Can AI Forecasting Help the Great Lakes? A University of Michigan Institute Hopes So

It’s difficult to encapsulate the magnitude of the Laurentian Great Lakes. As the largest lake system on Earth, it holds 84% of North America’s fresh water and creates distinct local weather patterns.

The five “inland freshwater seas” supply drinking water to 40 million people and support millions of jobs. It’s unsurprising, therefore, that dozens of organizations exist to protect, monitor, and study these ecosystems. 

One of these is the Cooperative Institute for Great Lakes Research (CIGLR), a University of Michigan-based partnership with NOAA that works “to achieve environmental, economic, and social sustainability in the Great Lakes.”

Considering the scale of the Great Lakes and the number of parameters CIGLR forecasts, the datasets are extremely complicated. But they’re also crucial, and CIGLR needs people to understand them. 

To bridge this gap and aid researchers using this data, the institute is turning toward a powerful but often controversial tool: artificial intelligence.

Difficulties with Great Lakes Forecasting

CIGLR operates in several research disciplines and industries, but one of its most prominent products is Great Lakes Forecasting. From ice cover to weather systems, the institute produces, trains, and distributes forecasting tools to scientists, resource managers, and the public.

“We collect and have access to vast amounts of Great Lakes data from various observing systems, including satellites, buoys, and monitoring programs,” says Mary Ogdahl, Managing Director of CIGLR.

Associate Research Scientist and Associate Director Ayumi Fujisaki-Manome is a principal investigator in their forecasting models. She and colleagues recently held a “Better Great Lakes Forecasts” workshop to tackle the dilemmas associated with such complex forecasting tools. 

“It’s a very popular product, but a very complex product, and not many people have enough preparation for that,” Monaome says. “So this workshop itself was initially trying to fill that gap.”

She continues, “And then one thing we realized is that AI will be a very powerful partner.”

So, the collaborative workshop allowed UM students and professors, AI experts, and NOAA researchers to work together on understanding CIGLR’s forecasting models. Manome says the students learned about running the models, data cleansing, and checking forecasts against real-world measurements.

But even with a deeper understanding, the forecasting models run into some unavoidable difficulties. The sheer size and complexity require forecasters to dutifully “clean” the data before it can be used, leaving less time for real-world application.

Yet, with the explosion of artificial intelligence and help from the Michigan Institute for Data and AI in Society, CIGLR hopes to change that.

CIGLR engineer Russ Miller and retired research scientist Tom Johengen service a western Lake Erie real-time NexSens water quality monitoring buoy that provides critical observations for CIGLR and GLERL research projects, including harmful algal bloom studies. (Credit: Cooperative Institute for Great Lakes Research)

Leveraging AI Forecasting in Environmental Datasets

“Sometimes people spend too much time on data processing, data cleaning, and all that, and have little time to actually interpret the results and think about the implications,” Manome explains. “So, I think with AI, you could have more time for that, which is a positive thing.”

She says that CIGLR hopes AI can help students and researchers alike use the forecast more efficiently. By simplifying the large, complex model, Manome says students may feel better about approaching it. That’s where the workshop was beneficial, as it helped future Great Lakes researchers encounter the forecasting tool in a collaborative learning environment. 

“We hope that the participants will be able to understand the data structure and be able to manipulate the data set and analyze it, but also be able to use AI responsibly and effectively,” Manome says.

She acknowledges that AI can be used as a “crutch” rather than simply a piece of the puzzle. The Great Lakes region is also growing as a data center hub, and concerns arise from their immense water usage. 

This is not lost on Manome and her colleagues. For an institute responsible for sustainably researching and managing the Great Lakes, they must be thoughtful with machine learning forecasting. 

Ogdahl also acknowledged that AI can still make mistakes, and that CIGLR is in the early stages of applying it to research. Therefore, students learning the forecasting tools also worked directly with AI experts.

“That is why we believe strongly in close oversight by expert scientists when using AI,” says Ogdahl. “Human knowledge, scrutiny, and judgement are critical along every step when using AI in research.”

CIGLR and partners gather at the Great Lakes Environmental Data Training Workshop on June 5, 2026. The workshop highlighted how environmental data, from forecasts to field observations, supports a better understanding and management of the Great Lakes. (Credit: Cooperative Institute for Great Lakes Research)

Training Future Scientists in AI Prediction and Computing

Ultimately, the goal is not for AI to forecast everything. It’s for AI to create a more efficient process, so that researchers spend more time unraveling what these forecasts mean. 

Whether that’s helping emergency managers understand storms or sharing environmental forecasts with local resource managers, Manome says they need students who understand these complex models.

She explains that many students conducting research in CIGLR are interested in environmental science. While this natural science background is useful, they might lack the technical understanding needed to tackle the huge dataset.

As the climate becomes more unpredictable and technology continues to evolve, Manome believes environmental science students will need to embrace this side of the field. 

“It will be a huge strength to the Great Lakes science community at large if we can somehow catalyze training at the earlier stage of their careers,” she explains.

CIGLR Associate Research Scientist and Associate Director Ayumi Fujisaki-Manome presents at the Great Lakes Environmental Data Training Workshop on June 5, 2026. The workshop highlighted how environmental data, from forecasts to field observations, helps researchers and managers better understand and manage the Great Lakes. (Credit: Cooperative Institute for Great Lakes Research)

Collaborations and Benefits of Great Lakes Forecasting

Monitoring such a large system is both incredibly difficult and rewarding. The difficulties arise from a complex ecosystem, large dataset, and training new students. But when applied correctly, the hydrometeorological and ecosystem forecasting can be incredibly beneficial.

For Ogdahl, that’s why CIGLR is such a collaborative institution. When protecting the Great Lakes requires both environmental science students and AI experts working together, collaboration isn’t just helpful; it’s crucial.

“No single organization or institution holds all of the data or expertise to understand and use them,” she says. “[…] The Great Lakes are so vast, so complex, and cover such a large geographic area that these collaborations are critical for understanding and protecting this precious resource.”

CIGLR engineer Russ Miller deploys an autonomous glider in Saginaw Bay, Lake Huron. The glider collects detailed observations of water conditions, helping researchers better understand lake processes and monitor changes throughout the water column. (Credit: Cooperative Institute for Great Lakes Research)