Making targeted weather forecasts highly available

TWC acquisition perfect complement to IBM’s Deep Thunder

Editor’s note: This article is by Lloyd Treinish, IBM Distinguished Engineer, Chief Scientist -- Environmental Modelling, Weather and Deep Thunder, IBM Research

Blue skies. Not just a perfect day for a picnic. But it’s also the term utilities use to describe a grid that’s perfectly balanced with energy from solar, wind, and other sources, or free of disruptions from a storm. With IBM’s acquisition of The Weather Company closing last month, utility companies could soon use TWC’s size, scale and expertise, and my team’s Deep Thunder precision forecasting, to predict those blue skies – and what they can do to rebalance their loads for days that aren’t so perfect. And that’s just one industry example.

IBM’s been in the weather business for decades. We delivered the first 701 “Electronic Data Processing Machines” to the US Weather Bureau in 1955, and got into the science with work on the mathematical and computational techniques for weather modelling in the mid-1960s. In the 1990s, several years after I joined IBM, my colleagues and I worked with NOAA to deploy weather models to help support forecasting during the Summer Olympic Games in Atlanta. That work would later evolve into what would become Deep Thunder. 

Deep Thunder is great at localized predictions for significant weather events. For example, our prediction for winter storm Jonas in New York City for Saturday, January 23 was close to the actual snowfall.

Snowfall (Inches)	Measured	Deep Thunder Forecasts
Central Park	26.8	22.0
JFK	30.5	20.2
LaGuardia	27.9	24.5
Newark	28.1	20.0

Deep Thunder made these predictions by using advanced physics and sophisticated use of a diversity of environmental data to develop high resolution, short-term forecasts. But the real power for business is how Deep Thunder can drive “coupled models,” which use the latest in machine learning and cognitive computing to predict the impact of weather. Utilities and municipalities, though, want to know not only how much snow is going to fall, but also how much wind power can they produce and use, or when and where pollution levels will have effects on people’s health.

IBM is working with power companies to combine micro weather forecasts with detailed local data to help predict where power outages will likely occur.

Putting IBM Research together with TWC is a bit like a neuroscientist working with a brain surgeon. Different but complementary skills steeped in a deep understanding of the science. My team of Ph.D. meteorologists, physicists, mathematicians and computer scientists speak the same language as TWC’s meteorologists (we’re all weather geeks at heart).  We are all looking forward to doing great things, together. – Lloyd Treinish

Pin-point predictions for everyone

Where physics and model coupling is our strength, size and scale is TWC’s sweet spot. They provide around-the-clock forecasts to industries from aviation to broadcast, and millions of consumers with high performance computing and cloud-based delivery. I envision a point in the near future where a consortium of utility companies, across multiple metropolitan areas, can couple TWC’s forecasts with Deep Thunder’s targeted models – driving our Outage Prediction and Response Optimization (OPRO) technology to determine, days ahead of time, where and when to share their repair crews so that the power can be restored from damage due to severe weather as soon as the storm passes.

And as weather events make a greater impact on our lives, businesses and industries not obviously in need of forecasts want something more than just weather. For example, a farmer wanting to reduce how much irrigation is needed, yet maximize crop yield, needs to understand the intensity and timing of rain, and the soil moisture. Deep Thunder can help do it, but in order to apply this idea to farms around the world, TWC’s industrial-scale forecast production is needed. As an initial step, my team prototyped a version of Deep Thunder, using IBM BlueMix on SoftLayer, that can be “spun up” at arbitrary locations and automatically configure itself to the local geography and weather. This way atmospheric scientists aren’t the only ones who can run a forecast for public broadcast or for a business, or analyze past weather events that had significant impact.

Future Forecasting With IoT and Mobile

Mobile devices and sensors are everywhere. But there are still too many variables to consider when trying to add reliable data, for example, from automatic windshield wipers or pressure sensors in smartphones to improve rainfall prediction. We need more sophisticated calibration to integrate such measurements. But there's tremendous potential. IBM and TWC’s coupled resources can now explore what was previously speculation, from high resolution forecasts for your local TV news, to a utility company investing in its next wind farm expansion. And perhaps soon, how to integrate IoT-generated weather measurements.

TED at IBM: This Weather Forecasting Model is Actually Accurate

IBM Research and TWC will host the 2016 IBM Weather and Environmental Sciences Conference from April 5-7 at the Thomas J. Watson Research Center in Yorktown Heights, NY. The conference is a gathering of researchers, developers, and business leaders to set strategy for the development of scientific and technical advancements related to weather and environmentally-sensitive services and business opportunities. Watch this space for more information, soon.