ul 2004
NW
nparing
latively
ange in
derived
[ESMA
le et al
2.
Average
Jifference
3 aerial
n dates
one Is
er well-
) rather
indicate
sures at
all four
ns have
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BY. Istanbul 2004
the smallest total mean difference from the reference data
aggregated over the test sites. The accuracy of the impervious
surface fractions is slightly lower in Datel, while the shade
fractions have the lowest accuracy in both dates. The overall
accuracy results are also consistent with the individual results
by site.
3.3 Results from Analyzing Change through Fuzzy Logic
Change measures were produced by subtracting the 1990
endmember fraction images from those of 2000. The advantage
of using the fraction images produced by MESMA lies in their
ability to reveal whether or not a change occurs, the direction of
change (increase, decrease), as well as classes of land cover
undergoing change. In order to improve interpretation of
results, it was appropriate to analyze change at the census tract
level rather than on a pixel-by-pixel basis. This is because
census tracts represent a reasonable spatial unit in Los Angeles
that accounts for the variations in demographic and
socioeconomic variables. Furthermore, analyzing change at the
census tract level helps establish a direct link between change in
urban land cover and underlying population forces, thus
offering an effective way for improving our understanding of
the impact of sustainable policies. Once, the average amount of
change per census tract was calculated from each image, fuzzy
set functions indicating the different magnitudes of change in
fraction (i.e. higher increase, lower increase, no change, lower
decrease, and higher decrease) were then applied to indicate the
degree of membership of each census tract in these sets. The
end product was an index of the severity level of change for
each land cover class assigned to each census tract. Based on
these indices, a threshold should value was used to classify the
census tracts according to the different levels of change
severity. This threshold value indicated the degree to which one
is certain about the compatibility of the final classification
results with the change concepts represented by the fuzzy sets.
The results presented in Figure 2 are based on a threshold value
of 0.70.
As shown in the figure, the types of change in urban land cover
vary remarkably between the core of the city and its periphery.
In addition, the magnitudes of change in different types of land
cover are generally limited to one of three levels: lower
decrease, no-change, or lower increase An exception to this are
the shade fractions that do not encounter any decrease and have
higher increases in both the center and southern parts of the
region. The observed patterns of change in vegetation and soil
fractions represent the typical trend one can find in megacities,
where rapid urban growth tends to occur first, followed by
slower internal modification in the old fabric of the city. In the
case of vegetation, the magnitudes of change tend to be low and
usually represent a decrease or loss of vegetation on the city
edges. As one would expect, this trend is coupled with a
contrasting pattern of change severity in impervious surface,
which experiences lower increases on the periphery of the city.
The effects of greenness policies in the southwestern region of
Los Angeles (Keil and Desfor, 2003; Pincetl et al., 2003) reveal
physical changes in vegetation cover and are reflected by an
increase in vegetation fractions. At the sane time, lower
decreases can be observed in the impervious surface of the city.
One explanation of this pattern is directly related to the
relatively low spatial resolution of the multispectral images
utilized in the analysis, compared to the varying character of
Los Angeles' urban morphology. That is, change in impervious
surface is observed where the morphological pattern of the
urban area is relatively sparse. Change in impervious surface is
difficult to observe where the pattern of the urban morphology
507
is dense. This is the case for downtown Los Angeles. This idea
is augmented by the higher increases in shade fractions around
the downtown, which can likely be attributed to the
densification processes taking place in these highly populated
areas.
Veg 5 "ein. Imp
Noe,
I High Decrease
Low Decrease
: ^
[CJ No Change Soil Shd
[] Low Increase
High Increase
Figure 2: Census tracts in The Metropolitan Area of Los
Angeles County Classified by the Magnitude of Changes in
Land Cover Fractions
3.4 Results from Applying Landscape Metrics
Landscape metrics were applied to change fractions to quantify
the ecological patterns (that is, the configuration) of land cover
change at different magnitudes. These patterns describe the
spatial character and arrangement, position, or orientation of
pixel encountering a change in land cover within the urban
locale. A full explanation of the results from this exercise is
beyond the scope of this paper. However, to illustrate the
usefulness of the proposed methodology, a sample of the results
from applying landscape metrics is shown in Figure 3. Figure 3-
A and 5-B respectively represent the degree of cohesion among
pixels encountering increases and decreases within census
tracts. The darker areas represent census tracts assigned to
higher values in the cohesion index, whereas lighter areas
represent lower values. Cohesion generally refers to the
functional connections among patches within a landscape. In
this example of vegetation change, functional connection may
imply organized activities aiming to increase the greenness of
urban areas (increase) or unsustainable practices (decrease)
activities. As shown in Figure 3-A, census tracts near
downtown and NW Los Angeles show a cluster of higher
connectivity among pixels that experience vegetation increase
(no matter what their magnitude is), which could be a reflection
of efforts aiming at naturalizing the region. On the other hand,
Figure 3-B shows a random pattern of pixels that experience
vegetation decrease, which cannot be linked to organized
activities that result in environmental degradation.
Comparing the patterns resulting from these two metrics with
patterns of change in the impervious surface and soil fractions
produced by either the same metrics or others, one can start
comprehending the degree to which sustainable policies have
become effective in Los Angeles County. Moreover, the fact
that these metrics are assessed and calculated at the census tract
level implies that one would be able to link these results to a
range of demographic, socioeconomic and policy variables and
thus to construct a full story of to what is going on in Los
Angeles County, a story in which change is the rule not only in
the periphery, but also in the core areas.
