Environmental Engineering Reference
In-Depth Information
Table 15.2 Statistical description of the statistical distribution of forest and urban patch sizes (ha)
within the Maryland Piedmont in 1992 and 2001
Year
p
1
Land cover type
N
Minimum Median Maximum Sav
1992
Forest
0.2820
16,145
0.09
0.9
14,072
2,742.5
1992
Urban
0.1346
33,926
0.09
0.18
26,706.7
13,753.1
2001
Forest
0.2734
14,807
0.09
0.9
13,247
2,479.4
2001
Urban
0.1429
32,816
0.09
0.18
29,308.5
15,410.9
Table 15.3 Land cover change within the Maryland Piedmont from 1992 and 2001
Land cover type
Change (1,000 ha)
Percent change
Open water
24.5
0.40
Urban
141.2
2.80
Barren lands
50.1
16.90
Forest
-516.6
-1.50
Grassland
155.1
12.50
Agriculture
244.4
1.00
Wetlands
129.6
1.20
remained constant over this 9-year period. However, the geometric average patch
size (Sav) declined by 263.1 ha and the size of the largest forested area decreased by
825 ha. The size distribution of urban areas showed an 11% increase in Sav from
13,753 ha to 15,411 ha and a simultaneous increase in the largest urban area from
26,707 to 29,309 ha. The increase in urban area occurred with a declining number
of urban patches from 33,926 in 1992 to 32,816 in 2001 (Table
15.2
). The pattern of
urban growth occurred as a result of an increase in the average patch size and the
subsequent joining of adjacent urban areas. Because the median value of the size
distribution for urban areas was constant (Table
15.2
), the process of absorption
appears to have affected all size classes of urban development.
The assessment of map accuracy for MRLC data is an important consideration, but
also a complex topic (see Homer et al. 2004, 2007; Vogelmann et al. 1998; Wickham
et al. 2004; Yang et al. 2001) that is beyond the scope of this analysis. If we assume
that open water changed little between 1992 and 2001, then the 0.4% increase in open
water over this time period (Table
15.3
) provides a simple and convenient index of
relative accuracy. Table
15.3
shows that the only category that lost area was forest,
with a net decline of 516.6 thousand ha; All other land cover categories increased in
area, with the greatest gain for the agriculture, grassland and urban categories.
Figure
15.1
illustrates the cumulative frequency distribution (cfd) of the size of
forest patches in 1992 with the cfd generated by the RwC model of Qrule. The cfd
for 2001 was similar in form and is not illustrated here. The random maps generated
by the RwC model had a greater number of small, isolated clusters with 50% of the
patches smaller than 0.18 ha, while the median patch size for the 1992 land cover
map was nearly a hectare in size (0.9 ha, Table
15.2
). The largest patch for the
random maps was 18.3 ha while the largest for the 1992 land cover data was
14,072 ha. It is clear that the random and empirical distributions (Fig.
15.1
) are
different, and the KS test confirms this result (
D
0.021,
0.02). The RwC
¼
a ¼
