I got a strange phone call (well, a voicemail actually) on Sunday, October 22nd from a guy named Ben asking if I would research the percentages for weather forecasts. As far as I can recall, this is the first time I’ve received a phone call about my blog. As a side note to Ben, if you’re reading, I prefer e-mail.
Here’s the voicemail if you want to listen to it.
There are a few issues related to the percentages in weather forecasts. The first is whether the percentage means there is a 30% chance of any rain or if it means 30% of the area will have rain. As is pointed out in an Ask Tom question, it doesn’t matter because they both mean the same thing as far as it raining on your head. A similar question was posed to a retired lead forecaster at the National Weather Service and his reply is below.
The use of probabilities of precipitation is one of the most misunderstood elements in weather forecasts. The concept of using percentages is to provide additional information for the user to make a risk-benefit decision. For example, a contractor might decide to pour cement if the chance of rain is only 20 percent, but if it’s 40 percent or higher, he wouldn’t.
KPIX meteorologist Brian Sussman tells about his first visit to a National Weather Service office. There were four forecasters
on duty, and only one thought it would rain, making it a 25 percent chance of rain. If it were only that easy!
Actually, the usage of probabilities of precipitation is first dependent on whether there will be widespread rain or just localized showers. In the case of rain, the probability depends on what the forecaster thinks the chance of getting measurable rain (i.e., at least .01 inches) is at any given point in the area. Using computer models and professional experience, a determination is made of how often the meteorologist thinks rain will occur from a similar meteorological situation. For example, if an approaching weather system is weak, the forecaster might expect rain only two times out of 10 from similar systems, and give the probability as 20 percent.
However, with showery weather the percentages relate to area coverage. Thus, a 50 percent probability for showers would mean that measurable precipitation is expected over half of the area.
The other question of the general accuracy of forecasts is a more interesting one, and fortunately a site called Forecast Advisor has done all the work. Just enter in a zip code and it will tell you the accuracy of the various sources of weather forecasts for that area. I spot checked a few zip codes around the country and found that most are in the 60-80% range. I heard about that site from a Wall Street Journal article by Carl Bialik, the numbers guy. He makes some excellent points regarding how the accuracy is measured. Most people probably don’t care if the high temperature for the day is a few degrees off, but are more concerned if it rains when they were expecting sun. The web site doesn’t really take that into account and both would be considered incorrect.
Still, virtually all of the forecasts are better than a more simplistic approach of checking the almanac for the weather in previous years. Let that be a lesson to you. In weather, just like the stock market, past results are no indicator of future returns (or in this case, weather).