Estimating forage utilization with drone-based photogrammetric point clouds

Authors: GILLAN, J.K. - University Of Arizona; MCCLARAN, M.P. - University Of Arizona; SWETNAM, T.L. - University Of Arizona; Heilman, Philip - Phil

Submitted to: Rangeland Ecology and Management
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/4/2018
Publication Date: 2/27/2019
Citation: Gillan, J., McClaran, M., Swetnam, T., Heilman, P. 2019. Estimating forage utilization with drone-based photogrammetric point clouds. Rangeland Ecology and Management. 72(4):575-585. https://doi.org/10.1016/j.rama.2019.02.009.
DOI: https://doi.org/10.1016/j.rama.2019.02.009

Interpretive Summary: Forage biomass and utilization are important indicators for evaluating livestock and range management in dryland ecosystems. Traditional field methods are typically obtained from few locations within a management unit because of large investment in travel and field time. This small spatial coverage and few samples can limit the accuracy of representing these indicators in a large management unit. To address this challenge of efficiently covering large areas without diminishing the quality of information, we deployed a small unmanned aerial system equipped with high resolution true color camera, operated with autonomous mission planning, and processed the data with advanced image analysis software capable of estimating indicator values. Our work occurred at the Santa Rita Experimental Range, a desert grassland savanna in southern Arizona. Prior to and immediately after a month-long grazing rotation of 80 head of cattle in an 150 ha pasture, we acquired very high-resolution RGB imagery (~ 1 cm ground sample distance). We used structure-from-motion photogrammetry methods to create 3D point clouds, digital surface models (DSMs), digital terrain models, and orthomosaics. Utilization was estimated by differencing the pre and post grazing DSMs. Imagery-based indicator values were compared to field methods known as comparative yield (biomass) and ungrazed plant method (utilization) for transects, plots, and the entire pasture. The high-resolution 3D models represented approximately 40% of the grass heights which contains 80% of the biomass. Preliminary results show 1) realistic estimates of biomass based on image-based extrapolations of volume based on field estimates and 2) estimates of utilization that are closely related to field methods at both the transect and plot scales. Our preliminary results are consistent with the promise of more efficiently obtaining high-resolution information for larger area than traditional field methods that under-sample the large of extent of rangelands.

Technical Abstract: Forage biomass and utilization are important indicators for evaluating livestock and range management in dryland ecosystems. Traditional field methods are typically obtained from few locations within a management unit because of large investment in travel and field time. This small spatial coverage and few samples can limit the accuracy of representing these indicators in a large management unit. To address this challenge of efficiently covering large areas without diminishing the quality of information, we deployed a small unmanned aerial system equipped with high resolution true color camera, operated with autonomous mission planning, and processed the data with advanced image analysis software capable of estimating indicator values. Our work occurred at the Santa Rita Experimental Range, a desert grassland savanna in southern Arizona. Prior to and immediately after a month-long grazing rotation of 80 head of cattle in an 150 ha pasture, we acquired very high-resolution RGB imagery (~ 1 cm ground sample distance). We used structure-from-motion photogrammetry methods to create 3D point clouds, digital surface models (DSMs), digital terrain models, and orthomosaics. Utilization was estimated by differencing the pre and post grazing DSMs. Imagery-based indicator values were compared to field methods known as comparative yield (biomass) and ungrazed plant method (utilization) for transects, plots, and the entire pasture. The high-resolution 3D models represented approximately 40% of the grass heights which contains 80% of the biomass. Preliminary results show 1) realistic estimates of biomass based on image-based extrapolations of volume based on field estimates and 2) estimates of utilization that are closely related to field methods at both the transect and plot scales. Our preliminary results are consistent with the promise of more efficiently obtaining high-resolution information for larger area than traditional field methods that under-sample the large of extent of rangelands.