Image Processing Reference
In-Depth Information
The metric results for the Soil and Bedrock class suggest that while the majority
of surficial soil and rock material in the urban area is not highly intermingled with
other land cover types (relatively low IJI), these areas are irregular in shape (rela-
tively high ED). The boundaries between urban and peri-urban regions are not as
sharply defined for the Soil and Bedrock class as for the other classes. This is to be
expected as the majority of the peri-urban region is comprised of desert soils with
low vegetation cover or exposed bedrock (GP2100
2003
). The Agriculture class has
minor representation in the urban region, however it exhibits low to moderate ED
in keeping with the generally rectilinear plan of agricultural fields in this region.
The Undifferentiated Vegetation class shows highest ED values along the Salt River
bed and in areas of mesic residential land cover/land use. While the Salt River bed
is generally dry, there is enough water released in association with an upstream
impoundment of water (Tempe Town Lake) to support a sparse riparian vegetation
community (Musacchio
2003
). The IJI values for the Undifferentiated Vegetation
class are similar to those for Asphalt; relatively high values are recorded in older
mesic regions, while variable results are associated with the remainder of the urban
area. Both ED and IJI values for the Built class are generally high throughout the
urban region. This suggests that this class is evenly distributed, highly intermingled
with other land cover types throughout the urbanized area, and has generally irregu-
lar boundaries with other classes at the 1 km scale of analysis. The land cover clas-
sification presented in Fig.
12.2
illustrates the degree of mixing present between
soil, built, and vegetation land cover types in the Phoenix region.
We hypothesized that the landscape structure of the Phoenix urban/peri-urban
area correlated with a variety of biophysical parameters at the 1 km scale of a
MODIS pixel. The results presented in Table
12.3
suggest that this hypothesis is not
generally correct. Table
12.4
presents descriptive statistics for the MODIS datasets.
Examination of the results for both the urban and nonurban grid cells indicates
generally low variance in the MODIS data; standard deviations are typically an
order of magnitude or more about the means. This suggests that at the 1 km/pixel
scale of the MODIS data, the Phoenix metropolitan area appears fairly uniform.
The exceptions to this general conclusion are the fPAR and LAI datasets. The
observed high variance of these datasets is however the result of small sample size
rather than actual variation across the metropolitan region (discussed below).
Nevertheless, there are indications of some weak correlations between MODIS
variables and the landscape metric results. We therefore limit the following discus-
sion of results to positive and negative correlations greater than 0.2 and −0.2 respec-
tively. Albedo measurements show a positive correlation with CA for the aggregate
Agriculture class in the nonurban area. This is a result of using the full wavelength
range albedo (0.3-5.0 mm) which includes the near-infrared reflectance peak from
plants. The Agriculture class also contains both vegetated and fallow agricultural
fields with high soil brightness. Finally, the Agriculture class is concentrated pri-
marily in the nonurban grid cells. Relatively strong negative correlation with albedo
is observed for both urban Asphalt and Built classes in the CA, ED, and IJI metrics.
Small (
2003
) demonstrates that the characteristic spatial scale of albedo variations with
land cover for 14 urban centers (including Phoenix) is 10-20 m using IKONOS data.
