forecasts and weather blogs I notice that meteorologists will usually predict a range for precipitation or wind ("1-3 inches snow", "gusts of 35-40 mph", eg) but temperature predictions don't come with an associated uncertainty or range. I don't know why that is.
posted by hattifattener at 1:43 PM on February 25, 2011

What you're really asking is how much confidence you can put into the high and low numbers. The concepts you need to understand are the target value, the error (or uncertainty) and the confidence in the uncertainty. Any prediction is a guess, the question is really how good that guess is.

A short bit of theory: a predicted temperature T, 20 degrees, might have possible range of values +/- t, say 5 degrees, 95% of the time.

T is the target value, a predicted high or low temperature, for example 20 degrees. This is the most likely or average predicted vale---not always the same thing, but that's the simplest case.

+/- t is the range the actual temperature is likely to be in: T-t to T+t. In our example, the temperature would be expected to be between 15 and 25 degrees most of the time. This range, t, is called the "uncertainty" or sometimes the "error".

Most of the time? Yes: the actual temperature will be in the predicted range 19 days in every 20. This is called "confidence". The ratio 19/20 or 95% is the default "certainty" in almost all reported measurements. Sometimes 99.5% is used too: the wrong answer 1 in 200 days. The trade off for more confidence is that the range of uncertainty, t, is much bigger.

This is complicated, which is why they don't do it in weather reports. The forecasters can calculate these numbers however, and present their best predictions of temperatures, T, without telling you the ranges of uncertainties or the confidences in the numbers. You can see a set of typical ranges here though.

The predicted temperatures are marked at the centre of of each vertical rectangle. Those are the values T. The ranges, t, are shown by how tall each rectangle is. Notice that the bars get bigger, the father into the future the prediction goes. Longer-term predictions are less certain than near-term predictions.

So, the answer to your question is that what's given on the weather report is an intentionally simplified version of a prediction. The range of variation and the confidence in that range appear implicity though. The cutoff for the number of days in the forecast is the maximum uncertainty the office is willing to report. Usually this is a few degrees. Too much more, and the predicted temperatures aren't very useful.
posted by bonehead at 1:47 PM on February 25, 2011

There are no confidence intervals with weather predictions, and many people share your understanding of what the forecast means. I'm with the group that says those are approximate values of what the temperature will peak at during the day.
posted by blue_beetle at 2:02 PM on February 25, 2011

It seems to me that these are just 2 different ways to phrase the same thing. In other words, if someone says temperature will fall within such and such range, the prediction makes no sense unless it means that with some degree of certainty, temperature will reach both lower and upper bounds. In other words, could they say: "We have absolutely no certainty whether temperature will or will not reach lower or upper bounds, but it will be within this range"?

The very idea that there is an imprecise prediction that temp will be in a range inevitably means that there is an imprecise prediction that it will reach both limits, unless the range given is so unhelpfully wide that you'd need an asteroid strike or something to reach the limit.
posted by rainy at 2:34 PM on February 25, 2011

« Older You have dancing shoes with nimble soles.   |   Festivals / must sees within an evenings train... Newer »