03 August 2008

Don't let the National Weather Service ruin your party

What good is a prediction if it's subject to change during the time period it is supposed to be useful? A prediction that keeps changing by large amounts before a predicted event happens (or doesn't happen) is practically useless for long-term planning and is an indication of either of 2 things:

1. The "predictions" are actually wild guesses or made-up numbers to satisfy those who are expecting a prediction.
2. The phenomenon that is being predicted is the outcome of the complex interactions of many factors that may be subject to random fluctuations and is, therefore, inherently unpredictable. Consequently, the "predictions" are actually wild guesses or made-up numbers to satisfy those who are expecting a prediction.

A typical example of changing predictions was provided yesterday by the National Weather Service. Early in the morning, when it was quite cloudy, they were predicting a 100% chance of thunderstorms for where I live.


Early in the afternoon, it was starting to get sunny and the chance of thunderstorms was now 60%.


The initial "100% Tstms" meant that we were absolutely, definitely going to have thunderstorms: if it had been possible to observe a large number of weather conditions identical to the one we had yesterday morning, on the average, each and everyone of those conditions (100%) would have produced thunderstorms. But when the "prediction" later became "60% Tstms Likely", it became clear that the earlier predicted absolute certainty was baloney, because it had been subject to change. If you say to me "The chances are 100%, but that may change later", then the chances are not 100% and the only thing that is certain is that you don't know what you are talking about.

If a prediction keeps changing, then it's not a prediction. The NWS's lowering of its prediction from 100% to 60% was quite substantial and clearly illustrates the unpredictability of complex events like localized thunderstorms.

General weather patterns are predictable, but are not of much use. For example, predicting that there will be some thunderstorms somewhere in the state of Maryland sometime during the next 24 hours is neither impressive nor useful.

It was important for us to have an accurate weather prediction for yesterday, because we had an outdoor party planned for the late afternoon. After I learned about the 100% chance of thunderstorms, we briefly considered postponing the party, but then decided to go ahead with it anyway.

At the end, we never had thunderstorms; it didn't even rain. We did feel a few drops for a few seconds at one point, but nobody even bothered to get up and go inside. So much for predicting the weather.

The party was good with plenty of food and wine.



Duane Smith said...

I wonder what how often their 100% predictions turn out to be right. 50% of the time? More? Less? I'm glad the party when well. How often is that the case? Nearly 100% of the time.

Duane Smith said...

What is that "what" doing there after "wonder?" And then "when" should be "went." I think I will go back to bed.

Anonymous said...

Hi Aydin,

I understand your frustration. However, as you know I am sure, weather predictions are only educated guesses, based on the available evidence. Being that the atmosphere is a fluid dynamics problem, it has far too many variables to be modeled accurately.

When they said "100%" chance of thunderstorms, they meant that it seemed that a big front with uninterrupted line of thunderstorms was guaranteed to roll across your whole area. But as it happened, in reality however, the line broke up into areas, with gaps between, and you went through a gap in the line and got almost nothing to speak of. The same thing happened here in NYC.

In my case I look at the radar picture and make my own predictions as the events unfold.

There are often times when the weather is more stable, and then the predictions are more reliable. Recently the weather patterns have been very unstable.

I have thought long and hard about whether the NWS could come up with a better alternative than saying "100%"...

I think the thing they need to stress is that the whole weather forecast itself is a prediction, an educated guess, a best guess, and it is never a certainty. After all they are not clairvoyant. So in effect they are doing something like making a 85% prediction of a 100% chance of thunderstorms, something like that.

best to you,

Susan H.

Funder said...

My parents live in northern Mississippi, and my dad is a "weather addict." He pops in at the computer probably 10 times a day to check the radar and see what the NWS has to say.

My dad was most amused all winter, because somebody kept a "30% chance of snow" two days out in the forecast from November til March. In reality, it's about a 30% chance of snow ONCE A MONTH in December, January, and February. He thinks the perpetual chance of snow is just to keep the kids excited. :)