2018 ESA Annual Meeting (August 5 -- 10)

COS 68-9 - Using a drone and multispectral camera to obtain accurate and complete information on reclamation success: Well pad and habitat modification studies in western Colorado

Wednesday, August 8, 2018: 10:50 AM
339, New Orleans Ernest N. Morial Convention Center
Tamera J. Minnick1, Richard Alward1,2, Alicia M. Langton3, Grayson Koenemann1 and Danielle B. Johnston4, (1)Physical and Environmental Sciences, Colorado Mesa University, Grand Junction, CO, (2)Aridlands, LLC, Grand Junction, CO, (3)EcoloGIS Consulting, Grand Junction, CO, (4)Colorado Parks and Wildlife, Grand Junction, CO
Background/Question/Methods

Quantitative assessments of the success of ecological restoration can be difficult, labor intensive, and a seemingly costly use of limited funds. However, inadequate evaluation of success and failure inhibits the implementation of adaptive management principles. Current quantitative monitoring methods typically consist of sampling (i.e., quadrats, line-point-intercept) and drawing inferences about a larger project area. The application of a rapidly emerging technology, namely small unmanned aerial systems (sUAS or drones), provides the restoration ecologist the opportunity to overcome the above limitations. Paired with miniaturized multi-spectral cameras, a small UAS can collect high-resolution (<8cm/pixel) vegetation and soil data on 10s to 100s of hectares in less than an hour. We are addressing two key questions from a wildlife habitat modification project and a natural gas well pad reclamation study in pinyon-juniper/shrub ecosystems. First, we compare the accuracy of these high resolution data acquired via the drone and camera to line point intercept data of vegetation cover, cover by functional group and individual species cover. Second, we address the ability to scale to the entire site from each type of data collection method.

Results/Conclusions

From a pilot study conducted in 2016 on six of the 28 plots of the wildlife habitat modification project, we found that overall cover calculated using the Optimized Soil-Adjusted Vegetation Index (OSAVI) was significantly related to estimated cover using line point intercept (p < 0.001, R2 = 0.77). There was also a significant relationship of Purshia tridentata (bitterbrush) cover with the two methods (p < 0.001, R2 = 0.90). We will also present results of data collected in 2017 from all 28 plots. From the well pad study, we found significant relationships from the two collection methods of herbaceous cover (p < 0.001, R2 = 0.90), shrub cover (p < 0.001, R2 = 0.90), tree cover (p < 0.001, R2 = 0.90), and bareground (p < 0.001, R2 = 0.90). When comparing cover of specific tree and shrub species using supervised classification, R2 values were between 0.41 and 0.89. We expect that object based image analysis (OBIA) will perform even better. We also found that on-the-ground, line-point intercept measures did not accurately extrapolate to the entire well pad, particularly in reference sites or those with high tree cover. The ability to extrapolate to larger areas is a benefit for reclamation work that requires standards of vegetation cover to be met.