2020 saw the worst wildfire season on the West Coast of the U.S. since records have been kept. By the end of the year, roughly 4% of California’s acreage had been burned, an area roughly the size of the state of Rhode Island. The fires also broke the weather forecasts.

When there are active wildfires, intuitively, there is a lot of smoke. Where that smoke lingers, it can block out sunlight, making it cooler on the ground than it otherwise would be. Most atmospheric simulation models used operationally to forecast the weather do not anticipate wildfire smoke or incorporate its effects, nor do the AI methods trained to correct and blend the model forecasts. So when AI tried to forecast the California weather after being trained on historical data that didn’t include instances of record-setting wildfires, it didn’t anticipate these temperature anomalies. The results were forecasts that skewed far too warm due to an omission that would have been obvious to anyone on the ground.

This episode highlights one of the main weaknesses of AI when it makes predictions on its own, without the additional insight of a human being. AI almost always makes the forecast better. But when it’s trained with historical information, AI is not so great at making sense of events that haven’t happened before, or have occurred only rarely.

When it comes to the weather, that’s an important limitation. Not only are extreme weather events getting increasingly common, but they’re having a growing impact. The year 2020 saw 22 weather disasters that caused more than $1B in losses each. The solution is to marry the speed and skill of AI with the wisdom of experienced human forecasters “over the loop.” In the case of last year’s wildfires, meteorologists at IBM’s The Weather Company recognized that our AI forecasts were too warm in smoky areas and jumped in to supply corrective nudges. The AI’s machine learning, which is constantly updating, began to correct for the new conditions as it learned from each new data point, then adjusted back again when the smoke cleared.

Incorporating the human perspective is a crucial part of both keeping automated forecasts in line with reality and tailoring them to the needs of individuals or businesses. When forecasts incorporate a human perspective, they can go beyond what the weather will be to what it will mean for people and their activities: for example, predicting flu transmissibility, pollen counts, the best time to go for a run or a safety threat that requires action. And humans are still unmatched in translating the forecast and communicating what people and businesses need to know, particularly when faced with an unprecedented weather event.

Harnessing AI’s strengths

One reason that AI is such a powerful tool is that AI and human beings actually have surprisingly complementary strengths.

While humans are better at subjective observations or perceiving meaningful anomalies, AI’s strengths in lightning-fast data processing, curation, and fusion are essential for taking the fire hose of weather information and making it comprehensible to humans in a timely way. Weather data comes from a wide variety of sources, from satellites and ground sensors to radars, cell phones and even connected vehicles, producing billions of data points every second from all around the world. Without AI, humans would find all that information overwhelming. For example, The Weather Company handles more than 400 terabytes of data daily. AI helps clean up data sets so that they are high quality and accessible for people to study and understand. It also combines disparate data sources and blends them together to make hyper-local weather predictions anywhere in the world in the blink of an eye.

Because it’s so good with numbers, AI is also an incredibly powerful tool for determining the range of possible weather outcomes and how likely different scenarios are. These weather probability predictions can have huge business implications. Take the example of a grower deciding whether to deploy blowers or mister spray to mitigate crop damage in the face of a possible freeze. Knowing the likelihood of freezing temperatures helps them weigh the cost of an intervention against the potential loss if no action is taken. AI can also help model the surviving crop value, predicting the expected cost and reward for each outcome and allowing the farmer to make a more optimal decision. Situations in which quantifying weather uncertainty helps people make better decisions abound in business, government and our personal lives.

Anticipating human needs

One of the main reasons people look to weather forecasts is to enhance their ability to plan, from the type of clothing they need to be comfortable on their morning commute to what time the weekend picnic should start. And here, as well, human perspective, along with AI, is important. AI may not be ready to help you plan a weekend by itself, but a meteorologist who understands what the weather means for activities in your location and season can interpret the forecast and provide just the guidance you need. Combining the human perspective with an AI-driven forecast of the weather, its uncertainty and its impact can automate this process, creating a personalized “decision service.”

At The Weather Company, we believe there is even greater potential for AI and humans to work together to create more accurate and relevant forecasts and use them to guide decisions. For instance, AI can combine the latest observations and atmospheric simulations to create precise forecasts for precipitation, down to the minute that the rain is likely to start in your neighborhood, for those who have opted to share that information. With human input, we can figure out who needs that kind of granularity, or what it may mean in terms of flooding hazard or evacuation routes. And with feedback on what people value, AI can learn to guide us to superior choices. By working together, humans and AI can enhance how we adapt to changes in the weather, a crucial step toward helping individuals, organizations and enterprises better prepare for the future.

