UB students mix fieldwork, AI to map and understand Greenland glacial ice

UB students are mixing fieldwork with the latest developments in AI to map and understand the ever-changing glacial ice in Greenland.

This work is taking place in the Glacier Modeling Lab in the Department of Geology, led by Kristin Poinar, assistant professor, glaciologist and student mentor.

“Data from NASA satellite missions has turbo-charged the amount and rate we can learn about Earth’s ice sheets,” Poinar explains. “We’re harnessing AI and UB’s supercomputers to do this even faster. At the same time, not everything is observable from orbit, so we complement this with a boots-on-the-ice approach.”

Over the past two years, PhD student Naureen Khan has worked to finetune a deep learning model to detect crevasse fields in the Pâkitsoq region of western Greenland. Previously, researchers looked at satellite imagery and determined the crevasses by eye, based on years of training and experience. Khan’s approach substantially improves the speed and precision of the procedure.

According to Khan, understanding crevasses is critical, as they are often the highway system moving meltwater to the bed ice sheet, which can cause more glacier sliding and ultimately increased sea levels.

“With the technology advancement in machine learning, especially in deep learning and AI,” Khan says, “we can now use that computational resource to detect complicated crevasse fields and streams from the ice region, which was not possible before.”

Khan presented her research at the American Geophysical Union’s fall meeting last December in San Francisco. While researchers at other universities are working on detecting Antarctic crevasses, which tend to be wider and clearer, Khan’s work in Greenland is unique and advanced.

The AI approach is complementing fieldwork, which has been perennially difficult to perform in the extreme conditions found on the Greenland Ice Sheet.

“Not only is [climate change] affecting the ice in general, but it’s also making conducting fieldwork and leaving instruments out a lot more difficult,” says postdoctoral researcher Jessica Mejía, who traveled to Greenland twice in 2023 to coordinate deployment of GPS stations on the ice. The devices, set up near crevasses and left out in the elements for months at a time, are measuring how quickly the crevasses are growing.

Mejía says extreme winds, as well as unexpected ice melt, delayed some of their work last summer. But with determination and flexibility, the team was still able to collect data. Mejía has conducted an initial review of the data and says they captured a new crevasse opening where they didn’t expect.

PhD candidate Courtney Shafer is also working in the field, studying how water is being transported inside the ice sheet.

“Understanding where these cracks are forming and allowing water to infiltrate the ice sheet allows you to understand where the ice sheet is experiencing these speedups,” Shafer says.

Her fieldwork is focused on a firn aquifer — liquid water inside the layer of old snow near the surface of the ice sheet. The liquid water can persist, even when air temperatures are below freezing, because the firn acts as an igloo, insulating it from the cold. She used seismoelectrical technology to achieve her research goal.

The seismoelectric method, an underused technology in earth science, uses an explosion or a sledgehammer impact to shake the ground, Poinar explains. As the compressional waves travel downward through the glacier, they rapidly vibrate the ice and any water it contains. This generates a small electric field, which the lab scientists measure with large copper electrodes placed at the glacier surface. They use the travel time of the downward waves to figure out the subsurface electric properties at hundreds of different levels below.

“We are using the seismoelectric to locate both the top and the bottom of the aquifer,” Shafer says. “Previous radar measurements are limited by the strong reflectivity of the water table, so they can only locate the top surface.”

The students’ research advances are in no small part supported by the culture of the lab.

Poinar, who joined the UB faculty in 2018, has worked hard to create a lab culture that she says values equal-access principles — openly sharing research information and experiences, with everyone invested in the research.

“Kristin might be the best mentor I’ve had,” Khan notes. “She understands the level of effort I’m putting in and is always there to push and support me.”

Khan says she has benefited from Poinar’s approach, as they worked together to problem-solve while building the AI modeling program. “Talking to her about my work helps me think as well,” she says.

For Mejía, gaining experience in the field was an important factor when she was looking for a postdoctoral position. She says she has really enjoyed working and learning in Poinar’s lab.

“She’s given me a lot of leadership opportunities in the field, which was really exciting to me,” Mejia says. “My opinions are taken and discussed, and I have been contributing and shaping the direction of the fieldwork. To be trusted in that way, it’s been really great.”

Likewise, Shafer says she’s also had many new experiences since joining the lab. Her research shifted to Greenland and she has also gained fieldwork experience.

“Having Kristin as an adviser, who is very passionate, it seeps into your own work,” Shafer says. “You get excited about it because that kind of strong mentoring is there.”