Image Processing Reference
In-Depth Information
Table 12.4
Statistics for MODIS data for urban (U) and nonurban (NU) grid cells
Minimum
Maximum
Mean
±1 s
Dataset (units)
(U)
(NU)
(U)
(NU)
(U)
(NU)
(U)
(NU)
Albedo (unitless)
0.10
0.05
0.30
0.31
0.19
0.20
0.03
0.04
fPAR (%)
0.10
0.08
0.81
0.82
0.28
0.26
0.10
0.11
LAI (m
2
/m
2
)
0.10
0.10
3.30
4.30
0.38
0.37
0.22
0.29
Day surface
temperature (K)
24.50
24.50
51.00
52.00
47.26
47.74
1.66
2.96
Night surface
temperature (K)
0.00
0.00
30.00
30.00
27.55
26.82
2.08
4.48
NDVI (unitless)
0.04
0.07
0.72
0.74
0.24
0.23
0.08
0.11
This characteristic scale is an order of magnitude below our scale of investigation,
however the weak negative correlations we obtain suggest that some effect of the
distribution of the Asphalt and Built classes is still discernable in the 1 km scale
reflectance measured by MODIS. Figure
12.4
illustrates the overall distribution and
high degree of mixing of the Built class with other land cover types in the urban
region; the metric results for the Asphalt class are similar. The results of Small
(
2003
) also suggest that the 15 m spatial scale of ASTER is adequate to capture
land cover variations (and by extension, landscape structure) of importance to
urban ecological and climatological studies.
The results for fPAR indicate that this variable is weakly correlated with urban
CA for both the Undifferentiated Vegetation and Built classes. Weak correlation is
also obtained for urban ED with the Undifferentiated Vegetation class, and for
urban IJI with the Built class. These results suggest a connection between fPAR and
older mesic residential neighborhoods which tend to have larger lot sizes and higher
amounts of trees, grass, and shrubs (GP2100
2003
). No correlations are observed
for LAI with any metrics or classes which most probably reflect the low proportion
of broad-leafed plant species in the Phoenix area (Whitford
2002
; Hope et al.
2003
). Somewhat surprisingly there is also a weak correlation observed for fPAR
with the urban Asphalt class for IJI. This may be a result of a high degree of asso-
ciation between asphalt roadways and residential areas. It should be noted that the
fPAR and LAI values associated with “urban” pixels, as classified using the
MODIS Land Cover data product, are not modeled and therefore invalid
(Knyazikhin et al.
1999
). A total of 466 out of 1,599 urban grid cells (29%) are
invalid. By contrast only 32 out of 1,781 nonurban grid cells (2%) are invalid. This
aspect of the fPAR and LAI datasets is unfortunate, as it limits detailed study of
these variables in urban core areas using these MODIS data products.
Daytime surface temperature shows a weak positive correlation with urban CA
for the Soil and Bedrock class. This may result from the presence of several urban
mountain parks (Piestewa Peak/Dreamy Draw, Papago, South Mountain, and
Camelback Mountain) with relatively little soil or vegetation cover. Relatively strong
negative correlations are obtained with both CA and ED for the urban Undifferentiated
Vegetation class. This reflects the relatively high extent of vegetated surfaces in the
Phoenix metropolitan region (as parks, golf courses, and mesic residential develop-
ments) and suggests that distribution of this land cover type may influence surficial
