159
Remote Sensing Drought Assessment in a Coastal Urban Region
cover (Weng et al. 2004; Chen et al. 2006; Xie et al. 2010). Extensive comparisons
between LST and VIs associated with the two LULC analyses (1987 and 2000) in
our study area help identify different relational patterns as a result of the distribution
of these heat fluxes at the ground level. Our study uniquely found that LST and VIs
were negatively correlated in most cases of low, medium, and high vegetation cover,
except for the case of high-density vegetation cover in 2000 due to the effect of UHI.
For detailed comparisons, if the average LST (30.02°C) of the study area in 1987
was set as a reference temperature, the difference of LST between the reference
temperature and specific land use is phenomenal: −0.75°C for farmland, 1.39°C for
built-up land, 3.93°C for saline–alkali land, −1.91°C for beach land, and −4.76°C for
water body. This individual finding supports the urbanization effect on LST from
1987 to 2000. In particular, the rise of 1.39°C, on the average, over built-up land
was phenomenal. If we set the average LST of the study area in 2000 (29.23°C) as
the reference temperature, the LST of farmland was 0.82°C lower, that of built-up
land was 3.12°C higher, that of saline–alkali land was 4.13°C higher, that of beach
land was 3.21°C lower, and that of water body was 3.21°C lower. The urbanization
effect on LULC in 2000 was far greater than that in 1987, and the proportion of
built-up land in 2000 occupied 3% more than that in 1987, and therefore, the UHI
effect in 2000 was proved more influential. This is further supported by a 3.17°C
higher LST of built-up land in 2000 when compared with its counterpart, which
was 1.48°C higher in 1987.
7.4.2 andvi effect on lSt and uhi
The turning point in scatterplot data for RWSI and TVDIs (Figure 7.10) can be used
as an example to explain the relationship between ANDVI and LST. The analysis
can be conducted by means of four subgroups (i.e., ANDVI < 0, ANDVI between 0
and 0.03, ANDVI between 0.04 and 0.19, and ANDVI > 0.20) across the scatterplot
of 2000. A comparative analysis of the areas with ANDVI > 0.20 in 1987 and 2000
yields some insight. The entire study area can be classified into four groups: 8.5% in
1987 and 24.2% in 2000 is the first, 13.8% in 1987 and 22.2% in 2000 is the second,
57.2% in 1987 and 49.4% in 2000 is the third, and 20.5% in 1987 and 4.2% in 2000 is
the fourth. A direct comparison of these groups shows that the density of vegetation
cover in 1987 was significantly higher than that in 2000, especially the much higher
ANDVI 1987 value (>0.19) compared to that in 2000.
The area with the ANDVI value > 0.19 in 1987 and 2000 can be used to support a
focused analysis (see Figure 7.11), indicating that the selected area is tied to the sec7.11), indicating that the selected area is tied to the sec.11), indicating that the selected area is tied to the sec11), indicating that the selected area is tied to the sec1), indicating that the selected area is tied to the sec), indicating that the selected area is tied to the second turning point that is probably tied to UHI (Figure 7.10). This area was covered
mainly by farmland (crop), some scattered rural settlements, and sporadic urban
built-up land, accounting for 19.2% LULC of the total study area in 2000. The area
with ANDVI values between 0.04 and 0.19 accounted for the additional 72.72% of
the study area. Together, both types of LULC accounted for 91.92% of the total study
area. In addition, the area with ANDVI values between 0 and 0.03 accounted for
5.91% of the total study area in 2000 compared to 0.36% in 1987. Collectively, these
scattered rural settlements and urban land contributed to the overall LST directly
and UHI indirectly in 2000.
Remote Sensing Drought Assessment in a Coastal Urban Region
cover (Weng et al. 2004; Chen et al. 2006; Xie et al. 2010). Extensive comparisons
between LST and VIs associated with the two LULC analyses (1987 and 2000) in
our study area help identify different relational patterns as a result of the distribution
of these heat fluxes at the ground level. Our study uniquely found that LST and VIs
were negatively correlated in most cases of low, medium, and high vegetation cover,
except for the case of high-density vegetation cover in 2000 due to the effect of UHI.
For detailed comparisons, if the average LST (30.02°C) of the study area in 1987
was set as a reference temperature, the difference of LST between the reference
temperature and specific land use is phenomenal: −0.75°C for farmland, 1.39°C for
built-up land, 3.93°C for saline–alkali land, −1.91°C for beach land, and −4.76°C for
water body. This individual finding supports the urbanization effect on LST from
1987 to 2000. In particular, the rise of 1.39°C, on the average, over built-up land
was phenomenal. If we set the average LST of the study area in 2000 (29.23°C) as
the reference temperature, the LST of farmland was 0.82°C lower, that of built-up
land was 3.12°C higher, that of saline–alkali land was 4.13°C higher, that of beach
land was 3.21°C lower, and that of water body was 3.21°C lower. The urbanization
effect on LULC in 2000 was far greater than that in 1987, and the proportion of
built-up land in 2000 occupied 3% more than that in 1987, and therefore, the UHI
effect in 2000 was proved more influential. This is further supported by a 3.17°C
higher LST of built-up land in 2000 when compared with its counterpart, which
was 1.48°C higher in 1987.
7.4.2 andvi effect on lSt and uhi
The turning point in scatterplot data for RWSI and TVDIs (Figure 7.10) can be used
as an example to explain the relationship between ANDVI and LST. The analysis
can be conducted by means of four subgroups (i.e., ANDVI < 0, ANDVI between 0
and 0.03, ANDVI between 0.04 and 0.19, and ANDVI > 0.20) across the scatterplot
of 2000. A comparative analysis of the areas with ANDVI > 0.20 in 1987 and 2000
yields some insight. The entire study area can be classified into four groups: 8.5% in
1987 and 24.2% in 2000 is the first, 13.8% in 1987 and 22.2% in 2000 is the second,
57.2% in 1987 and 49.4% in 2000 is the third, and 20.5% in 1987 and 4.2% in 2000 is
the fourth. A direct comparison of these groups shows that the density of vegetation
cover in 1987 was significantly higher than that in 2000, especially the much higher
ANDVI 1987 value (>0.19) compared to that in 2000.
The area with the ANDVI value > 0.19 in 1987 and 2000 can be used to support a
focused analysis (see Figure 7.11), indicating that the selected area is tied to the sec7.11), indicating that the selected area is tied to the sec.11), indicating that the selected area is tied to the sec11), indicating that the selected area is tied to the sec1), indicating that the selected area is tied to the sec), indicating that the selected area is tied to the second turning point that is probably tied to UHI (Figure 7.10). This area was covered
mainly by farmland (crop), some scattered rural settlements, and sporadic urban
built-up land, accounting for 19.2% LULC of the total study area in 2000. The area
with ANDVI values between 0.04 and 0.19 accounted for the additional 72.72% of
the study area. Together, both types of LULC accounted for 91.92% of the total study
area. In addition, the area with ANDVI values between 0 and 0.03 accounted for
5.91% of the total study area in 2000 compared to 0.36% in 1987. Collectively, these
scattered rural settlements and urban land contributed to the overall LST directly
and UHI indirectly in 2000.