Geoscience Reference
In-Depth Information
Fig. 4.6
Relationship between F-ratio (predicted chestnut presence/expected chestnut presence)
and habitat suitability
cross-validation on the 10 sub-regions random partition. The CBI indicated that our
chestnut habitat suitability model was very robust (CBI equal to one indicates a per-
fect model). In addition, the relationship between the F-ratio (predicted/expected
ratio) and probability of habitat suitability derived from the cross-validation also
indicated that the chestnut habitat suitability model was robust (Fig.
4.6
) because
the F-ratio was low when habitat suitability was low, F-ratio was high when habitat
suitability was high, and F-ratio was nearly monotonically increasing (Boyce et al.
2002
).
We also used partitioning (Fielding and Bell
1997
) to validate the model based on
locations of the 719 chestnuts that were not included in the habitat suitability model.
Over 87% of the chestnuts were located in the areas that were classified as favorable
chestnut habitats (Fig.
4.5b
), and nearly 90% of the chestnuts were located in the
favorable habitats if the habitat suitability map was smoothed by a 3
3 filter with a
mean value algorithm. This further indicates that our chestnut suitability model was
robust.
Our results indicate that geospatial analysis was effective in determining both the
chestnut site affinities and habitat map in Mammoth Cave National Park. American
chestnut sprouts have a relatively narrow niche in Mammoth Cave National Park.
Strong site affinities were found for American chestnut sprouts in Mammoth Cave
National Park. They have a very low presence on relatively young abandoned agri-
cultural lands, but most often occur in less disturbed forests on relatively steep mid
to upper slopes near the boundary of limestone and sandstone formations, but with
greater preference for the sandstone soils. We believe this model can be a valu-
able tool for locating surviving American chestnut trees for use in chestnut breeding
programs. Another potential use of this model is to identify restoration sites where
planted chestnut seedlings may have a good chance to survive. The current distri-
bution of chestnut sprouts appears to be a combined result of land use history and
actual chestnut site affinities. The potential restoration sites predicted by the model
may be suitable areas for chestnut trees to survive but may not be the best sites for
chestnut trees to thrive.
×
