250 m resolution. No obvious changes were observed in zones 5 and 6 across the
scales, except a temporary decrease at 250 m indicating that area percentages of both
zones seemed to be more affected at the 250 m resolution during the aggregation
process. For the image date October 3, 2000, zone 1 had a slight decrease in patch
percentage from 250 to 500 m resolution. The percentages of zones 2 and 4 showed
much more obvious but still mild changes across the scales. Zone 3 as the largest zone
in area did not show significant variations across the scales. There was a slight
increase in patch percentage in zone 5 at 250 m. The patch percentage of zone 6
remained consistent with some minor increases after 250 m resolution.
The patch densities of LULC types became lower and lower from 15 to 1000 m
resolution for each image date. For each season/image date, urban, forest, and
grassland experienced notable decreases before the scale reached 60 m, and then
became much gentler toward 250 m resolution. No clear variations could be observed
after 250 m. Meanwhile, water, agriculture, and barren lands experienced decreases in
patch density across the scales. Patch densities of LST types did not undergo obvious
changes as the scale changed from 15 to 90 m followed by notable decreases. It is clear
that the higher the patch density, the more variations in the temperature zone across
the scales. All LST patch densities seemed to be equal at 1000 m resolution, indicating
a maximum patch aggregation at the scale of 1000 m.
No clear seasonal changes could be observed in landscape aggregation based on
the measurements of landscape shape index for both LULC and LST maps. The
aggregation level of all LULC types kept decreasing across the scales for both image
dates. It appears that highly disaggregated LULC types (urban, forest, and grassland)
experienced more changes during image resampling compared to those less aggregated LULC types (water, agriculture, and barren lands). A decrease in values was
observed in each LST zone since 90 m resolution. Zones with greater aggregation
levels (zones 2–4) appeared to be more affected during aggregation process than
zones with lower aggregation level (zones 1, 5, and 6).
Unexpected measurements of perimeter–area fractal dimension index were
obtained for both LULC and LST maps for each image date. As for the LULC
maps, the fractal dimension of each LULC type seemed to increase overall. However,
the aggregation process during image resampling intended to decrease but not
increase the overall level of fractal dimension. The results of the measurements of
perimeter–area fractal dimension indices on LST maps seemed to be more complicated than those on LULC maps. No distinct changes were shown at each temperature
zone for each image date as the resolution changed from 15 to 30 m. The values
slightly dropped at 60 m resolution and kept increasing since then until 250 m
resolution. The changes became inconsistent after 250 m resolution for both image
dates.
11.2.3.2 Scaling-Up Effect on Landscape-Level Landscape Metrics According
to Figure 11.4, the landscape patch density continuously decreased across the scales
for both image dates. It could be explained by the fact that the LULC patches became
more and more aggregated and the overall landscape contained fewer and fewer
interspersed patches during the scaling-up process. The October image had a
224
SCALING ISSUES IN STUDYING THE RELATIONSHIP
scales, except a temporary decrease at 250 m indicating that area percentages of both
zones seemed to be more affected at the 250 m resolution during the aggregation
process. For the image date October 3, 2000, zone 1 had a slight decrease in patch
percentage from 250 to 500 m resolution. The percentages of zones 2 and 4 showed
much more obvious but still mild changes across the scales. Zone 3 as the largest zone
in area did not show significant variations across the scales. There was a slight
increase in patch percentage in zone 5 at 250 m. The patch percentage of zone 6
remained consistent with some minor increases after 250 m resolution.
The patch densities of LULC types became lower and lower from 15 to 1000 m
resolution for each image date. For each season/image date, urban, forest, and
grassland experienced notable decreases before the scale reached 60 m, and then
became much gentler toward 250 m resolution. No clear variations could be observed
after 250 m. Meanwhile, water, agriculture, and barren lands experienced decreases in
patch density across the scales. Patch densities of LST types did not undergo obvious
changes as the scale changed from 15 to 90 m followed by notable decreases. It is clear
that the higher the patch density, the more variations in the temperature zone across
the scales. All LST patch densities seemed to be equal at 1000 m resolution, indicating
a maximum patch aggregation at the scale of 1000 m.
No clear seasonal changes could be observed in landscape aggregation based on
the measurements of landscape shape index for both LULC and LST maps. The
aggregation level of all LULC types kept decreasing across the scales for both image
dates. It appears that highly disaggregated LULC types (urban, forest, and grassland)
experienced more changes during image resampling compared to those less aggregated LULC types (water, agriculture, and barren lands). A decrease in values was
observed in each LST zone since 90 m resolution. Zones with greater aggregation
levels (zones 2–4) appeared to be more affected during aggregation process than
zones with lower aggregation level (zones 1, 5, and 6).
Unexpected measurements of perimeter–area fractal dimension index were
obtained for both LULC and LST maps for each image date. As for the LULC
maps, the fractal dimension of each LULC type seemed to increase overall. However,
the aggregation process during image resampling intended to decrease but not
increase the overall level of fractal dimension. The results of the measurements of
perimeter–area fractal dimension indices on LST maps seemed to be more complicated than those on LULC maps. No distinct changes were shown at each temperature
zone for each image date as the resolution changed from 15 to 30 m. The values
slightly dropped at 60 m resolution and kept increasing since then until 250 m
resolution. The changes became inconsistent after 250 m resolution for both image
dates.
11.2.3.2 Scaling-Up Effect on Landscape-Level Landscape Metrics According
to Figure 11.4, the landscape patch density continuously decreased across the scales
for both image dates. It could be explained by the fact that the LULC patches became
more and more aggregated and the overall landscape contained fewer and fewer
interspersed patches during the scaling-up process. The October image had a
224
SCALING ISSUES IN STUDYING THE RELATIONSHIP