Extent of the global grassland biome at high resolution

Authors: MACDOUGALL, A - University Of Guelph; VANZANT, B - University Of Guelph; SEABLOOM, E - University Of Minnesota; BORER, E - University Of Minnesota; Fay, Philip; Flynn, Kyle; Rowley, David

Submitted to: Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/1/2025
Publication Date: 1/27/2026
Citation: Macdougall, A.S., Vanzant, B., Seabloom, E.W., Borer, E.T., Fay, P.A., Flynn, K.C., Rowley, D.W. 2026. Extent of the global grassland biome at high resolution. Science. https://doi.org/10.1038/s41559-025-02955-6.
DOI: https://doi.org/10.1038/s41559-025-02955-6

Interpretive Summary: Accurate estimates of the area of grasslands on Earth's landmasses are critical to accurate estimates of global biodiversity and the global carbon sink. Despite advances in remote sensing, estimates of the distribution of grassland varies by millions of square kilometers. Using fine-scale data from six continents and evaluations of remote-sensed imagery by regional grassland experts, we show sources of mapping error for global grasslands, estimate coverage, and discuss implications for C stocks based on current United Nations (UN) FAO estimates. We estimate grassland coverage at ~31 million km2, which deviates from many widely cited totals, which are often much larger and continue to serve as integral components of global models for terrestrial C. Grassland coverage errors derive from a range of sources centering on how grasslands are defined and detected and variously affect each of the high-resolution platforms that we examined. Our findings illustrate the benefits and shortcomings of using high-resolution remotely resolved products in the absence of corroborating field data, and illustrate the susceptibility of estimated and modelled measures to how grasslands are defined and mapped. Resolving these issues will increasingly be important for understanding grassland distribution and their role in providing key ecosystem services. This importance is heightened because grasslands are one of the most transformed and at-risk biomes on the planet because of increased crop production, urban expansion, desertification, but which also gain in extent from deforestation and land retirement.

Technical Abstract: Land cover data are commonly used to model the terrestrial sink for carbon (C), yet these data have wide margins of error that significantly alters estimates of global C storage. We demonstrate this data vulnerability in grasslands, whose vast extent and productivity make them a critical component C cycling but whose estimated distribution varies by millions of km2. Using fine-scale data at 10 m resolution from six continents, we show sources of mapping error for global grasslands, estimate coverage, and discuss implications for C stocks based on current UN FAO estimates. We found grassland mapping to be commonly affected by inconsistent definitions and widespread detection error, with true coverage most likely at ~23% (30.6 million km2) of the earth’s ice-free surface. This correction would alter UN projections of C sequestration to 1.41 Pg C based on an estimated sequestration rate of 0.47 tonnes ha-1 yr-1. These findings underscore the challenges of biome mapping for ecosystem accounting and policy, with solutions requiring the matching of remote and local data rather than just counting pixels.