Riparian habitats along rivers of the southwestern United States are threatened with invasions by non-native trees, including Tamarix spp., Elaeagnus angustifolia, Ailanthus altissima, and Ulmus pumila. Removal of invasive species frequently is followed by new invasions by herbaceous non-natives including Cirsium arvense, Acroptilon repens, Cardaria draba, and others instead of the desired Salix and Populus species. Timely monitoring of riparian restoration progress can be challenging due to distance to sites, restricted site access, lack of trained personnel, scarce economic resources, as well as the widespread nature of the threats. Remote sensing may be used to address these challenges, but freely available Landsat aerial imagery may be too coarse to be useful for evaluating the early stages of recovery or re-invasion. We are using a very high resolution multi-spectral camera (<8 cm GSD) mounted on an unmanned aerial system (UAS, aka “drone”) platform to assess riparian restoration at a nearly 50 ha site along the Colorado River near Grand Junction, Colorado. Prior to removal efforts, tamarisk canopy cover ranged from 20% to nearly 95% in riparian and wetland habitats.
Results/Conclusions
Tamarisk was cleared in 2015 and Salix, Populus, and Acer spp. poles were planted in 2016. On 16 July 2017, at approximately “peak-green,” we collected five-band multi-spectral aerial imagery of the entire site requiring just over 30 minutes of flight time. We processed the images into a 5-layer georectified orthomosaic in preparation for analysis and vegetation classification. Due to high correlations in reflectance among bands, we extracted three principal components that accounted for 99.6% of the total variation; this reduced data set was used to conduct a supervised classification. On-the-ground species identification and locations were collected on 19 July 2017 to provide training data for the classification. We found resprouting Tamarix and other undesirable species on 12.5 ha (25.7% cover) while desirable species, including Salix exigua and Populus fremontii, Distichlis spicata, and others totaled 24.8 ha (51.0%). Bare soil and gravel, woody debris, dead vegetation, and open water comprised the remaining 11.4 ha (23.3%). Adopting this new technology will help to ensure better restoration outcomes through improved capabilities of cost-effectively collecting detailed data over large areas, at ecologically significant times. Furthermore, multi-spectral data may be used to address questions about plant physiology, horizontal and vertical structure, and change detection.