Elizabeth Smith presenting her research.
Plant and Soil Sciences Ph.D. alumna, Elizabeth Smith studies soil respiration using computational tools and machine learning. Smith’s research could have far-reaching implications for climate science; identifying soil respiration patterns could provide a more accurate foundation for predicting how ecosystems will respond to climate change.

Uncovering Earth’s respiration

April 13, 2025 Written by Nya Wynn | Photo by Monica Moriak

Plant and Soil Sciences Ph.D. alumna Elizabeth Smith uses machine learning to investigate soil respiration and the carbon cycle

One of the world’s most critical processes is soil respiration. That’s when soil releases carbon dioxide into the atmosphere. While a Plant and Soil Sciences Ph.D. student at UD, Elizabeth Smith investigated soil respiration using computational tools and machine learning, with the goal of creating a greater understanding of the role soil plays in the global carbon cycle. 

“Soil respiration is kind of like the Earth breathing,” said Smith, who was also an NSF Graduate Research Fellow. “It’s how carbon moves from the soil into the atmosphere, and understanding that process helps us better predict changes in carbon levels over time.”

Smith employed machine learning techniques, a type of computer-based analysis, to study how soil releases carbon dioxide over time and across different locations. Instead of manually looking for patterns within large amounts of soil respiration data, Smith used these machine learning models to help make predictions, the uncertainty of these predictions and find patterns within the data. 

According to her thesis advisor at the time, Rodrigo Vargas, professor of ecosystem ecology and environmental change, these techniques are constantly changing how environmental scientists  look at and analyze data. He explained that new computational techniques are opening doors to understanding environmental phenomena that were previously unstudied.

“We live in a data-rich era of environmental science information,” Vargas said. “Elizabeth's work exemplifies how novel analytical techniques can be applied to test current paradigms and shed light on new ones.”

Her work spanned regions within the National Ecological Observatory Network (NEON) across the United States, using 20 years of data collected from 2000 to 2020 to create detailed maps of soil respiration trends.

“NEON divides the United States into 17 regions based on factors like temperature, precipitation and vegetation,” Smith said. “I analyzed these regions to identify patterns, but I didn’t find the increasing trends in soil respiration we expected to see.”

Instead of a consistent rise or fall in soil respiration over time or across regions, Smith found that in some areas, soil respiration levels remained the same. This could mean that the factors expected to drive change in these environments, such as temperature, moisture or land use, were not having a dramatic impact on soil respiration. While unexpected, this discovery was far from a dead end. 

Smith’s findings could have far-reaching implications for climate science. By identifying uncertainties in soil respiration estimates, her work could provide a more accurate foundation for predicting how ecosystems will respond to climate change. 

“Rates of bottom up soil respiration vary, complicating our understanding of its role in the carbon cycle,” Smith said. “Using these computational models, I can estimate a level of bottom up respiration for a region and the level of uncertainty around that estimate. Regional models such as the one in this project can help constrain uncertainty in larger, global models.”

Since defending her dissertation in July 2024, Smith is now working as a postdoctoral researcher at Georgia Tech in the School of Earth and Atmospheric Sciences.

“My work at GT is similar to my work at UD in that it is also computational, but different in that it’s on a smaller scale,” Smith said. “I’ll be managing the data for a joint Georgia Tech and Yale University USDA funded project to research and scale up enhanced rock weathering (ERW) in agricultural settings.” 

ERW is a carbon-capturing technique that can be used to help address climate change by taking carbon out of the air and storing it in rocks. With her work at UD, Smith was looking at how the carbon budget, a measure of how much carbon dioxide is released into the atmosphere, could be reduced with further understanding of soil respiration. Both fields of study are critical for addressing the growing problems that accompany climate change. 

Smith is also passionate about science communication and education. Through her partnership with Four Youth, a local nonprofit organization in DE, she worked with local schools to teach environmental science to students. 

Smith was also recently selected as a Grassroots Science and Policy Intern for the American Geophysical Union, where she will help connect science with policy by advocating for science forward legislation.

“A lot of scientific research is full of jargon and big words, which makes it hard for the public to connect with,” Smith said. “I want to bridge that gap by presenting research in a way that’s not only understandable, but also exciting.”


Related News

  • In our backyard

    August 07, 2026 | Written by Molly Schafer | Photos courtesy of Madeline Brown, Alex McBreen, Catherine Morrissey, Scarlet Shifflett, J.D. Zimmerman and the Department of Natural Resources and Environmental Control Division of Parks and Recreation
    This year marks the 75th anniversary of Delaware State Parks, and for faculty and students at the University of Delaware, there is much to celebrate. For Delawareans, the state parks are a source of peace, but at UD, they’re also a source of purpose. The parks provide a living laboratory for research, opportunities for experiential learning and a space for community service.
  • Global crops and drought

    July 30, 2026 | Written by Adam Thomas | Illustration by Jeffrey C. Chase
    A new study from the University of Delaware’s Kyle Davis examined production losses for 17 major crops accounting for three-quarters of the world’s food production and found that portions of the Midwestern United States, eastern Brazil, eastern Spain, and central and northern India were among the areas most sensitive to droughts.
  • AI flower power

    July 27, 2026 | Written by Katie Peikes | Photo illustration by Jeffrey C. Chase
    Yin Bao, University of Delaware assistant professor of digital agriculture, built a specialized AI tool called FLORIST to automatically analyze thousands of plant photos taken by everyday people, instead of manually sorting through 44,000 images. FLORIST efficiently filtered out the vast majority of photos that did not show active flowering, narrowing the dataset. Bao says the main takeaway is blending widespread public observations with deep genetic research gives a more complete understanding of how species evolve to survive in the wild.
View all news

Events