Who has the best forecast? This is a question we get a lot. In 2012, the No.1 smartphone task was to check the weather — and we want it to be right!
Depending on what you’re interested in, some apps and services do better than others, and here’s a website to help you decide which forecast might be the best for your weather needs and your location:
ForecastAdvisor is a third-party verification website. It is not associated with any of the actual forecasters, like Weather Channel, Weather Underground or AccuWeather. It merely grades the forecasts based on what actually occurred.
Temperature forecasts are pretty easy to slap a rank on. ForecastAdvisor considers a temperature within three degrees of the forecast a good forecast. So if Weather Underground is forecasting a high of 75 degrees, it would be correct if the actual high temperature reached anywhere from 72 to 78 degrees.
How hard is it to get a temperature forecast exactly right? Just 14.7 percent of AccuWeather’s high-temperature forecasts were perfect in 2015, according to ForecastAdvisor.
Precipitation tends to be the big question, but it’s the toughest forecast to grade. Technically, the “probability of precipitation” is a combination of spatial coverage of rain — what percentage of the area will see any amount of rain — and the confidence level that at least somewhere in the area will see measurable precipitation.
What comes out in the wash (besides bad puns) is a percent-chance it’s going to rain in your location. If there’s a 10 percent chance of rain, and it actually rains, is that a good or bad forecast?
Yes.
The thing is, you can absolutely grade rain forecasts, but the results depend on how you slice it.
For simplicity’s sake, ForecastAdvisor considers any mention of any precipitation of any kind a “positive” precipitation forecast, meaning the forecast is calling for rain whether it’s a 10 percent chance or a 100 percent chance. The score is then calculated based on how many times it rained on the days when the forecaster called for rain.
That simple grading scheme is good for ForecastAdvisor because it’s “something that was easy to understand but still useful,” said Eric Floehr, CEO of the Columbus-basedForecastWatch, the business-to-business arm of the company that offers more customization and detail.
“Everyone needs to evaluate for themselves,” Floehr said. “There is no such thing as the ‘best’ forecast, just what’s ‘best’ for your needs.”
Let’s start with Washington. AccuWeather led the forecasting pack last year, getting the high temperature right within three degrees about 75 percent of the time. But the Weather Channel and Weather Underground are in close second, with 73 percent.
There is likely no statistically significant difference between 75 percent accuracy and 73 percent accuracy. In other words, based on this metric, it’s not really going to make or break your plans if you choose one service over the other.
There are also some groups that have obviously fallen to the bottom of the pack, including the Dark Sky app, run byforecast.io. Dark Sky has a beautiful interface and some handy tools, but, at least according to ForecastAdvisor, it struggles on the actual weather.
The National Weather Service Digital Forecast is computed using model output statistics — a fairly archaic method of forecasting based on model data and climatology. It’s not touched by a human forecaster and is not typically what you would see if you visited weather.gov, so it’s not surprising that it does not perform well in most cases.
みんなの最新日記