Atlanta— If you go to The Weather Channel to check out the sun or rain, you get temperature, precipitation, wind and other conditions for one of 10,000 places across the country. Pretty good.
That’s about to change dramatically. On April 4, The Weather Channel put into operation a new generation of its High Resolution Aggregated Data, or, HiRAD, system, that will calculate conditions for 1.9 million distinct points across the country. Move over two miles and your weather report will change.
“You have the ability to zero in on something,’’ says the channel’s chief information officer, Brian Shield. This, even if the reporting station nearest to the ski slopes in Vail, Colo., is 40 miles away. “You can now start talking about what the weather is in locations where they don’t have forecasts.”
It’s “backyard weather,” as Ian Miller, senior vice president of weather systems, put it — or at least the mathematically generated equivalent.
And it’s being driven by the new types of screens to which The Weather Channel must cater. BlackBerry e-mail devices, mobile phones, desktop computers and portable equivalents are causing the meteorologists at TWC to produce “personally relevant weather” for every viewer, on each device.
Now, “we have to build platforms that allow the consumer to be in charge,” said Miller.
The Weather Channel used to pretty much repackage National Weather Service data. Now, using its HiRAD system, its is creating its own highly localized sets of weather reports that pinpoint conditions between home, school and office. And with a set of computer models called The Weather Channel Operational Post Processing System, or TOPPS, it’s created a series of its own factories for creating forecasts that Miller believes match or beat the federal ones.
Which is fine with the National Weather Service. “We are not in competition,” said Dennis Feltgen, a meteorologist and public affairs specialist at the National Oceanic and Atmospheric Administration, which operates the service.
To develop its own forecasts, The Weather Channel since 1998 has been building its TOPPS system on computer servers at its headquarters here.
The system was designed and written largely by one weather systems manager, Dave Faciane. The 14 forecasting factories in the system take in measurements of the state of the atmosphere such as heat, moisture, pressure, motion and wind, from 1,500 airports and other National Weather Service reporting stations. Each factory turns out slightly different results, based on the mathematical formulas and statistical methods of forecasting.
This “statistical ensemble” is then melded into an initial unified forecast, weighted essentially by how well each factory has performed in the last 90 days. Human weather forecasters then get a shot at the data through a “digital global intervention tool,” aka Digit, that lets them reshape and edit the final forecast.
The current conditions are also built on ground observations from the 1,500 reporting stations. Those observations then can be triangulated to give a sense of conditions every 2.5 kilometers across the country.
But the ground observations are compared to information from the sky: radar and satellite images. The satellite images show clouds, almost to the inch. The radar shows where precipitation is occurring.
Comparing the radar results to the ground observations calibrates what’s going on, according to Bruce Rose, vice president of weather systems. Sometimes it does represent what’s going on and sometimes it doesn’t; constantly comparing the ground and air observations and mixing in lightning strikes in the last 10 minutes provides the final calculations for each of the 1.9 million points.
Make no mistake. These are “synthetic” reports of conditions: Extrapolations and interpolations based on actual readings at the 1,500 reporting stations and what is known about weather patterns, elevation, clouds and other factors. But that does not mean they aren’t as sound as the real thing, if the models and comparisons are working well, Miller said.
“Simulated fur is inferior to fur. Synthetic (weather) observations should not be inferior to observations,’’ contends Miller.
They certainly can be more frequent. The Weather Channel used to only generate three forecasts a day; now it is prepared to generate as many as 22. And current conditions? The HiRAD system aggregates and chews on reporting station data, radar and satellite imagery three times an hour. So it creates new “synthetic current conditions” every 20 minutes.
It’s a “cutting-edge system, with the intention of filling in the gaps that the current observation network leaves open,’’ according to Thomas Downs, director of Expert Weather Investigations, a New York consulting firm.
But the “gaps” aren’t what’s motivating Weather Channel so much as the drive to produce a weather service that suits several different delivery “platforms,’’ according to Miller. In this case, the latest platform to be reckoned with is the phone carried around in pocket or purse at all times.
“The sweet spot on forecasting, particularly now with mobile devices, is the next three to six hours,’’ says Miller. “Telling someone it’s just going to be nice this afternoon and a high of 81 is no longer totally relevant.”
Tailoring information for each kind of device thus becomes critical. Web pages feature five and seven-day forecasts; and 90-second reports from talking meteorologists. Mobile phones get three-day forecasts and 60-second clips, according to vice president of broadband products Jody Fennell. One large data table feeds all devices.
All this leads to 30 trillion bytes of weather information now being stored by The Weather Channel at any given point. That will reach 45 trillion by year’s end.
“All the different platforms are demanding to consume the weather in different ways. It forces us to be much more extravagant in how we approach creating our content,’’ said Miller.
The saving grace? Relatively little data is kept, when each 20 minutes is up.
“After ten minutes, it’s useless,’’ Miller said.