Open AccessImproving the efficiency and accuracy of evaluating aridland riparian habitat restoration using unmanned aerial vehicles
Author(s) -
GómezSapiens Martha,
Schlatter Karen J.,
Meléndez Ángela,
HernándezLópez Deus,
Salazar Helen,
Kendy Eloise,
Flessa Karl W.
Publication year - 2021
Publication title -
remote sensing in ecology and conservation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.191
H-Index - 21
ISSN - 2056-3485
DOI - 10.1002/rse2.204
Subject(s) - normalized difference vegetation index , riparian zone , vegetation (pathology) , environmental science , remote sensing , multispectral image , satellite imagery , vegetation classification , habitat , ecology , geography , leaf area index , biology , medicine , pathology
Abstract Unmanned Aerial Vehicles (UAVs) offer new opportunities for accurate, repeatable vegetation assessments, which are needed to adaptively manage restored habitat. We used UAVs, ground surveys, and satellite imagery to evaluate vegetation metrics for three riparian restoration sites along the Colorado River in Mexico and we compared the data accuracy and efficiency (cost and time requirements) between the three methods. We used an off‐the‐shelf UAV coupled with a multispectral sensor to determine Normalized Difference Vegetation Index (NDVI) and vegetation cover. We were unable to accurately classify vegetation by individual species, but by grouping riparian species of interest (cottonwood‐willow, mesquite, shrubs), we achieved high overall model accuracies of 87–96% across sites (Kappa = 0.82–0.95). Producer’s and user’s accuracies were moderate to high for target vegetation classes (69–100%). UAV and ground‐survey vegetation percent cover differed due to differences in methodologies (UAVs measure aerial cover; ground surveys measure foliar cover) and sources of error for each method. Correlations between UAV and ground survey vegetation cover were moderate (rs(90) = 0.24–0.58, p < 0.05). UAV NDVI (0.50–0.61) was significantly higher than Landsat NDVI (0.40–0.45) for all sites (p < 0.0001), likely due to presence of shadows with high NDVI values in UAV imagery. UAV NDVI, Landsat NDVI and UAV total vegetation cover were strongly correlated (rs(90) = 0.72–0.85, p < 0.05). UAV surveys were more labor‐ and cost‐ intensive than ground surveys in the first year, but were slightly less so in the second year. We conclude that UAVs can provide efficient, accurate assessments of riparian vegetation, which can be used in restoration site management. Due to UAV limitations to assess vegetation in a multi‐layered canopy and inability to classify individual riparian species with similar spectral signals, we recommend a combined approach of UAV and ground surveys.