I’m surprised no one has discussed how they predict the weather.
There are basically two methods used to predict anything, including weather: case-based reasoning (CBR) and model-based reasoning (MBR). Both methods are often used together, and (not surprisingly) each method has its strong and weak points. Both methods rely on measurements (wind direction, humidity, temperature, precipitation, etc.).
MBR relies on a model. The better your model, the better your predictions will be. For predicting weather, your model would consist of a myriad of complex equations (based on physics) that correlate wind direction, humidity, temperature, precipitation, etc. As you might guess, this method pretty much sucks for predicting weather.
CBR is a better approach, and is conceptually simpler than MBR. To perform CBR:
Continuously take lots of time-stamped meteorological measurements (wind direction, humidity, temperature, precipitation, amount of sunlight, etc.) in a variety of geographical locations. The more measurements you take, and the more locations, the better. Store all values (including absolute time stamps) in a database. Over many decades, this database will amass millions of measurements, and will only get bigger over time.
Take meteorological measurements right now.
Search your database for past times when the measurements were very close to whatever you’re measuring now (the closer the better). For each of these times, look at what happened to the weather one day later. If it snowed in 70% of the cases, for example, then you predict there is a 70% chance it will snow one day from now.
Pretty nifty, huh?
Except there are many obvious pitfalls. Like random variability, limited amounts of data, and questionable measurement accuracies. But it’s still a pretty cool technique.
I read the Cecil Adams link and I say he’s wrong and the pilot is right.
Turn on the Weather Channel the next time the weather bureau predicts 40%, 50%, or 60%, or greater chance of rain. Look at the radar of the upcoming system, and you will see that is the % of area being covered by rain. This is empirical evidence as to what the Weather Bureau means by 40% chance of rain.
I’m not sure I agree with Cecil regarding his statement about the weatherguys being right 100% of the time. Under this definition, you could also be right 100% of the time, simply by figuring out the average number of days in a yearthat it rains, and predicting this percentage likelihood every single day. The point of the weather people is to predict which days will diverge from the average, and by what likelihood. Now suppose under conditions such as we have now it has historically rained 30% of the time. Presumably there is a reason why it has rained on some of those days and not on the other 70%. To the extent that the weatherguy has failed to differentiate which are which, and has failed to see the reasons why on this particular day it will/will not rain, he has failed as a predictor, and the fact that under his self-defined set of conditions he is right about the probability is not too meaningful.
Imagine the increase in viewers if rumor has it, “THE STORM OF THE CENTURY WILL STRIKE (insert City here) TOMORROW!!!”
Living in Seattle,WA. I have some sympathy for the local weathermen in the area. As an Iowa native and Navy meteorologist told me once, “when you live in the Midwest, you just look to the West and it is relatively easy to predict what weather will come tomorrow.” In the Pacific Northwest, we have a gigantic ocean, swirling and churning, pumping low and high pressure systems at us in a chaotic and random pattern.
Regardless of my lack of faith in weathermen, I still tune in when the threat of a storm is discussed around the water cooler. Therefore making the advertising on my local T.V. station more valuable, which makes it easier to sell more ads, etc., etc. ad nauseum.
Sensationalism sells in all aspects of news broadcasting, please leave the weather alone!
I’ll recommend “The Fortune Sellers” by William A Sherdan
it has a chapter on meterology/meterologists.
The way predictions are presented is important. Short range predictions (up to 3 days ) are pretty accurate, but are often still presented in vague terms e.g. “40% chance of precipitation”, “mostly overcast” or “sunshine and showers”, so can be claimed as accurate whatever happens, especially since the areas defined are also left fairly vague e.g. “the east”.
Long range forecasting is nigh on impossible, although it doesn’t stop a lot of almanacks claiming that they can do so sucessfully. Longjohn was right regarding perception, your view of accuracy will be coloured by what you want to believe and how much you want to believe it, rather like visiting a psychic.
I’m not knocking weathermen - predicting is difficult, and meterology is probably the best researched and best defined of the various prediction businesses (much more so than the stock market, for example). I research into predicting about software, eg development effort or defects, and have respect for and indeed envy of those guys.
Altho the forecast for the amount of snow NYC was to receive was inaccurate, it did snow in NYC, and most of the forecast region got the predicted amounts. It really is amazing to see how far meteorology has improved with the use of computers and more knowledge. They were able to predict the general parameters and features of the storm days before the Low that produced all that snow formed in the Atlantic. To even predict the formation of the Low is amazing. Beats the old days when predictions were based upon how much grass the cows were eating.
I see now that the Weather Bureau (www.weather.com) now has a 10-day forecast. As noted by Sir Doris, 3-day forecasts are quite accurate, but a 10-day forecast? Don’t take it to the bank.
Hey…just because you didn’t get hammered in the city, doesn’t mean the rest of us didn’t get what was predicted. Here in central NY state, we got ~2 feet from that storm, which is almost exactly what was predicted for us.
I support the kids rattling off Chaos theory mumbojumbo. Basicamente, we gots a big butterfly of possibilities that constitute the entire history of weather, and all the constraints for future weather. If you take the current weather as a line or path within one wing of johnny butterfly and the weather continues on that path, then the weatherman can get the facts right for damn once. Howeva, if a disturbance pushes a weather system off its current course, we may end up wearing shorts on a snow day. The weatherman gives you probabilities because that the best dude/dudette can do…they’re probabilities of the the path not fluctuating around, so really they’re probabilities of maintaining the current weather within the butterfly constraint.
Lots of info about paths,no? Taoism and chaos theory go together like peas and carrots, man.