DSI Researchers Win Microsoft AI for Earth Grant
Two researchers from the Data Science Institute (DSI) have been awarded a Microsoft AI for Earth grant that gives them access to the company’s cloud-computing platform, Azure, and AI tools. The two will use Microsoft’s tools to conduct ecological research in Puerto Rico, where they’re studying the effect of Hurricane Maria on El Yunque National Forest, the only tropical rainforest in the U.S. Forest System.
In September 2017, the hurricane left an unprecedented swath of destruction in the forest, felling thousands of trees. The researchers will call upon Microsoft's machine-learning tools to analyze data regarding how tree species in the forest fared during the hurricane – which species withstood the winds and which didn’t – and how the topography of the forest protected some trees from wind damage. The 28,000-acre forest is home to some 240 tree species and thousands of plants.
It’s virtually impossible to survey a forest of that size using ground-based observations and “the human eye,” said Maria Uriarte, a professor of Ecology, Evolution and Environmental Biology at Columbia, one of the two researchers on the team. The Microsoft tools will help them analyze data generated from advanced imaging and remote-sensing technologies to create a clear portrait of the forest.
“We can use Microsoft's tools to study the trees remotely, which makes our job immensely easier yet more effective,” said Uriarte, a DSI affiliate and member of the Earth Insitute. “There are large tree species and small ones, and the canopy of the bigger ones often covers the smaller ones, so we need 3-D images and aerial images. It’s a lot of data and thus a challenge that calls for a data-driven, AI-intensive approach.”
Uriarte has also partnered with NASA, whose satellites and fly-over planes will give them millions of images of the forest – data that have to be clustered, sorted and analyzed.
Tian Zheng, the other researcher on the team, has a graduate student already analyzing imagistic data from the forest. And as professor of statistics and associate director for education at DSI, Zheng is an expert in big data. The grant will give them cloud credits to access the many deep-learning and machine-learning tools on Azure, she said, which are time consuming to install and maintain on one’s own computer system, especially during the exploratory stage of a research project. The grant will thus save the two time, money and effort and allow them to design an efficient data-science workflow for their research project. The research addresses a growing public concern since, due to global warming, hurricanes will become more intense in the future, said Zheng. So it’s essential to understand how hurricanes affect forests, particularly the 3-D structure that closely captures biomass and carbon-storage capacity.
“We can try out different Microsoft tools and see which ones work best for us,” added Zheng. “We have data imagery from the forest that my student, Chengliang Tang, is working on. But it’s difficult to extract meaningful visual features from the images of forest canopies and to discern different tree species. That presents us with a unique AI computer-vision challenge, and the Microsoft grant will give us the tools to experiment with and develop an optimal data-analysis approach for this important ecological research.”
--By Robert Florida