Location: Range and Meadow Forage Management Research
Project Number: 2070-21500-001-018-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Jul 1, 2026
End Date: Oct 31, 2028
Objective:
This project will advance rangeland plant community monitoring by developing a deep learning pipeline leveraging drone imagery to produce vegetation maps across western U.S. rangelands. Current tools based on satellite imagery are not highly accurate at the meter to centimeter scales needed for fuels estimates, forage productivity, and invasive species detection relevant to rangeland management.
Approach:
Our project objectives are to 1) Develop an Artificial Intelligence model (convolutional neural network) trained to identify rangeland plant species from drone imagery; 2) Integrate the output into sophisticated statistical models (hierarchical Bayesian models) to account for classification uncertainty and provide reliable species abundance estimates; and 3) Apply statistical techniques to extend applicability across varying drone sensors and conditions, minimizing the need for extensive field data.
The resulting open-source platform will enable land managers, ranchers, and agencies to estimate numerous rangeland vegetation metrics at fine-scales in real time. Our pipeline will be validated within the context of a long-term grazing study, allowing us to quantify the effects of livestock grazing on plant communities.