IBM Research’s AI-driven risk and impact predictions help businesses adapt to climate change
This article is the second of a two-part series. We’re exploring how IBM Research is confronting climate change, helping businesses improve sustainability and deal with the negative impacts of the changing climate.
This article is the second of a two-part series. We’re exploring how IBM Research is confronting climate change, helping businesses improve sustainability and deal with the negative impacts of the changing climate.
Hendrik Hamann remembers hearing the forecast the morning of September 17, 1999, and wondering what “several inches of rain per hour” looked like. Hamann, just nine months into his tenure at IBM Research at the time, would soon find out. Hurricane Floyd turned the T.J. Watson Research Center’s Yorktown Heights campus — and much of New York’s Westchester County — into a swirl of sideways rain and swaying (or uprooted) trees.
“I had no concept of what the forecast meant until I actually saw it,” recalls Hamann, a distinguished researcher and chief scientist for IBM Research’s Future of Climate initiative.
“Most roadways were unpassable because they had turned into massive running streams,” he says. “It turned out the only way to get home was to park my car and then to basically walk around a massive lake, which was too deep to drive through.”
Efforts to explain climate change’s profound impact on the planet often meet the type of reaction Hamann had prior to Floyd. “Someone can tell you that in 10 years you’re going to have 10 more days of 40°C (104°F) weather per year,” he says. “But that number doesn’t tell you the kind of sustained increase in temperature has — on the economy, on your life.”
As much of the world struggles to grasp how climate change will affect our lives, Hamann and his IBM Research colleagues are developing AI and cloud-based tools to help businesses understand and adapt to increasingly severe climate conditions. “In general, there’s a realization that it’s time to wake up, and it’s time to do something,” he says.
“Everyone is now experiencing climate change in some way. It’s already here.”
Launched in 2020, the Future of Climate initiative’s goal is to accelerate the discovery of climate change solutions. That includes longer-term approaches to mitigating humanity’s negative influence on the environment, as well as near-term adaptation to help organizations adjust their operations and resources for climate changes already underway. This latter objective is where Hamann and his team focus most of their effort — applying AI, cloud, high performance computing, and data to help IBM clients predict and prepare for increasingly severe weather events.
“Businesses have managed to digitize finance, social networks, media, and other industries, and you see how they’ve been impacted — they changed and became smarter,” Hamann says. “That’s what IBM can do for the environment. We are now in full flight digitizing the physical world.”
IBM plays multiple roles in promoting adaptation. As the world’s largest provider of weather information, IBM helps the agriculture, energy and travel industries make real-time decisions for things like harvesting, infrastructure investments, and transportation scheduling. The key to successfully navigating consistently warmer, drier seasons or more extreme coastal storms is using data about the changing environment to create new business processes that account for those changes.
Today’s technology struggles to integrate massive layers of data collected by satellites and other sensors.
“Take wildfires: You may have thunderstorms that set them off, but how much do you know about the condition of the soil around those trees, the amount of surrounding brush and other factors that contribute to the intensity of such an event?” Hamann asks. “That level of information represents terabytes of data that have to be put into a geospatial context to be meaningful.”
IBM introduced its Physical Analytics Integrated Data and Repository Services (PAIRS) in 2018 to help address that challenge, and it now plays a central role in the company’s adaptation offerings. The technology aggregates and analyzes massive amounts of data from aerial imagery, drones, Lidar, and satellite data to predict the risk, resiliency, and potential impact of upcoming weather events. Information like this can help government agencies, companies, and other organizations better anticipate and understand how changes in seasonal flooding and other forces of nature will impact their supply chains, agricultural production, and regional operations.
PAIRS anchors IBM’s new Environmental Intelligence Suite (EIS), a set of AI-driven applications to help organizations prepare for and respond to weather and climate risks, while also better assessing their own impact on the environment. This is the first time such a comprehensive set of weather, climate modeling, and carbon accounting tools has been brought together in a single offering available via the cloud.
PAIRS also serves as the foundation for IBM’s Climate Impact Modeling Framework (CIMF), which is part of the EIS. CIMF uses the geospatial data analytics platform for AI models to predict the risk and potential impact of upcoming weather and longer-term climate hazards in a way that’s much more efficient, standardized, and integrated than existing methods.
IBM created CIMF to help researchers run data models, regardless of whether the information is geospatial or temporal, says Anne Jones, a member of Hamann’s team who is based at the IBM Research Daresbury lab, in the UK.
“If we want to examine drought, for example, we just put the data in the framework,” she says. “The idea is for clients to be able to interact with CIMF via API calls. Those companies can use the technology directly, rather than having to feed their data to IBM.”
As part of Future of Climate, Jones and her UK colleagues played an important role in developing CIMF. “For the initiative, unlike a lot of modeling, which is done independently by researchers according to their needs, we focused on building CIMF as a core framework that could be used more broadly,” she says.
“I’ve always been interested in how modeling and technology can help with better decision making,” says Jones, who earned a Master of Science in weather, climate, and modeling at the University of Reading in England, before earning her PhD in climate-sensitive disease forecasting from the University of Liverpool.
