Earth systems models are essential for examining the intricate processes happening on our planet, including interactions between the atmosphere and the biosphere. These models aid researchers and policymakers in grasping issues like climate change. However, to enhance accuracy in simulations, it is often necessary to collect vast amounts of data, sometimes requiring the painstaking effort of compiling millions of data points.
Researchers, such as UConn’s Assistant Professor James Knighton from the Department of Natural Resources and the Environment, Pablo Sanchez-Martinez from the University of Edinburgh, and Leander Anderegg from the University of California, Santa Barbara, have devised a technique to avoid the need for data collection across more than 55,000 tree species, allowing for a clearer understanding of how plants impact global water flow. Their research findings have been published in Nature Scientific Data.
Plants play crucial roles in Earth’s systems—they absorb carbon and supply oxygen for other organisms, including humans. According to Knighton, plants contribute to the movement of water, with an estimated 60% of precipitation returned to the atmosphere through a process called transpiration. This extensive movement of water through vegetation is intricate and is currently represented in Earth system models (ESMs) in a simplified manner, where all plants in a region are treated as a single entity known as a plant functional type (PFT).
“PFTs are used since our knowledge about individual plant species is limited,” explains Knighton, who is part of the College of Agriculture, Health, and Natural Sciences. “Creating a detailed vegetation map across a continent and assigning specific values for each species is too complex, so it’s more convenient to categorize them as a common PFT.”
The challenge with PFTs lies in the fact that different plant species exhibit unique hydrological traits—essentially how water is processed within plants. This oversimplification may hinder the models’ ability to accurately predict future conditions. Scientists have attempted to address this by developing databases like the TRY Plant Trait Database, which compiles trait information. However, Knighton highlights that only a small fraction, around 5,000 to 15,000 plant species, have had their traits comprehensively documented despite centuries of plant research.
“There are approximately 60,000 to 70,000 tree species globally, meaning we have knowledge on just around 5 to 10% of what’s out there after 200 years. If we keep this pace, it could take another 2,000 years to learn about all the necessary plants. By then, the impacts of climate change will already be upon us, and that is not tenable. We cannot just rely on field researchers to gather data at a slow rate; while field studies are valuable, they won’t yield results quickly enough,” he emphasizes.
In light of this issue, Knighton and his team set out to find a faster solution by analyzing existing trait data, such as tree height, root depth, and water flow rates within plants. They then conducted a phylogenetic analysis, assessing the traits preserved in the evolutionary history of related species.
“We examined the similarity of traits among closely related species, based on the idea that critical survival traits would have been maintained through evolution, rather than being randomly distributed,” Knighton explains. “For instance, if deep roots are essential for a particular plant’s survival, closely related species will likely share that characteristic, with similar root structures across their family or genus.”
The researchers applied this analysis across all traits and discovered a significant level of consistency throughout the phylogenetic tree, suggesting that closely related species share comparable trait values.
“Then we created a phylogeny mapping all plant species on Earth, detailing their relationships,” he states.
With this information on related species, Knighton explains they could estimate trait data without the need for extensive field measurements.
“Using various machine learning techniques, we constructed a comprehensive database containing critical trait values for 55,000 tree species worldwide,” he adds. “For global modeling that requires detailed vegetation data, this serves as an excellent starting point. There’s no longer a need to rely solely on the generic ‘one plant species per continent’ model; researchers can explore more intricate approaches to see how different species perform.”
Knighton acknowledges this endeavor is a preliminary step, but a significant first move. As additional data is gathered through field studies, it can refine and enhance the accuracy of their interpolated data.
This research is part of a larger project, beginning with a proof-of-concept study at a more local scale. That initial phase confirmed the viability of imputing hydrological traits, and Knighton indicates the next aim is to compare these imputed values with the data collected on-site at UConn Forest and other locations across the United States.
Knighton notes that there are ten sites throughout the country where extensive data is collected, which will be used as test cases. Master’s student Caroline Stanton ’26 is currently developing ecosystem models for each location, calibrating high-resolution models to estimate traits and comparing them to two decades of previously gathered data. They will analyze how the quality of their models is influenced by these estimated plant traits versus the observed data from each site.
Ultimately, the researchers aspire to extend their method to forested regions globally, studying the factors that drive variation in plant traits. Understanding these variations could enhance model accuracy while providing insights into the mechanisms behind different trait distributions.
Knighton expresses that he and his team hope that climate modelers find their work valuable, and they are eager to contribute to a deeper understanding of the Earth’s systems and the vital roles that plants play in them.
“Plants regulate our environment to an astounding extent,” he concludes.
