For businesses, weather is the ultimate force to be reckoned with. It can disrupt transportation and supply chains, throttle productivity, upend pricing and demand models, and destroy property. Accurately forecasting the weather’s impact can mean the difference between profitability and loss for any given business — and the more localized the forecasts, the better.

Of course, that’s easier said than done. For the longest time, weather was considered complex to the point of being chaotic — essentially random and unpredictable. But over the past several decades, IBM has increasingly pushed the accuracy and specificity of forecasts by harnessing key advancements in data gathering and visualization techniques, computational power, analytics software and physics to create models that go far beyond the realm of traditional meteorology.

The company’s efforts have empowered countless companies and government entities to anticipate and plan for weather events while strengthening IBM’s relationships in industries ranging from aviation and insurance to public utilities and agriculture. While IBM’s efforts to calculate weather forecasts date back as early as the 1950s, the company’s focus sharpened in the 1990s with its Deep Thunder™ forecasting initiative. Deep Thunder helped establish the company as a go-to source of weather-related insights in an era of increasingly perilous events driven by climate change.

The need for such expertise has never been so acute. In 2023, the world saw more than USD 300 billion in weather disaster damages.¹

IBM’s earliest forays into localized forecasting date back to 1995, when it partnered with the National Oceanic and Atmospheric Administration (NOAA), parent of the US National Weather Service, to build the first parallel processing supercomputer for operational weather modeling. Based on the IBM RS/6000 SP and installed at the National Weather Service forecasting office in Peachtree City, Georgia, the system ran software adapted from NOAA, Colorado State University and IBM to produce multiple daily forecasts for all of the venues of the XXVI Olympiad and Paralympics during the summer of 1996 . Among the goals was the ability to predict wind velocity for diving and sailing events, and rain on Olympic parades in a given three-hour window. This successful effort led many of the government weather centers around the world to adopt IBM supercomputer systems for their operational forecast models.

The initiative underscored for Lloyd Treinish, who had come to IBM Research after 12 years at NASA’s Goddard Space Flight Center, the profound difference between creating local versus regional forecasts. As IBM’s chief weather scientist, he decided to pivot the project away from its earlier hardware emphasis to a system to assist organizations in optimizing their weather-sensitive decisions — with a primary goal of helping industry be safer, proactive and efficient. He recalled that “we started to focus on specific applications that lacked reliable, precise weather forecasts and information about their impacts.”

Treinish and the team realized that the right combination of precision forecasts and insights would enable airlines and airports to better manage the logistical nightmare of weather-generated delays. Flights could be rerouted or consolidated more efficiently. Equipped with specific information about wind, temperature and other factors, firefighters could potentially save lives and property, and farmers could increase crop yields and lower costs. More precise and accurate sunshine and wind-velocity forecasts would help anticipate the availability of renewable energy.

A year later, after IBM Deep Blue defeated the world chess champion, a journalist dubbed the IBM weather project “Deep Thunder.” The name stuck.

In 2001, the Deep Thunder team established an operational test bed on a grid of thousands of blocks in the New York City metropolitan area. Each area, 1 square kilometer in size, received highly localized forecasts based on a unique set of data and calculations.

The team also began developing a system to pair forecasts with data visualizations to help businesses make faster and smarter logistical, planning and operational decisions. As a way to better prepare a public utility for the impact of a storm, the team would mine and model historical data of damage to power lines or telephone poles. By coupling this data with hyperlocal forecasts, IBM enabled the utilities to plan for how to staff and deploy repair crews to more quickly restore power after outages.

The group soon began partnering with other analytics-driven IBM projects such as Smarter Cities. In 2010, a coastal storm in Rio de Janeiro with heavy rains and mudslides killed more than 200 people, left 15,000 homeless and caused widespread disruption of transportation systems. The storm prompted the city to develop a plan to create an operations center to improve responsiveness to emergencies, part of which involved IBM’s high-resolution weather forecasting.

Working with colleagues in the IBM Research - Brazil and IBM Research - India, the team spearheaded a project to better anticipate flooding and predict where mudslides might be triggered by severe storms. These operational forecasts were at an unprecedented scale with 1km resolution for the weather and 1m resolution for the flood risk with almost two days of lead time. The weather data was also incorporated into city information systems to determine where and when to deploy emergency crews, make optimal use of shelters, and monitor availability of hospital beds.

With colleagues in the IBM Research Lab in Dublin, the team developed and deployed the first operational system to address the integration of renewable energy into the electric grid with a consortium of utilities in Vermont. Deep Thunder forecasts enabled precise prediction of energy demand and power from individual wind turbines and small solar farms to then determine the stability of the transmission system.

The scientific, technological and business successes of Deep Thunder helped lead to the acquisition of The Weather Company in early 2016. Soon after that, IBM announced plans to combine hyperlocal forecasts developed by IBM Research with The Weather Company’s capabilities to create a new class of global weather forecasting models. The Weather Company once again became a standalone company in 2024 but its data is still in use by IBM’s sustainability software business.

In parallel to that effort, Deep Thunder continued to focus on hyperlocal forecasts — at a 0.2- to 0.6-mile resolution — and incorporate environmental data such as vegetation and soil conditions to better predict the weather, which enabled more refined predictions of weather impacts. Further advances led to new applications such as water quality, which were pioneered through The Jefferson Project at Lake George. This included Deep Thunder driving forecasts of the runoff of pollutants from precipitation events and the three-dimensional dynamics of the water flowing in lakes and reservoirs.

Deep Thunder is also used to train machine learning–based models to help businesses predict how even modest variations in any weather conditions could affect their business, from consumer buying behavior to how retailers should manage their supply chains and stock shelves or how insurance companies can analyze the impact of weather damage. Accurate, tailored and detailed weather forecasts with lead times of a few days coupled with business data can transform the way industries manage their processes. They can tailor services, change routes and deploy equipment — anticipating and minimizing the effects of major weather events on clients and constituents, reducing costs, improving service and even saving lives.

While the Deep Thunder team continues to advance the local predictability of extreme weather events, they are applying the technology to better understand the impacts of climate change. This includes the ability to evaluate the effectiveness decades from now of adaptation strategies being deployed today, and to examine how past weather events could evolve differently in a warmer future climate.