Take a look back at NBC10 First Alert Weather chief meteorologist Glenn "Hurricane" Schwartz came to love the NBC10 App for neighborhood weather.
NO, IT’S NOT BECAUSE I’M JEALOUS
Here’s a trade secret: Most meteorologists hate most weather apps. They are the incredibly popular and wildly misused forecasts that make meteorologists seem less accurate than they really are. Rather than explain it myself, I’ll let one of the nation’s most respected TV meteorologists, Dan Satterfield do it in his great blog.
(It’s better for him to make specific accusations than for me to do it-I have a history with one of the companies mentioned. He also uses a phrase that’s perfect for the subject, but one I couldn’t get past the censors).
THE PROBLEM WITH (MOST) WEATHER APPS
No matter how much computer models improve, human forecasters can still “beat” them. The numbers that come out of various computer models are called “guidance” in our business. The top synonym for “guidance” in the dictionary is “advice”. It’s not the final word. It is a piece of advice to forecasters: “this is what the_____model suggests as the high temperature”, for example.
Different models will give different “advice”. No model is perfect. Almost all models have some sort of bias: consistently too warm or cold, too fast or slow with fronts, over-developing storms, etc. Then it’s up to the forecaster to take that all into account and make a decision. Statistics show that human forecasters can “beat” the computer models at all forecast periods (tomorrow, next weekend, etc.). So, as Dan said in his blog, you are not getting the best forecast possible when you check out most apps.
And will those apps get better? Probably not much, if at all. Who’s there to monitor those thousands of forecasts? If a particular weather model is used for the app, it will make the same mistakes OVER & OVER. The main U.S. model, the GFS, has had a bias of being too cold the farther you go out in time. It also takes a current extreme weather pattern and forces the numbers to get closer to average after about 5 days. All the time! And it’s been making the same type of mistakes for more than 20 years!
Wouldn’t it be great if there was a way to get around the idea that your possibly favorite forecast was straight out of a computer? There is! And we have it!
OK, so what’s so different with our app? I’m happy to tell you, especially because it is so much work to do what we want to do. Like other apps, we get detailed, automated forecasts for all parts of our area via our weather partner, The Weather Company (formerly WSI Corporation). Those forecasts are known as “default” forecasts. If we’re too busy with tornado warnings, for example, the automated forecasts will go out to nbc10.com and our app. But that happens less than 1% of the time.
On a normal day, we write down the “default” forecast from The Weather Company for Philadelphia. We also write down the forecasts from other computer models, including the most accurate in the world overall, the European. We have exclusive rights for some of the data from the European that shows what their “advice” is for the high and low temperatures and precipitation chances for each day. Then, based on an understanding of each model’s biases, and with many years of forecast experience, we change the “default” numbers to our own. We go out to 10 days. That forecast goes immediately into nbc10.com and our app.
That’s for Philadelphia. That is not where it ends-it’s just the beginning. We then make similar adjustments to FIVE other forecast areas: PA suburbs, Lehigh Valley & Berks, inland New Jersey, the Jersey Shore, and Delaware. Below is a map of our forecast breakdowns, so you know which area you live in (or travel to).
So, now we have in effect made 45 forecasts: 10 days for Philadelphia, and 7 days for 5 different areas. 10+35=45. But we’re not done. If we want to truly call it a “Neighborhood Forecast”, we have to go into more detail. We then chose 3 different neighborhoods from each area. 3x6=18
Our simple math is: 45+18=63
Yes, we do make 63 separate forecasts EVERY DAY. We try to figure in the effect of sea breezes, fronts that only affect parts of our area, the “urban heat island effect” of Philadelphia, the cooler, sandy soils in New Jersey, the effect of east winds on the higher elevations North and West of Philadelphia, etc. Sometimes a couple of degrees will mean the difference between rain and snow, and those differences will show up in our Neighborhood Forecasts. There are times when the forecast numbers will be close, but other days (especially in winter) where you’ll see 20-30 degree spreads across our region. Of course, our forecasts won’t be right all the time. We can’t be perfect-but we can be the best. And we strive for that every day.
This process takes hours, and a lot of concentration while the phones are ringing, promos are taped, and on-air graphics made. The web people are asking for a blog on the coming storm. The producers want to know whether the weather is important enough to be the lead story. The promo people want us to tape something for the 4pm news. Reporters want to know if lightning will affect their live shots, and we often check any scripts of weather-related stories for accuracy.
We can’t make those forecasts earlier in the day. We have to wait for the latest computer model data. So, for example, the period from 2-4 p.m. is constant, detailed, high-pressure work.
We could just sit back and let the computer models and default forecasts take over, and then go on TV implying that we’re giving you our own forecasts. But, now that we have the technology to deliver detailed Neighborhood Forecasts, we will not settle for anything but our best effort. And you shouldn’t settle for it in other Weather Apps, either.