IBM’s The Weather Company has been selected as part of a team led by the University of Oklahoma to establish a new U.S. National Science Foundation AI Institute for Research on Trustworthy AI in Weather, Climate and Coastal Oceanography. The team has been granted a $20 million, five-year award.

Dr. John K. Williams, senior manager, forecasting sciences at The Weather Company, helped write the winning application with lead researcher Dr. Amy McGovern and other private, public and academic groups. Here, he explains the importance of trustworthiness in artificial intelligence and how the institute is meant to help foster it.

What are some of the challenges the institute aims to address?

Big Data is one of the big challenges of our enterprise. In environmental science and weather and oceanography, we have immense amounts of data flowing in from satellite systems, ground observing stations, numerical weather models, even cell phones around the world. To take all that information and get value out of it is not something that human beings can do alone anymore. It’s something that requires artificial intelligence.

The Weather Company is a global leader in AI applied to weather at huge scale,  and we have been at it for 20 years. But as the use of AI explodes—we see it everywhere now, even making decisions about the ads and news we see—it’s really important to make sure that it’s unbiased and trustworthy in the information it’s providing, and that it really does what it says it’s going to do.

How does trustworthiness and bias apply to weather forecasting?

It could be that we haven’t thought enough about how different decision makers and cultural communities use forecast information or want weather information to be presented to them. That’s one reason I’m really excited that the institute will include social scientists, which is rare for this kind of research. Studying what information people need and how they use it will help us develop the best AI methods for providing it.

Another way that bias can creep into weather prediction is when you tend not to forecast extreme events because, mostly, there aren’t extreme events. So, if you train your forecast on the historical data that you have, your predictions tend to converge on what’s average. But you don’t want to bias your forecasts towards what’s usual. The risk is that you don’t capture those very impactful events that are unusual, which is what you really care most about.

What sort of work will the institute do?

It will be a combination of advancing AI methods for weather prediction and improving scientific understanding of the environment. But a large element will be driven by specific applications. For example, we’re providing IBM’s weather data to a group that is using AI to help predict when water temperatures in the Laguna Madre off Corpus Christi, Texas, might drop so that ships can avoid running over sea turtles stunned by the cold and volunteers can be mobilized to help rescue the turtles.

Extreme weather—things like tropical cyclones, tornadoes and winter storms—will be a focus, along with predicting seasonal variability. In many cases we already have numerical weather prediction models, including our own IBM GRAF, that provide forecasts. But they’re often not tuned to the variables that are most of interest to a user for making a decision, like a farmer or outdoor event planner who wants to know whether it’s going to hail. The model might not provide a very accurate hail forecast by itself, but with AI it can make a more skillful prediction. Developing better ways to communicate uncertainty and risks is also a key theme.

There are several academic, government and other industry partners involved. What is IBM’s role?

There’s a lot that we can learn from and share with each other—methods, intellectual property, code. Bringing in IBM helps ensure that the methods developed by the institute are addressing challenges that our customers and fans care about and will be conveyed to people who can benefit from them.

We’re planning an open source project that will be an essential part of what the institute does. That’s something obviously that’s in IBM’s DNA, especially now through Red Hat, and hybrid cloud could become an important element of our program to make the software truly portable.

There’s so much value in all the research being done at all these different institutions. If we can create an open collaboration and figure out a way to bring it all together and harness it, it will be a fantastic opportunity to cross the valley of death from research to operations and create value for society.

For more information: University of Oklahoma news release.

For Misha Sulpovar, an AI product leader at IBM’s The Weather Company, COVID-19 was personal.

His wife, Jiyoung, is Korean, and they have family in South Korea, as well as friends in China. And so the couple had closely followed the outbreak in those countries from the beginning. By early March, when it was evident the pandemic was becoming global, Sulpovar, who is based in Atlanta, had further cause for concern.

Last year, his team helped develop Flu Insights with Watson, a tool that uses The Weather Company data to provide a 15-day flu forecast. He became familiar with that ailment’s epidemiology. And so, when COVID-19 came along, Sulpovar realized the world was facing a much more serious threat: a highly contagious disease, with a relatively high mortality rate, that can spread before those who are infected develop symptoms.

Sulpovar knew that scientists, doctors and the public would need robust data tools to track the disease, if they hoped to slow its spread—tools that IBM and The Weather Company could create.

