Environmental Engineering Reference
In-Depth Information
and habitat type in comma delimited format, including the ten indices summarized
in rulerun.log; and patch_cfd.dat, the cumulative frequency distribution (cfd) of
patch sizes for each iteration and habitat type in summary form (i.e. the cumulative
frequency in each size class).
Manipulation of output files generated by Qrule by other programmes is neces-
sary for visualization and statistical testing. To assist in this sometimes tedious
process, a series of programmes have been written in R (R Team 2008) to display
results and test for significant differences. The statistical results and visualization
reported here were produced by these programmes.
Maryland Piedmont Maps
Data from the National Land Cover Database for the Piedmont of Maryland were
downloaded and were selected for this analysis ( http://www.mrlc.gov/download_
data.php ). The multiple categories for 1992 and 2001 were aggregated into seven
classification level comparisons: open water, urban, barren land, forest, grassland,
agriculture and wetlands. The 30-m resolution of the maps resulted in grids with
dimensions of 4,365 rows and 5,550 columns. These two maps were first analyzed
with Qrule using the next-nearest-neighbor rule for patch identification. Qrule
results provide a select set of indices characterizing patterns (see Gardner 1999
for a listing and description of these indices). The indices reported here include
three calculations characterizing the largest cluster: LC.sz, the size of the largest
cluster in hectares; LC.ed, the amount of edge for the largest cluster in meters; and
LC.frc, the fractal index of the largest cluster. Also considered are five general
indices for each land cover type: S.frq, the total number of pixels for each land-
cover type (pixel units); T.cltr, the total number of patches; C.len, the average
correlation length of patches in meters; Sav, the area-weighted average patch size in
hectares; T.eg, the total amount of edge in meters.
Many comparisons could be constructed from this information-rich data set. The
analysis reported here focuses on patterns of loss of forested areas because this
landscape-attribute has been threatened by population growth and urban develop-
ment within this region (Lookingbill et al. 2009) and has resulted in special
legislation to protect forested areas ( http://www.dnr.state.md.us/forests ).
15.4 Analysis and Hypothesis Testing
Forested areas occupied 28.2% of the Maryland Piedmont in 1992, declining to
27.3% in 2001 (Table 15.2 ), a change equivalent to the loss of nearly 6,000 ha.
Urban areas showed the opposite trend, increasing from 13.5% in 1992 to 14.3% in
2001, a change that produced an additional 5,673 ha of urban development. The size
distribution of forest patches also shifted although the median value of 0.9 ha
Search WWH ::




Custom Search