Study: Automated extraction of leaf mass per area from digitized herbarium specimens (DOI: 10.1111/nph.70292)
Report: State of the World's Plant and Fungi 2026
Scientists who study plant physiology and evolution have a new tool in their toolkit: a machine learning algorithm that can scan digital plant specimen collections and quickly measure leaf size and thickness.
The algorithm, developed by William Weaver, a Schmidt AI in Science fellow at the University of Michigan, helped U-M researchers show that temperature appears to be one of the primary factors that contribute to leaf thickness. The algorithm will be incorporated into software developed by Weaver called LeafMachine2, which was designed to extract leaf traits from digital plant datasets such as those held at the U-M Herbarium.
The new algorithm measures a leaf's petiole, the part of the stem that attaches to the leaf.
"If you're sitting with a plant in person, you can measure the petiole with calipers. If you wanted to measure it manually in a photo, you could just calculate the distance between the pixels," Weaver said. "But we created an algorithm to automate that so we could look at some leaf mass per area trends at a global scale across a lot of different plants using herbarium specimens."
The study, appearing in New Phytologist, is part of the State of the World's Plants and Fungi report published by Royal Botanic Gardens, Kew. The 2026 edition of the report "reveals the state of global biodiversity and explores how digital tools are now transforming our ability to understand and respond to the climate and biodiversity crises, expose critical gaps in scientific knowledge, and highlight where action is most urgently needed to safeguard plants and fungi," according to the organization.
Leaf traits and climate
The size and thickness of a plant's leaf can tell you a lot about where it lives. Plants that live in more tropical climates tend to grow thicker leaves-leaves with higher mass per area-than plants that lose their leaves at the end of summer. That's because it makes more sense to invest more resources in growing a thicker leaf if you're going to hang onto it throughout the year than it might if your leaves are going to fall off at the end of the growing season, Weaver said.
To develop the algorithm, the research team, which includes U-M Herbarium associate chair Thais Vasconcelos and U-M researchers Aly Baumgartner, Zoë Bugnaski and James Boyko, examined leaf area and petiole width from more than 22,000 leaves, representing 1,580 species of "woody angiosperms"-flowering plants with bark such as vines, shrubs and trees. They then used those measurements to estimate predicted leaf mass per area, a proxy for construction investment.
Their analysis found that predicted leaf mass per area was more strongly associated with temperature-related variables and solar radiation than with precipitation-related variables at a global scale. The result is consistent with previous studies of leaf mass per area and climate, and supports the use of automated herbarium measurements for studying broad trait-environment patterns.
Plant leaves and paleoclimate
Paleobotanists use plants that are currently living to study plants that lived millions of years ago. It's difficult to deduce the thickness of plant leaves from fossil impressions. Likewise, researchers can't measure mass based on photographs of plants in an herbarium collection. Instead, proxy methods-in this case, based on certain traits of currently living plants-developed for fossil leaves can be instrumental for allowing researchers to compile and analyze large datasets of digitized specimen images.
"We study patterns of modern plant traits to develop proxies that can then be applied to fossil plants, which allows us to understand similar patterns that cannot be measured directly." said Baumgartner, an applied paleobotanist. "For example, the leaf petiole has to hold up the blade of the leaf, so the greater the mass of the leaf, the thicker the petiole needs to be to support it. It's simple physics, but it's a game changer if you're in a situation where you cannot measure the mass of the leaf."
Understanding how leaf thickness correlates with where extant plants are living can tell paleobotanists a lot about the paleoclimate in which plants lived millions of years ago, Weaver said.
"If you're able to get a really strong link between a certain leaf morphology and a certain present-day climate, then you could reason that a similar climate probably existed in the past when you're looking at a fossil assemblage in a past place," he said.
The algorithm and LeafMachine2 also show the evolving ways in which herbarium specimens can be used.
"The fact that we've collected plants in a systematic way for hundreds of years lets us learn a lot about our natural environment, about how the world is changing and how people have changed the world," Weaver said. "At the University of Michigan, we have specimens that go back well into the 1800s. We have a huge historical record that definitely gets overlooked and undervalued."
Baumgartner is the collection manager of vascular plants at the U-M Herbarium and the Department of Ecology and Evolutionary Biology. Bugnaski is a research assistant at the Chicago Botanic Gardens, and Boyko and Vasconcelos are assistant professors of ecology and evolutionary biology.