“One reason infectious diseases are more difficult to forecast than the weather is because you typically have more unknowns and less data to work with," Jones says. "With weather and climate, we have vast datasets from simulation models and satellites. The challenge is how to leverage them optimally for any given use case.”
Jones focuses on impact modeling, using data-driven AI and physics-based simulations to translate weather and climate variables to hazards such as floods, heatwaves and wildfires, and then to their consequences to human and natural systems. Her goal is to generate robust, accurate, and spatially precise information on these impacts that can then be used for risk quantification. This could, for example, enable an insurance company to combine IBM’s weather data with different flood models and its own data on a property’s location, vulnerability, or infrastructure to create more accurate risk assessments.
Jones’s risk models rely heavily on climate variables she gets from fellow IBM researcher Campbell Watson. “Adaptation is accepting that there will always be risk and finding ways to adapt to that risk,” says Watson, who joined IBM Research in 2014 as a postdoctoral research scientist. Following a stint with the Forecasting Sciences team at The Weather Company, an IBM Business, Watson turned his attention to physics-informed AI to enhance weather and climate predictions.
“These technologies are especially important to adaptation because we can no longer rely on information about past weather patterns to predict future weather, and yet the future feels very uncertain,” he adds.
Adaptation is particularly important to ensuring businesses have resilient supply chains, an aspect of business that’s saturated the news this year thanks to the ongoing chip shortage and its potential to disrupt several different industries for years to come.
“Businesses need to diversify their supply chains, particularly as some regions of the world begin to experience more extreme flooding or extreme droughts,” says Watson, who is also part of the Future of Climate initiative.
Watson, who grew up surfing the coast of Melbourne in Australia, became interested in weather at a young age. The devastating Black Saturday bushfires that swept across Victoria in February 2009 heightened his interest in extreme environmental phenomena. Among the continent’s worst-ever bushfires, the combination of an intense heatwave and high winds helped ignite hundreds of fires that resulted in 173 fatalities and left many more homeless.
“The whole state came to a halt,” Watson recalls, adding that his experience in the aftermath of the fires motivated him to turn his interest in weather into a career. After earning a PhD in atmospheric sciences at the University of Melbourne, Watson pursued his postdoc at Yale University, where he studied the physics of mountain-triggered convection and precipitation above Dominica, a small volcanic island in the Lesser Antilles.
An area of research that Watson is particularly excited about is weather generation. IBM WeatherGen, created in collaboration with a team based in the IBM Research Brazil lab, uses deep generative modeling to create synthetic weather scenarios uncannily similar to the real thing. IBM WeatherGen is integrated with CIMF and provides extreme weather scenarios to various impact models.
“Our weather generators can supplement the observational record by generating a raft of 1-in-100-year storms,” Watson says. “It is helping us understand and characterize what could happen during extreme weather events.”
Watson and his colleagues are also building an AI-based streamflow modeling system to help with water management challenges, including flooding. The physics-informed AI model learns how to emulate water flowing through a river network. The system, integrated with PAIRS and CIMF, enables the team to quickly build physically consistent and reliable AI models that are accurate and very fast to run.
“Its unique architecture ensures an exciting realism: the model enforces physical constraints, like the conservation of mass, and mimics the fast and slow responses of water,” Watson says.
The importance of climate-specific AI cannot be overstated, Hamann says. “We’ve been analyzing climate data using general-purpose AI that doesn’t necessarily consider the context — doesn’t factor in the laws of physics — and therefore hasn’t given us the best results,” he adds. “CIMF and a lot of the research we are doing focuses on adapting AI specifically to climate problems.”
Adaptation influences how businesses go to market, how they interact with customers and how their brands are perceived, Hamann says. “That’s not something you can address in one project; you have to have a strategy,” he adds. “You must consider what investments you need to make now to be ready for the world as it will be in a year, in five years.”
To do that, people must understand the meaning behind the numbers.
“After Hurricane Floyd, if I hear in the forecast it’s going to be more than five inches of rain, you better believe I won’t be going to work,” Hamann says, thinking of IBM Research’s role in helping businesses adapt to climate change. “I’ll be buying sandbags for my home.”
Watch the replay: Emerging technologies to make our world more sustainable
To learn more about the ways that IBM Research is developing cutting-edge climate change mitigating technology, watch a replay of the November 24 panel discussion from The Future for AI & Quantum for Accelerated Discovery event. IBM researchers focus on scientific breakthroughs in materials science, cloud computing and AI aimed at helping the world deal with the negative impacts of climate change and improving sustainability.
- Dr. Peter Staar, Manager of the Scalable Knowledge Ingestion group, IBM Research Europe-Zurich
- Dr. Kommy Weldemariam, Impact Science, Future of Climate, IBM Research–Africa
- Dr. Edward O. Pyzer-Knapp, STSM, WW Research Lead, AI Enriched Modelling and Simulation, IBM Research Europe-Daresbury
- Dr. Matteo Manica, researcher in Accelerated Discovery, IBM Research Europe-Zurich
- Dr. Laura Gardiner, researcher, application of ML and informatics for life sciences, IBM Research-Daresbury \
AI for Business Automation: We’re working on tools to help businesses automate much of their daily processes, to save time and drive new insights through trusted, safe, and explainable AI systems.