Developing large-scale pasture approaches to quantify forage mass in rangelands using drones

Authors: PAGE, MICHAEL - Texas A&M University; PEROTTO-BALDIVIESO, HUMBERTO - Texas A&M University; ORTEGA-S, J. ALFONSO - Texas A&M University; TANNER, EVAN - Texas A&M University; Angerer, Jay; COMBS, RIDER - Texas A&M University; JOHNSTON, BRADLEY - Texas A&M University; RAMIREZ, MELAINE - Texas A&M University; CAMACHO, ANNALYSA - Texas A&M University; DIMAGGIO, ALEXANDRIA - Texas A&M University; DANIELS, WAYLON - Natural Resources Conservation Service (NRCS, USDA); KIMMET, ANTHONY - Natural Resources Conservation Service (NRCS, USDA)

Submitted to: Rangeland Ecology and Management
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/9/2025
Publication Date: 4/10/2025
Citation: Page, M.T., Perotto-Baldivieso, H.L., Ortega-S, J., Tanner, E.P., Angerer, J.P., Combs, R.C., Johnston, B.K., Ramirez, M., Camacho, A.M., Dimaggio, A.M., Daniels, W.D., Kimmet, A. 2025. Developing large-scale pasture approaches to quantify forage mass in rangelands using drones. Rangeland Ecology and Management. 100:111-120. https://doi.org/10.1016/j.rama.2025.03.005.
DOI: https://doi.org/10.1016/j.rama.2025.03.005

Interpretive Summary: The use of drones for monitoring and managing rangelands has increased in recent years. These instruments have cameras and sensors that provide opportunities to map pasture forage in pastures at high-resolution which can be helpful in estimating the amount of forage available for grazing. In this study, conventional methods for forage estimation (vegetation clipping and double sampling) were compared with drone-based estimation methods. In addition, these methods were compared using differing numbers of drone-derived samples, along with estimates of the time required to collect data among the different methods to compare time efficiency. Six methods were tested on rangelands in South Texas which included: 1) vegetation clipping; 2) double sampling (i.e., clipping and visual estimation); 3) drone estimates from 50-m above ground level (AGL) combined with double sampling; 4) drone estimates from 100 m AGL with double sampling; 5) drone estimates from 50-m AGL with vegetation clipping and 6) drone estimate from 100 AGL with vegetation clipping. Statistical analyses indicated that all six methods were similar in their ability to estimate forage biomass. Variability in the drone-based forage estimates at the pasture scale decreased with lower flight altitudes (which increased image resolution) and with increasing sample sizes. Vegetation clipping combined with drone sampling was found to be the most time efficient method for pasture biomass estimation. Results from this study indicate that drones can be used successfully and efficiently to quantify forage mass over larger areas and potentially less accessible terrain.

Technical Abstract: The use of unmanned aerial vehicles (hereafter ‘drones’) has exponentially increased in recent years for monitoring and managing rangelands. High-resolution cameras and improved sensors provide an opportunity to investigate pasture-scale sampling methodology as an operational approach to estimate forage mass on rangelands using 3D models derived from drones. Our objectives were 1) to compare double sampling and vegetation clipping methods with drone-based forage estimation methods, (2) to compare forage mass estimation between methods using different number of drone-derived samples, and (3) estimate time efficiency of each one of these methods. To accomplish this, we acquired drone imagery in a 1,060-ha pasture in the South Texas Plains ecoregion in June 2020. We compared six forage mass sampling approaches: double sampling (DS), vegetation clipping (VC), drone-double sampling at 50 m (drone-DS50) and 100 m (drone-DS100) AGL, and drone-vegetation clipping at 50 m (drone-VC50) and 100 m (drone-VC100) AGL. Regression analyses were used to evaluate relationships between drone derived vegetation volume and the forage mass derived from DS and VC. The six forage mass estimation values were not statistically different 6,624 ± 141 kg · ha-1 (DS), 7,628 ± 481 kg · ha-1 (VC), 8,411 ± 217 kg · ha-1 (drone-DS50), 8,008 ± 250 kg · ha-1 (drone-DS100), 7,601 ± 350 kg · ha-1 (drone-VC50), and 7,667 ± 288 kg · ha-1 (drone-VC100). We also compared three sampling sizes: ground-based quadrats (DS and VC), 700, and 3,500 random points. Drone-VC50 with 700 and 3,500 random points provided the smallest and least variable forage mass estimations at a large-pasture scale (5818 ± 78 kg · ha-1 and 5653 ± 34 kg · ha-1 respectively). Number of samples per hour increased from 22 to 52 with the DS methods and 1.2 to 38 with the VC methods. A combination of DS and VC with drone data collection may be a good approach for future drone forage estimation.