Drylands cover 41% of the global land surface and provide ecosystem services to 38% of
the world’s population. Dryland ecosystems have already been degraded or threatened by the
increased rates of wildfire and invasive annual grasses, as well as changes in precipitation
patterns. We cannot protect, mitigate, or restore drylands without comprehensive vegetation
surveys. To understand ecosystem processes, we need to know the composition of vegetation at
the patch and plant scales. Field observations are limited in coverage, and are expensive and
time intensive. Data from Unmanned Aircraft Systems (UAS) will fill the niche between field
data and medium scale remotely sensed data, and support the potential for upscaling. UAS-based
remote sensing will also help extend the spatiotemporal scope of field surveys, improving
efficiency and effectiveness. This study aims to test UAS methods to estimate two important
vegetation metrics (1) fractional photosynthetic cover and (2) fractional cover of plant
functional types.
For both objectives, a series of surveys were conducted using fine-scale spatial resolution
(1-4 cm pixel-1) multispectral UAS data collected in Reynolds Creek Experimental
Watershed in Southwestern Idaho, USA. Data were collected at three sites along an elevation
and precipitation gradient. Each site is characterized by a different type of sagebrush:
Wyoming Big Sage, Low sage, and Mountain big sage. The first study in this thesis tests
multiple vegetation indices at each site to assess their accuracy in modeling photosynthetic
cover. We found the Modified Soil Adjusted Vegetation index (MSAVI) had the highest accuracy
when modeling photosynthetic cover at each site (62-93%). The modeled photosynthetic cover
was compared to field data consisting of point frame plots (n = 30) at each site. Correlations
between field and UAS-derived cover estimates showed significant positive relationships at the
Low Sage (r = 0.75, pr = 0.55, p = 0.002), but not at Wyoming Big Sage (r = 0.10,
p = 0.61). These results demonstrate methods to estimate photosynthetic cover at fine scales in
three types of sagebrush using UAS imagery. Additionally, these results suggest that UAS surveys
has high correlation with field measurements at mid and high elevation sagebrush sites, but
more studies are needed in low elevation sites to understand the potential of integrating UAS
and field observations of photosynthetic cover.
Our second study quantified fractional cover of plant functional types in the same three
sagebrush sites listed above. First, we tested Object-Based Image Analysis (OBIA) for
classification of UAS surveys into plant functional types. We assessed the accuracy of
the maps using confusion matrices; overall classification accuracies were strong: Wyoming
Big Sage (70%), Low Sage (73%), and Mountain Big Sage (78%). The classified maps of plant
functional types were compared to data from field plots (n = 30) at each site. We found
significant positive correlations for shrubs (r = 0.58; 0.83), forbs (r = 0.39;
0.94), and bare ground (r = 0.61; 0.70) at our Low Sage and Mountain Big Sage.
However we did not find significant relationships for the gramminoid class at any site (r =
0.18; 0.3; 0.32). Second, we tested the application of OBIA to sum shrub abundance from
UAS imagery. Abundance data from field plots (n= 24 per site) were tested for agreement
with UAS imagery. We found no correlation at any site with field observations at the 10m2 scale
(r = -0.22; 0.12; 0.26). Overall, we were able to calculate percent cover for
large-unit plant functional types, such as shrubs, trees, and some forbs. Accuracy for
gramminoid classification was low due to small plant size, confounding soil reflectance,
and grasses that grew beneath shrub canopies.
This research demonstrates that UAS methods can be used to estimate photosynthetic cover
and map plant functional types. Using UAS surveys also increased coverage and sampling
density of data when compared to traditional field observations. These findings help land
managers, restoration experts, and other researchers who monitor, manage, and protect
dryland ecosystems by demonstrating an accurate and less expensive approach to collecting
ecosystem data.