Vegetation Classification and Mapping Project Appendices - USGS

USGS-NPS Vegetation Mapping Program
Waterton-Glacier International Peace Park
Variables selected for the environmental stratification were elevation, aspect, soils, and disturbance (as
captured in fire history; see the three tables below—Stratification Variables and Classes—for classes
delineated for each variable). Elevation and aspect are derived from DEMs. National Park Service staff
provided coverage of soil classes and fire history of GNP. These datasets were transformed into grids
with the same cell size and reference as the DEM. All coverages and grids were in the Universal
Transverse Mercator (UTM) zones 11 and 12 projection, North American Datum of 1983 (NAD 83). For
the gradsect analysis, all grid data were projected in UTM zone 12, again using NAD 83.
For each variable, the grid cells were reclassified into the classes established by the working group. The
grids were then intersected to find the unique spatial combinations of variables that constituted polygons,
termed Biophysical Units (BPUs), which became the units of landscape (environmental) stratification. An
example of a BPU, which could have many polygons, would be “Deep soils of argillite or quartzite,
between 1351–1700 m in elevation, with pre-1840 fire, on north aspects.”
In the year 2000, the environmental stratification was modified to (1) smooth the 30-meter DEM to
remove noise, (2) reduce the artificial proliferation of BPUs by renumbering the classes (data layers and
classes unchanged), and (3) combine polygons (BPUs) less than 0.18 ha with their nearest neighbors. To
ensure that the full environmental gradient east of the CD was sampled, it was divided into the Hudson
Bay and the Missouri River (northern-portion and southern-portion) watersheds. Each BPU east of the
CD was then attributed by drainage. This coverage was joined spatially to the plots already collected, and
the resulting BPUs previously sampled were removed from consideration for the 2000 field sampling.
BPUs occurring in only one drainage were deleted from consideration. Remaining in the sampling pool
were those BPUs completely unsampled or sampled only in one portion of range of occurrence. West of
the CD was not comparably stratified to accommodate north-south variation.
The year 2000 analysis resulted in 17,808 polygons attributed to 377 BPUs across GNP east and west of
the CD. To winnow this number of BPUs to one capable of being efficiently accessed and sampled, the
coverage of BPUs was joined to an accumulated cost surface of GNP. Sites were selected to complete the
full sampling of each BPU based on size and travel cost (selecting BPUs in physical proximity of one
another). Several BPU types were deleted as targets because no examples of sufficient size occurred (all
polygons of 0.18 ha in size or smaller were dissolved into the adjacent polygons), or because the only
available polygons would have been hazardous for the field crews to sample. If possible, at least three
polygons of each BPU were selected for sampling. These were distributed throughout the range of the
BPU, thus replicating examples of each BPU type. Polygons covering hazardous area were masked from
selection for sampling; for example, polygons where the average slope was 50o or steeper were not
selected. The average polygon size was 7.45 ha and on average required approximately 800 m (about 0.5
mi.) of hiking on moderate terrain to access.
Appendix D: Gradsect Sampling Design Methodology D-2