Environmental Engineering Reference
In-Depth Information
Correlated Metrics
C.len
S.frq
Sav
T.eg
T.cltr
LC.rms
LC.frc
LC.eg
LC.sz
Fig. 15.4 Scatter plots of landscape metrics from ten iterations of Qrule using a neutral model
with constraints (the RwC model) derived from the 1992 map of forested areas in the Maryland
Piedmont. See text for a description of these metrics
simulations. Therefore, one interesting test is to examine a subset of the association
matrix: the frequency of urban areas associated with forested areas from 1992 to
2001. The observed frequencies of urban:forest association for the actual land-
scapes were 2.14
10
-3
for 1992 and 2001, respectively; the
expected frequency for the RwC simulations were 1.2
10
-3
and 1.82
10
-2
and 1.23
10
-2
,
2
respectively. Converting to counts, we can test these differences with a 2
2
w
10
9
and 7.1
10
9
for 1992 and 2001, respectively.
contingency table yielding 8.9
2
value for
The critical
0.0001with 1 d.f., as estimated by Conover (1971,
w
a ΒΌ
10
7
. Thus, there is significantly less urban:forest association than
expected by chance alone for both time periods. The decline in forest occurred
with the simultaneous increase in urban areas, producing a loss of 235 km of forest
edge over this 9-year period. Applying an
p. 367), was 2.4
2
test to this change gave a value of
w
