Although weather forecasting has become increasingly accurate due to radical advances in data science, the ability to deliver hyper-local and contextual information to businesses, governments and consumers remains elusive. Agencies such as the National Weather Service and European Center for Medium Range Weather Forecasts aren’t designed or equipped to deliver the granular data that’s increasingly in demand.
As a result, The Weather Company, an IBM Business, is now looking to take forecasting to a more granular and practical level by bringing “high-resolution forecast models to localized areas,” says Mary Glackin, senior vice president for Science & Forecast Operations at the company. The system currently produces forecasts for about 2.2. billion locations worldwide, totaling as many as 26 billion forecasts per day.
The Weather Company uses sophisticated models that analyze data from around the world. It relies on more than 100 terabytes of third-party data from Weather Underground’s network of nearly 200,000 personal weather stations, as well as atmospheric pressure data from smartphones and traditional radar and satellite data.
By plugging all this data into a cloud-based high-performance computing environment, The Weather Company generates specialized forecasts that can be put to use by businesses—such as commercial airlines, agribusiness, engineering firms and insurance companies—as well as government entities.
For example, “It’s possible to use high-resolution models to fine-tune flight schedules and more fully understand the costs associated with different cancellation options,” Glackin points out. Similarly, the technology can be used to better understand where to locate wind farms or solar facilities, and to help utilities use historical data about storms with hyper-local forecasts to better prepare repair crews in preparation for a hurricane or other emergency.
The Internet of Things and Machine-Learning Models
These detailed and customized forecasts—which wouldn’t be possible without the internet of things (IoT)—help analysts and business leaders better understand the full impact of a weather system in a particular location. That should result in faster and better decisions.
The technology can also target areas retrospectively by using machine-learning weather impact models. This capability helps organizations better understand how various weather conditions impact demand, sales and their supply chain.
The Weather Company delivers refreshed data and guidance every three hours. The underlying algorithms are highly customized using hyper-local data and criteria—typically providing visibility at a resolution of 0.2 to 1.2 miles—and take into account other relevant environmental data such as vegetation and soil conditions. Glackin points out that the real-time data is valuable for organizations as they try to more precisely predict how even modest variations in temperature, rainfall, humidity, air quality and other factors could affect business conditions.
“This hyper-local platform takes weather forecasting to a level that allows businesses to act and react rapidly,” Glackin concludes. “As we see regions such as South America, the Middle East and Africa adopt personal weather stations—and the underlying models improve—the way we view weather data and use it to make decisions will evolve and change.”