Geoscience Reference
In-Depth Information
habitat suitability (0-100) and does not provide a threshold of favorable habitat.
We used the maximum cumulative frequencies difference method (Browning et al.
2005
, Thompson et al.
2006
) to estimate the threshold. We first obtained habitat
suitability values for both the 1437 chestnut locations and 1437 random locations,
calculated the cumulative frequencies of locations by habitat suitability, and then
located the maximum difference between the two cumulative frequencies to define
the threshold.
The chestnut habitat suitability model derived from ENFA was then evalu-
ated using two approaches. The first approach was a cross-validation conducted in
Biomapper 3.1. The study area was partitioned into 10 sub-regions to validate the
model. To minimize the spatial-autocorrelation among the partitions, locations of
the partitions were randomly assigned. In the second approach, we used the 719
trees that were not included in the habitat modeling process to test the robust-
ness of the model derived from ENFA. Percentages of trees located in each of the
probability zones in the habitat model were then calculated.
4.3.2.2 Site Affinity
Land use history had a strong influence on the current distribution of chestnut
sprouts. Based on the 1936 vegetation map, about 41% of the area (8057 ha) was
classified as abandoned agricultural land. A total of 89% of the chestnut sprouts
at Mammoth Cave National Park were located within areas that were classified as
non-agricultural land. The remaining 11% of the chestnuts were located within a 10
m buffer around non-agricultural land. Thus, all surviving chestnuts were located in
or close to the edge of historically non-agricultural land.
Based on ENFA, chestnut distribution was strongly associated with elevation,
geological formation, slope steepness, and TPI (Table
4.2
). The three factors derived
from ENFA explained 98.3% of the total variance. Factor 1 which explained most
of the variance (78.8%) was mainly due to the Euclidean distance from the edge of
sandstone and limestone formations. Factor 2 was mainly due to slope steepness and
elevation, while Factor 3 was mainly contributed by elevation and TPI. Geological
formation, as expressed as the Euclidean distance, had a much stronger association
with the distribution of chestnut than the other variables.
Different distribution patterns by variables derived from chestnut locations and
random locations further confirmed the strong site affinities of chestnut sprouts.
American chestnut sprouts were more frequently located near the border line of
sandstone and limestone formations, with a buffer zone of 40 m on each side
of the boundary. Chestnut sprouts were distributed in a relative narrow elevation
range of 200-240 m. Chestnut sprouts were also more frequently located on steeper
slopes between 25 and 40
◦
. The distribution patterns between chestnut locations
and random points of TPI do not have an apparent difference compared to the other
three variables. However, relatively low chestnut presence on ravine and ridge sites
compared to higher presence on steep mid-slopes and upper slopes was observed.
A comprehensive view of chestnut distribution on a topographic relief geological
map further confirms the association between chestnut sprouts and environmental
variables (Fig.
4.4
).
