New artificial intelligence-assisted technology is planned to track permafrost in Alaska, beyond

Tracking changes in permafrost can take years and sometimes decades, lags that cannot keep up with the transformations in the rapidly warming Arctic. Now scientists will be developing new technology to track those changes in real time, thanks to a project funded by Google.

Tracking changes in permafrost can take years and sometimes decades, lags that cannot keep up with the transformations in the rapidly warming Arctic.

Now scientists will be developing new technology to track those changes in real time, thanks to a project funded by Google.

The company has awarded a $5 million grant to the Massachusetts-based Woodwell Climate Research Center to create a system combining satellite data with artificial intelligence to spot the changes as they occur. The project is led by Anna Liljedahl, an Alaska-based Woodwell climate scientist.

There are compelling reasons to better track changes in permafrost. Its thaw is causing myriad effects around the North. That includes expensive damage to important infrastructure. In Alaska, costs of replacing thaw-damaged sections of roads, runways and railroad would mount up to $24.5 billion by midcentury under the current climate trajectory, according to a recent study by researchers at George Washington University.

An example of how artificial intelligence can help, Liljedahl said, is the difficult task of tracking the polygon ice wedges that cover much of the Arctic tundra. Scientists have mapped over 1 billion of them, but trying to monitor their changes would be overwhelming without help from artificial intelligence, she said.

“The datasets that we are creating are so large that we need help from AI in order to keep up with the data and milk the information that lies within the data. The datasets are too large for us humans to comprehend, even if all of us Arctic permafrost scientists across the countries would come together,” she said by email.

For example, permafrost scientists are interested in whether polygons are high-centeredor low-centered because those features indicate different ice-melt patterns. High-centered polygons are shaped when surrounding ice wedges melt, and low-centered polygons are shaped when ice in the center melts away,  according to the National Snow and Ice Data Center.

By using artificial intelligence, computers could identify and outline individual polygons, determine whether they are low- or high-centered, detect changes to them and, if there are changes, help determine the cause, such as unusual summer heat, deep winter snowpack, rainfall or wildfire, she said.

“If we did not use AI then we would be stuck doing what we are already doing: Looking at isolated locations across the Arctic and guesstimating that those few areas (0.001% or less of the Arctic) are representing the entire Arctic. Assumptions are so unnecessary when we can do better!” Liljedahl said by email.

Woodwell has numerous academic and scientific partners in the project. Alaska institutions that are part of the project are the University of Alaska Fairbanks and the Alaska Native Tribal Health Consortium. The project will collaborate with the Permafrost Pathways program led by Woodwell Climate and supported by the Arctic Initiative at Harvard University’s Kennedy School of Government and the Alaska Institute for Justice.

Alaska Beacon is part of States Newsroom, a network of news bureaus supported by grants and a coalition of donors as a 501c(3) public charity. Alaska Beacon maintains editorial independence. Contact Editor Andrew Kitchenman for questions: [email protected]. Follow Alaska Beacon on Facebook and Twitter.

You can read the original here.