On March 13, as he and many other IBMers had begun working remotely at the company’s urging, Sulpovar went into action. He and a colleague, IBM Distinguished Engineer Bill Higgins, who was working from his home in Raleigh, N.C., started brainstorming online.

Within a few days they had virtually assembled an ad hoc team of developers, designers and other IBM experts, who started bouncing ideas off each other in a COVID-19 data channel they set up on Slack.

The Need for Trusted Data

At the same time, separate teams at The Weather Channel and IBM Cognos Analytics were already building proofs-of-concept for mapping the outbreak. They were hampered, though, by having access only to limited and incomplete data from various health, government and academic websites.

“We saw that there was no global repository for trusted COVID-19 data, at a time when everyone around the world was desperate for that information,” Sulpovar said. “Together, we could make a real contribution by building tools that automatically gather COVID-19 information from trusted sources—including national, county and local health departments—and displaying it in a consumable form.”

He and Higgins took their proposal to IBM’s chief data officer for cloud and cognitive software, Seth Dobrin, who was operating from home in Ridgefield, Conn. Dobrin brought it to IBM’s COVID-19 Task Force, which had been formed under Director of Research Dario Gil’s leadership to help fight the pandemic.

One of the task force’s goals is helping businesses, governments and citizens around the world make better decisions during the crisis by providing them with trusted data. Recognizing the value of the proposal Dobrin had brought forward, the task force immediately gave approval.

“Once Seth gave us the green light on behalf of Dario and the task force, we shifted into a higher gear,” Higgins said. “We created a Slack channel called #covid-data-war-room, started a Webex group, and simultaneously started to design and code what would become the initial release, while recruiting other IBMers with relevant skills to more quickly divide and conquer on the many tasks we’d need to complete. The race was on.”

It was now March 16, and the project was quickly building momentum. But success would depend on the team’s ability to represent complex data in simple, yet powerful, visuals.

Sheri Bachstein, vice president and head of The Weather Company, led the way for The Weather Channel digital team to create and host the consumer-facing applications. Sam Wong, head of business solutions for IBM Business Analytics, made the Cognos-based interactive dashboard he and his team had been designing available on the IBM Cloud.

Agile Collaboration

IBM in recent years has been a practitioner of Agile development—an approach to business methods that emphasizes collaboration among self-organizing, cross-functional teams. In this case, the agility culture once again proved its value. About 100 IBM coders quickly organized themselves into four groups.

A team from The Weather Company was responsible for design, user experience and building a scalable data applications programming interface, or API.

A Watson and Weather AI team took charge of collecting data and organizing it into usable form, and designing the data pipeline using Watson Natural Language Processing (NLP) technology.

A Cloud Platform team created the network architecture and designed the data repository on the IBM Cloud.

A team from the Chief Data Office handled data governance, to ensure that IBM had the permission to use the data in the tracking tools.

Collaborating via Slack, Webex and other digital team methods, the group created two complementary COVID-19 tracking tools in just five days—a volume of work that might normally require months to complete.

Two New Data Offerings, Free to the Public

One of the new offerings is called the Incidents Map for The Weather Channel app and weather.com website. Available for free, it provides COVID-19 data and statistics, including confirmed cases and more by U.S. state and by county, where available.

The Incidents Map also includes the latest COVID-19 news and video from The Weather Channel’s editorial team, as well as public health information and patient education materials. The team recently added international information and country-level data; it expects to provide local data for selected countries soon.

The team’s other innovation was an interactive dashboard, driven by IBM Watson AI capabilities and built on the IBM Cognos Analytics platform. Also available at no charge, the dashboard is designed to help users such as data scientists, researchers, media organizations and more conduct a deeper analysis and filtering of geographical data. Available global data includes confirmed cases and recoveries, where available. Users can drill down to the country, region, state and county level as needed.

Embraced by Tens of Millions

The COVID-19 hub of data and news on The Weather Channel app and weather.com has been visited more than 127 million times, with an average of nearly four million daily visitors. In addition, millions have visited the dashboard.

“The way the organization came together is a monument to IBM’s unique culture,” Sulpovar said. “IBMers who didn’t necessarily know each other teamed up in an atmosphere of complete trust to tackle a tough project—and succeeded.”

His partner in launching the effort, Higgins, also acknowledged the teamwork involved, under trying conditions.

“This experience has been especially inspirational because it demonstrates the unmatched value of one of IBM’s most cherished practices: Unite to get it done now,” Higgins said. “I’ve never been prouder to be an IBMer.”

