Satellite imagery and remote sensing technologies are being used to monitor the health and condition of vegetation across ecosystems. Data collected across multiple spectral bands by the long-running Landsat program allows researchers to calculate the Normalized Difference Vegetation Index (NDVI), which is used to track changes in plant health and land cover over time.
Two recent efforts highlight the versatility of this approach: one tracks the spread of exotic annual grasses across western U.S. rangelands, and the other applies machine learning to NDVI to detect early signs of stress in coastal marshes.
Tracking the spread of invasive grasses in the Sagebrush Biome
The Sagebrush Biome of the western United States is under increasing pressure from invasive annual grasses. Invasive grasses like cheatgrass (Bromus tectorum) spread rapidly, crowd out native plants, and significantly increase wildfire risk.
To monitor this growing threat, researchers have developed a weekly data product that estimates fractional cover of exotic annual grass (EAG) species using Landsat imagery. Fractional cover refers to the proportion of ground surface covered by a particular type of species or land cover within a given area, usually expressed as a percentage or a decimal between 0 and 1.
This dataset provides weekly maps from mid-April through late June, capturing near real-time conditions with a lag of 7–13 days after satellite acquisition. The analysis relies on NDVI and other spectral indices calculated from harmonized Landsat and Sentinel-2 (HLS) imagery. 16 exotic grass species and one native perennial grass species are tracked each week. This dataset helps land managers identify areas at high risk of degradation or fire and prioritize where to focus mitigation or restoration efforts.
The data:
Dahal, D., Boyte, S.P., Megard, L., Postma, K., and Pastick, N.J., 2025, Early Estimates of Exotic Annual Grass (EAG) in the Sagebrush Biome, USA, 2025 (ver. 10.0, June 2025): U.S. Geological Survey data release, /10.5066/P14VQEGO.
Early warnings of marsh decline through belowground modeling
Scientists are also using Landsat and NDVI data, but to detect much subtler changes in the coastal salt marshes in Georgia. A study published in Proceedings of the National Academy of Sciences describes the development of the Belowground Ecosystem Resilience Model (BERM). This machine learning model integrates Landsat-derived vegetation indices with environmental data to forecast declines in belowground biomass. The data is used to analyzed the health of the root systems that hold marsh soil together.
What makes this work significant is that many marshes appear healthy from above, even as their belowground structure deteriorates. By comparing NDVI and other spectral data over time, the model can predict where subsurface stress is likely occurring. These predictions were validated through field sampling of root biomass and hyperspectral measurements.
This approach provides an early warning system for marsh ecosystems that might otherwise go unnoticed until collapse is already underway.
The study:
Runion, K. D., Alber, M., Mishra, D. R., Lever, M. A., Hladik, C. M., & O’Connell, J. L. (2025). Early warning signs of salt marsh drowning indicated by widespread vulnerability from declining belowground plant biomass. Proceedings of the National Academy of Sciences, 122(26), e2425501122. DOI: /10.1073/pnas.242550112
Additional reference
Roche, M. D., Crist, M. R., Aldridge, C. L., Sofaer, H. R., Jarnevich, C. S., & Heinrichs, J. A. (2024). Rates of change in invasive annual grass cover to inform management actions in sagebrush ecosystems. Rangelands, 46(6), 183-194. DOI: 10.1016/j.rala.2024.10.001