Tracking Climate Change Through AI

This project originated with Kropp, who specializes in the intersection of environmental studies and data science. For Koplik, it was the perfect opportunity to combine his burgeoning interest in environmental studies with his established computer science skills. Last spring, he took Kropp’s Environmental Spatial Analysis course, which introduced him to Geographic Information System (GIS) software. “That piqued my interest, because it requires an expensive license,” he said. “So it was really cool to have that opportunity.”

Through this course, Koplik got to work with both the software and Kropp, two connections that facilitated his participation in summer research. Much of Koplik’s effort thus far has been concentrated on training an algorithm to better identify different elements of satellite images — shrubs, water, trees, and tundra, for example. To do this, he first labels these elements by hand before feeding the marked images to the computer. For this “training” process, he said, “the algorithm takes each pixel in the image and uses the surrounding pixels, with the surrounding squares, to guess which category it fits into.” 

By constantly fine-tuning the program in this way, Kropp and Koplik will develop a tool capable of identifying change over time in satellite images. They are using two sets of images, one from the 1970s and the other more recent, to understand how climate change and rising temperatures have affected terrain and vegetation in the Arctic over the past 50 years. As the training process continues, the algorithm will become more and more accurate. “So we’re hoping that will result in a good data set,” Koplik said. 

“This is the perfect balance for me — a little less hardware coding and more cool results, more very fine coding, which I like.”

This summer’s focus will mostly be on the earlier set of images, given the time and effort needed to adequately develop and train the computer algorithm. “The culmination of the summer will hopefully be a paper,” he explained. “We’re going to submit an abstract in the beginning of August to get everything moving.”

Entering Hamilton, Koplik planned to study computer science. But Hamilton’s creation of the new data science major allowed him to pursue a concentration more aligned with his interests. “This is the perfect balance for me — a little less hardware coding and more cool results, more very fine coding, which I like,” he said.

The insight-oriented angle of data science fits productively with environmental studies, an area in which Koplik found himself taking more and more classes as time went on. “I’m passionate about climate change, and being able to research it in a way I’m good at is really cool,” he said.