Geoscience Reference
In-Depth Information
4.3.2 Geospatial Modeling
4.3.2.1 Input Variables and Geospatial Model
Seven geospatial variables were used in our spatial model (Table
4.1
). Slope curva-
ture, elevation, topographic position index (TPI), slope steepness, and topographic
relative moisture index modified (TRMIM) were derived from a 10 m resolution
digital elevation model (DEM). Slope curvature, elevation, and slope steepness were
calculated using tools in ArcGIS 9.2 (ESRI, Redlands, CA, U.S.), TPI was calcu-
lated in ArcView 3.2 (ESRI, Redlands, CA, U.S.) using an extension developed
by Jenness (
2006
), and TRMIM was calculated in Arc Grid (ESRI, Redlands, CA,
U.S.) based on Parker's (
1982
) method.
Fine resolution geologic data were also used in the model. Geological formations
were classified into sandstone and limestone families, and Euclidean distance from
the boundary of the sandstone and limestone formations was then calculated and
used in the spatial model. Land use history was reconstructed based on the 1936
vegetation map provided by Mammoth Cave National Park. Based upon the park's
history, we classified the non-restocking and restocking categories as historical agri-
cultural areas (previously cultivated fields and pastures) and the other forest types
as non-agricultural areas. Because land use history is a human driven factor, we did
not include it in the habitat modeling; however, it was incorporated as a mask to
build a map to locate surviving American chestnut sprouts.
In this study, only locations that had American chestnut were recorded. Hence,
we employed ecological niche factor analysis (ENFA) to compute the habitat suit-
ability map for American chestnut based on the presence-only data in Biomapper 3.1
(Hirzel et al.
2004
). Data partitioning was applied to all the 2156 located American
chestnut sprouts as a validation technique (Fielding and Bell
1997
). Two thirds
of the specimens (1437) were used in Biomapper to build the spatial model, and
one third (719) were used to validate the model. Distributions of the variables that
were recognized by ENFA to have strong association with chestnut were further
compared between chestnut locations and random locations (1437 spatially random
points generated in GIS) in the study area. Biomapper only provides a range of
Table 4.1
Variables used in spatial modeling of American chestnut habitat at Mammoth Cave
National Park, Kentucky, U.S.A.
Variables
Resolution (m)
Description
Curvature
10
Convexity/concavity based on DEM
Elevation
10
Elevation from DEM
Topographic Position
Index
10
Topographic position based on DEM (Jenness
2006
)
Slope
10
Slope steepness (degree)
Topographic relative
moisture index
10
Dryness-wetness index based on DEM (Parker
1982
)
Land use history
-
Land use history (based on 1936 vegetation
map)
Geology
-
General bedrock formation (1:24,000 scale)
