Wyoming and California Research Team Maps Ever-Changing World of Tundra

By Jack Imel and Kimberly Mann Bruch

It’s easy to imagine the tundra as a barren and motionless landscape. After all, it is literally frozen. But those invested in these landscapes — Indigenous peoples, natural resource 

developers and researchers like Ivan Sudakow and Jian Gong — will quickly explain why this common perception is far from the truth. Tundra landscapes’ permafrost-based structure quickly translates changes in temperature to changes in geography. 

“Autumn in the Yamal Peninsula, Siberia can look like a Van Gogh landscape, a smattering of deep blue lakes contrasting against rich yellow grasses,” said Ivan Sudakow, a principal investigator at the Carl Sagan Center for Research, SETI Institute. “If you could take a birds-eye view of this landscape over the past century and condense that into just a few minutes, you’d see something like ‘Starry Night’ — swirls and ripples of geographic transformation, a canvas of change on a grand scale.” 

As the structures of permafrost in regions like the Yamal peninsula transform over time, so do the prospects of those living and working on the land. The best way to understand these changes is to see them, and that’s what Sudakow and Gong set out to do when they started their U.S. National Science Foundation (NSF) FARR-supported project, titled “Curating Multi-source Time Series Image Dataset for Tundra Lakes in the Siberian Arctic.” 

“The goal of our project was to integrate historical topographic maps with modern satellite observations to extend the timeframe of available data for evaluating changes in sensitive tundra environments like the Yamal,” said Gong, a research scientist in the School of Computing at the University of Wyoming. “In other words, that time-condensed, birds-eye view ‘Starry Night’ that Sudakow mentioned already exists, albeit in pieces. Our work was to put those pieces together.” 

While Gong makes it sound simple enough, the process presented certain challenges. For instance, Sudakow and Gong had to acquire and organize historical maps from a vast array of sources. During this effort the researchers established a metadata standard for cataloging and referencing the materials they compiled. That is, they created  a set of guidelines that provided a common method of describing geologic materials — historical maps and modern satellite imagery — to ensure that they can be easily traced and used by others in the future. Another challenge was extracting data from modern satellite sources. To do this, Sudakow and Gong developed automated workflows for image retrieval and preprocessing, and shared these workflows with the public via GitHub.

After compiling the maps and satellite images, the researchers tested tundra lake feature detection algorithms on certain regions, determining the detection accuracy of these algorithms against manual mapping and recent deep learning segmentation approaches. 

Gong and Sudakow presented their findings in a short talk at the 2024 FARR Workshop in Washington DC. Their processed, georeferenced historical map dataset will be archived at the NSF Arctic Data Center, ensuring open access and long-term preservation.