As Greenland’s massive ice sheet melts, sea levels are expected to rise more than seven meters, potentially spelling catastrophe for coastal cities from New York to New Orleans.
This alarming process is already underway and currently it is responsible for approximately a third of current sea level rises. But we don’t know how fast it will progress, said Marco Tedesco, who is a polar climate scientist at the Lamont-Doherty Earth Observatory, part of the Columbia Climate School, and an affiliate of the Data Science Institute (DSI).
As their capstone projects, a group of third semester DSI students worked with Tedesco to uncover the answers locked up in Greenland’s frigid “cryosphere,” using drone imaging to analyze the changing reflectiveness of the ice sheet, and the varied climatic and organic processes that are causing the ice to melt faster. This work presents a new way to investigate important unknowns.
“DSI students are extremely skilled professionals looking for a problem to solve,” said Tedesco.
He set students the challenge of helping to understand how drone photography and Artificial Intelligence can efficiently hunt for features that are invisible on satellite images. These include small holes in the ice sheet filled with water and microorganisms, called cryoconites, which cause the ice to reflect less light and drastically speed up the rate of melting.
Devan Samant, a member of the Class of 2024, was part of the team of five who joined Tedesco in Fall 2023 to work with Tedesco on this research problem. (Full student list below.) Samant had already spent the summer after his second semester using machine learning to better understand how oceans absorb carbon. “But I had never dealt with research on the cryosphere,” he said, which includes all frozen areas of the planet, such as ice caps and glaciers, sea ice, and permafrost.
The capstone team worked on three sets of drone images that Tedesco had collected during fieldwork in Greenland to better understand how to identify cryoconites and other important features. The students brought their well-honed technical skills to the real-world challenge, where they learned about climate-specific indicators, including the geophysical, climatic, and biochemical drivers of ice melting. “They become experts in understanding what drives melting in Greenland,” Tedesco said, in order to learn how their skills can be applied.
Using Tedesco’s data to identify fast-melting features was only the start of the challenge. Students were asked to critically investigate the data in order to guide future image-collecting practices. They were trying to determine what image resolution Tedesco’s algorithms required to detect the key features that are his data points.
“If we fly a drone at 100 feet, or 300 feet, or 1,000 feet, individual pixels can represent areas of different sizes – two centimeters, one inch, three inches,” he said. “Higher resolution images provide more data but require more compute, so we can’t collect as many images. As we get higher resolutions, when do we stop? It is almost like we are refining a recipe – if artificial intelligence is the basic recipe, what are the optimal ingredients to make the recipe?”
Investigating Tedesco’s data, Samant and his team determined that even when the drone flew at the highest altitude, key features were still visible, meaning that researchers in future can capture more data in a single trip.
Samant is proud of his work, having contributed to research answering such an important question, but he is eager to understand how these methods can progress further.
“There’s a lot of future research to be done, combining this with on-the-ground measurements, potentially looking at upscaling or downscaling resolution through machine learning, and combining drone images with satellite data.”
Tedesco said that capstone projects give students the freedom to critically analyze how data is being acquired and used, an essential part of the skills required for long term success in data science. Rather than an education that simply trains students to become skilled technicians, students from DSI develop strong critical thinking skills alongside the technical knowledge, preparing them to confront complex challenges.
“DSI students approach problems with a big picture perspective, then dissect them into components, solving them one by one.”