Juniper expansion across large land areas of the western United States has been well documented and mapped using coarse imagery from satellite and airborne remote sensing platforms. As regional land managers look to develop catchment-specific conservation goals to mitigate the future effects of climate-change and increase drought resiliency in headwater streams, a more fine-scale monitoring of juniper encroachment into riparian habitats is necessary. The emerging technology of unmanned aerial systems (UAS) capable of capturing very-high-resolution (VHR) imagery has the potential to provide scale-appropriate data to inform land managers of detailed landscape characteristics of riparian habitats. The aim of this study was to evaluate the effectiveness of UAS imagery and supervised image classification methods to map vegetation composition and estimate above-ground-biomass (AGB) of individual juniper trees along a low-order stream in Southwest Montana. Imagery from two sensors, a slightly modified off-the-shelf camera with traditional red-green-blue spectral information and a specialized multispectral camera equipped with a near-infrared sensor were compared.
Results/Conclusions
Detailed orthoimagery and elevation models were created from the 3-cm ground resolution RGB and the 11-cm ground resolution multispectral datasets for a 1.4km section of riparian habitat threatened by juniper expansion from the drier up-land landscape. Supervised image classification methods using the Random Forest algorithm have produced promising vegetation maps with overall accuracies of approximately 80%, compared with reference data, which is comparable to existing remote sensing products. Estimation of individual juniper tree parameters, such as tree height and canopy area are similarly promising, especially for those trees growing in sparse canopy cover. Discrimination between deciduous and evergreen cover in dense and mixed canopy of riparian areas directly adjacent to the stream has shown to be an area where further investigation into classification methods is needed. More research into best-practices of very-high-resolution image acquisition and classification are needed to advance UAS imagery to a point where it can replace more traditional and expensive remote sensing techniques, but the results of this study show that the technology is on its way to being a useful tool for conservation planners and ecologists.