Fig 1.
Map of the area covered by the entire Clearwater–Nez Perce 3DEP 2016 lidar acquisition, highlighting the QL1 data boundaries and the 24 randomly-selected PLSS sections.
Background map is a composite of Landsat 8 Operational Land Imager (OLI) Collection 2 Tier 1 Level-2 Science Product (L2SP) surface reflectance scenes (bands 4, 5, and 6) downloaded from USGS EarthExplorer [77]. Fig 1 was created using ArcGIS Pro version 3.0 from Esri.
Fig 2.
Example PLSS section illustrating an intermediate step in the voxel-based metric calculation process.
The section has been divided horizontally and vertically into 5-m resolution voxels. Statistics were then calculated to summarize points contained within each voxel. In this example, each point represents the mean height within the respective 5-m voxel. This data is used to subsequently calculate the voxel-based metrics shown in Table 1 for the entire PLSS section.
Table 1.
Lidar DEM-, point cloud-, and voxel-derived metrics.
RT: all = all vegetation returns ≥ 0.27 m; first = first vegetation returns ≥ 0.27 m; last = last vegetation returns ≥ 0.27 m.
Table 2.
Satellite-derived metrics.
Fig 3.
Example locations of the one mobile and five stationary goTenna Pros that may result in each of the six connectivity levels.
The mobile goTenna is shown at various locations along the diagonal walking path within a PLSS section. Panels A–F represent connectivity levels Con_6–Con_1, respectively. These connectivity levels correspond to instances in which five, four, three, two, one, or zero stationary devices are connected to the mobile goTenna Pro, respectively.
Table 3.
Descriptions of the six compositional response variables representing the six connectivity levels and corresponding calculation methods.
Table 4.
Uncorrelated Boruta-selected variables specific to each connectivity level used to build initial LIDSAT, LID, and SAT Dirichlet regression models.
Table 5.
Summary statistics of the six connectivity levels of the dependent variable (proportion of time connected).
Fig 4.
Bar chart of the number of times variables in each category were selected by the 100 iterations of the Boruta algorithm for the LIDSAT, LID, and SAT datasets.
Fig 5.
Bar chart of the number of variables from each category that were selected by the 100 iterations of the Boruta algorithm for the LIDSAT, LID, and SAT datasets.
Table 6.
Summary of LIDSAT Dirichlet model regression coefficients for each of the six connectivity levels.
Table 7.
Summary of LID Dirichlet model regression coefficients for each of the six connectivity levels.
Table 8.
Summary of SAT Dirichlet model regression coefficients for each of the six connectivity levels.
Table 9.
Accuracy metrics calculated for each connectivity level using LOOCV.
Fig 6.
Boxplots of observed and predicted proportions of the six connectivity levels.
Predictions were obtained using the final LIDSAT, LID, and SAT models fitted to the entire dataset.