3.4 Beta Diversity (β)
For “pairwise plots” (see Methods), adonis2 tests revealed that
dissimilarity increased significantly from the pre-fire to the post-fire
sample for total β, species turnover, and nestedness (Figure 3). Species
turnover contributed much more to total β than did nestedness: mean
distance from centroids were 0.387 for turnover and 0.070 for nestedness
for pre-fire data and 0.445 and 0.093, respectively, for post-fire data.
Total β and turnover both increased for juniper woodlands and piñon
woodlands tested separately but not for pine-oak forest (Figure 4).
Nestedness increased for juniper woodlands and pine-oak forest but not
for piñon woodlands. Total β and turnover exhibited significant temporal
and spatial trends, more so than did nestedness. If fire were associated
with the temporal increase in β, we would expect β to increase over time
for burned but not for unburned plots—a hypothesis supported for total
and turnover results, but not for nestedness, which increased for burned
and unburned plots alike (Figure 5). For both pre-fire and post-fire
data, total β declined in a curvilinear manner with elevation, increased
linearly with TRMI, and declined moderately with TRI (Table 2, Figure
7).
For “matched plots,” the timestep component is embedded in the
individual plot pre-fire/post-fire dissimilarities. Mean dissimilarity
differed among vegetation types for total β and turnover, but not for
nestedness (Figure 3). For total and turnover β, juniper woodland plots
changed more than those in piñon woodland, but other pairwise vegetation
type comparisons were not significant. If fire were a driver of temporal
changes in β, we would expect higher pre-fire/post-fire dissimilarity
for burned than unburned plots for β—a hypothesis supported for total
β (Figure 4). In fact, matched plot beta diversity increased with dNBR
up to high severity, at which point it declined slightly (Figure 7).
When plot dissimilarity vs. dNBR was analyzed separately for each
vegetation type, this relationship was significant only for pine-oak
forest, which exhibited the same curvilinear relationship of β with dNBR
as found for all plots (Figure 6). Finally, plot dissimilarity decreased
from low to high elevation for all β components, increased with TRI for
total only, and was unrelated to TRMI (Table 2, Figure 7).
For the “mantel test”, one analysis supported the hypothesis that
pyrodiversity begets β among plant communities, but another analysis did
not. As dissimilarity in fire severity between two plots increased so
did the amount of change in woody plant community dissimilarity from
before to after the Horseshoe Two Fire (mantel r=0.12, P =0.001).
These same pairwise plot contrasts in fire severity were not, however,
correlated with post-fire plot dissimilarities in communities alone
(mantel r=0.04, P =0.12). We rejected the null hypothesis that
these matrix relationships were artifacts of correlated effects of
physical distances among plots (mantel r=0.02, P =0.16). Including
distance as a second variable in partial mantel tests also did not
change the outcomes of the original analyses.