New Modeling System GRAF Offers More Timely and Precise Local Forecasts, Democratizing Weather Around the Globe

ARMONK, N.Y. and ATLANTA, Nov. 14, 2019 — IBM (NYSE: IBM) and its subsidiary The Weather Company today announced the global rollout of a new supercomputer-driven weather forecasting system that will provide fresher, higher quality forecasts in parts of the world that have never before had access to state-of-the-art weather data.

Experience the interactive Multichannel News Release here: https://www.multivu.com/players/English/8247557-ibm-the-weather-company-graf-forecasting-system/

Known as GRAF, the Global High-Resolution Atmospheric Forecasting System can predict conditions up to 12 hours in advance with detail and frequency previously unavailable at this global scale.

GRAF will provide much finer-grained predictions of the atmosphere and update its forecasts six to 12 times more frequently than conventional global modeling systems. Current global weather models cover 10-15 square kilometers (6.2-9.3 miles) and are updated every 6-12 hours. By contrast, GRAF forecasts down to 3 kilometers (1.9 miles) and is updated hourly.

An August 2018 monsoon in India, shown at left by the best current weather model that operates at 13-kilometer resolution. At right, The Weather Company’s new Global High-Resolution Atmospheric Forecasting System (GRAF) operates at 3-km resolution and updates 6 to 12 times more often.

This level of forecasting precision has been available in the U.S., Japan and a handful of Western European countries. But the launch of GRAF marks the first time such enhanced forecasts cover more of the globe, including Asia, Africa and South America, areas among the most vulnerable to the increasingly intense extreme weather resulting from climate change. GRAF is the world’s first operational high-resolution, hourly-updating model that covers the entire globe.

“We view the launch of GRAF as a true inflection point in forecasting science, where technology helps democratize weather data for the good of society,” said Cameron Clayton, head of The Weather Company and general manager of IBM’s Watson Media and Weather. “The enhanced forecasts could be revolutionary for some areas of the world, such as for a rural farmer in India or Kenya. If you’ve never before had access to high-resolution weather data but could now anticipate thunderstorms before they approach your fields, you can better plan for planting or harvesting.”

Collaboration helps create improved global modeling of the atmosphere

To build the new modeling system, The Weather Company collaborated with the National Center for Atmospheric Research (NCAR) to create GRAF based on NCAR’s next-generation open-source global model, the Model for Prediction Across Scales, which uses state-of-the-art science to forecast the atmosphere down to thunderstorm level on a global scale.

As the world contends with climate change and more intense severe weather events globally, timely and accurate weather information is increasingly important. To help meet future challenges, strong public-private partnerships across governments, businesses and research institutions and open-source collaboration can continue to help advance science and technology at a more effective pace.

Advanced supercomputing allows for higher workloads

Forecasting weather is a complicated mathematical problem, requiring high-performance computing to solve complex equations. Traditionally, most weather models use high-performance computers built only with CPUs (central processing units). To handle its increased resolution and update frequency, the new GRAF system runs on an IBM POWER9-based supercomputer optimized for both CPUs and GPUs (graphics processing units), powerful compute engines widely deployed for demanding high-performance computing and AI applications.

IBM is applying the same technology behind some of the world’s most powerful supercomputers to weather forecasting. The Weather Company and IBM, together with NCAR, the University of Wyoming’s Department of Electrical and Computer Engineering, and others applied OpenACC directives to MPAS to take advantage of NVIDIA V100 Tensor Core GPUs on an IBM Power Systems AC922 server.

This is the world’s first global weather model to run operationally on a GPU-based high-performance computing architecture.

A better picture of weather globally

Other models may be high-resolution or update often, but the resulting forecasts only cover one country or region of the world. This is the first time a full global model exists to provide forecasts for the day ahead at this scale, resolution and frequency.

A clearer sense of exactly when and where impactful weather will surface can help when planning and preparing for weather. Whether it’s an airline, a utility company, a daily commuter, a retailer, a government decision-maker, or a farmer, GRAF predictions can help people, governments and businesses around the world make more informed weather-related decisions.

Leveraging GRAF to help in decision-making also requires technologies such as AI, cloud and analytics. Combining these additional technologies enables predictions from GRAF to help power IBM weather offerings for businesses and to drive weather content within apps and websites by The Weather Channel (weather.com) and Weather Underground (wunderground.com).

SOURCE IBM