Presentation in the 2019 2nd International Conference on Civil Engineering and Architecture on September 21-23, 2019, Seoul National University, South Korea
Assessment of Urban Green Space Structures and Its Effect on Land Surface Temperature in Chiang Mai City Area, Thailand
1. Paper ID: CEA048
2019 2nd International Conference on Civil Engineering and Architecture
September 21-23, 2019, Seoul National University,South Korea
Manat Srivanit
Livable Environment & Architectural Design Laboratory (LEAD-Lab)
Faculty of Architecture and Planning, Thammasat University
* Corresponding author, E-mail address: s.manat@gmail.com
2. CONTENTS
1. Introduction
2. Research question
3. Objectives
4. Research methodology
Quantifying Green Spaces and Land Surface
Temperature (LST) Retrieval
Quantifying the Effects of Urban Green
Patterns on LST
5. Results
6. Discussion and conclusion
7. Further research is needed
2
3. 1. Introduction
Therefore, understanding the influences of
greenspace on temperature reduction is of
great interest in examining the relationship
between greenspace structure and cooling
effectiveness at the patch-level.
The following three major landscape metrics
were used to examine landscape structures of
individual green patches including;
1) Size
2) Shape
3) Core (Kennedy et al., 2003)
Greenspace can form an “Urban Cool Island (UCI)” within an urban area through
evaporative cooling and can directly influence an urban microclimate by reducing surface
and local air temperatures, which in turn affects localized cooling (Chang et al. 2007;
Oliveira et al. 2011; Feyisa et al. 2014). “The same size but not necessarily
the same shape!!”
3
4. The Chiang Mai is the largest city in northern Thailand, rapid
urbanization and urban sprawl extending into several neighboring
districts has resulted in a remarkable surface urban heat island
(SUHI) effect with mean temperatures increasing by 6.22 °C.
(Srivanit and Auttarat, 2016)
4
(b) Land surface temperature (LST)
(a) Urban form
Impacts of urbanization on the urban thermal environment in Chiang Mai City
0
5
10
15
20
25
0
10
20
30
40
50
60
70
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
hours,days
Temperature(°C)
Month
Climate data for Chiang Mai (1981–2010)
Average rainy days Average high (°C)
Daily mean (°C) Mean daily sunshine hours
The temperature difference can be
as much as 12-15°C
5. A B
Chiang Mai Airport
Old Town
Old CBD
Highway
Hypermarket
Wat Gate
district
Wualai
district
LST 2006LST 2006LST 2000LST 2000
Distance (meters)
76600
34.5924.96
500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
29.96
A B
SurfaceTemperature
Centigrade)
LST2000
LST2006
Distance (meters)
76600
34.5924.96
500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
29.96
A B
SurfaceTemperature
Centigrade)
Distance (meters)
76600
34.5924.96
500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
29.96
A B
SurfaceTemperature
Centigrade)
0.223-0.426
NDVI/NDWI
LST
NDVI
NDWI
Remark: LST (land surface temperature), NDVI (an index of greenness), NDWI (index of water content in water bodies)
Night Bazaar
Averaged Surface Temperature Profile of Chiang Mai City Center (Srivanit M., 2010)
5
Golf course
Large green
patch
6. Climate
information for
improved
urban
landscape
planning
Surface
urban heat
island (SUHI)
Estimating
cooling effect
(+) increase or
(-) reduce the
temperature
Landscape
metrics
(Size, Shape,
Core)
Landscape
structure of
greenspace
3. Objectives
1) to identify the characteristics of greenspace and corresponding
land surface temperatures, and
2) to determine the cooling potential of greenspace patterns and
its relationships with the effectiveness of greenspaces in the Chiang
Mai city area.
2. Research question
6
Define the task Analysis and synthesis of data Interpret and report
“How the landscape structure of green
space in urban areas can efficiently reduce
the urban temperature?”
7. 4. Research Methodology
Step I: The data were obtained from Landsat-8 images taken on 8 April 2015, is the hottest
month of the year (TMD 2016);
We used the visible wavelength red band (Band 4) to the near-infrared band (Band 5) to
calculate the Normalized Difference Vegetation Index (NDVI), and were used to identify
green patches.
The satellite collects thermal imaging data from a thermal infrared sensor (TIRS) was
provided by band 10 used to retrieve Land Surface Temperatures (LSTs).
Step II: The landscape metrics were employed using Patch Analyst software developed by
Rempel et al. (2012) embedded into the ArcGIS 10.0 platform.
7
4.1 The area of study
Centrallatitude/latitude located at 18° 47′46.1148′′N, 98° 58′45.3468′′E
8. 4.2 Research
Framework
Landsat-8 images taken
on April 8th, 2015
Red band (Band 4)
Near-infrared band (Band 5)
Thermal infrared band
NDVI
(ranged from +1 to -1)
LST (Celsius)
NDVI>0.5
Selected green patches (N=3,416)
Yes
No
LST<mean LST
Urban cool island (UCI)
Yes
NoUrban heat island (UHI)
Patch Analyst Tool
(Landscape metrics calculations)
Landscape structure of green patches
If a UCI in an area contained
over 60% of green patch
Greenspace cool islands (GCIs) (N=1,155)
Yes
No
Not GCIs (N=2,261)The area of UCI
Mean temperature reduction
Pearson rho’s
correlation coefficient
(i) to examine the bivariate relationship
between pattern metrics of all green patches
Stepwise multiple
Regression analyses
(ii) to determine the best metric
indicators for UCIs as the cool island effect
Not green patch
8
Aspect of pattern
I. Size
II. Shape
III. Core
※ Kong et al., 2014; Zhang et al., 2017
Step 2: Identify greenspace cool islandsStep 1: Defining of urban cool islands
Step 3: Quantifying the cooling
potential of greenspace
patterns and its relationships
※ Sun and Chen, 2017; Zhang et al., 2017
9. Table 1. Definitions and equations of landscape metrics for measuring greenspace
patterns at the patch level
Patterns Landscape metrics
Abbrevia
tion
Description Units Equations (Range)
Size Patch area PA The green patch area km2 (PA > 0, no limit)
Edge length TE The total length of edge
segments of a green patch
km (TE > 0, no limit)
Shape Shape index SI Normalized ratio of the patch
perimeter to area
unitless SI=P/(2√ PA); (SI ≥1, no limit)
Perimeter-area ratio PAR Simple ratio of the patch
perimeter to area where
patch shape is confounded
with patch size
unitless PAR=P/PA; (PAR>0, no limit)
Fractal dimension FD Another normalized shape
index based on perimeter-
area relationships for which
the perimeter and area are
log transformed
unitless FD=2ln(0.25P)/ln(PA); (1≤FD≤2)
Core Core area CA The area of the patch in the
core area
km2 (CA > 0, no limit)
Core index CI The percentage of the patch
within the core area
percent CI=(CA/PA)x100; (0≤CI≤100)
Note: P denotes the perimeter of a green patch. 9
10. 10
Fig. 1 Spatial distribution of (a) the NDVI and (b) green patches within the study area
derived from a Landsat-8 image
11. Urban zones Count Size Shape Core
PA (km2) TE (km) SI PAR FD CA (km2) CI (%)
Core zone 39 0.003 0.205 1.217 0.105 1.390 n.a. n.a.
Inner zone 670 0.008 0.373 1.313 0.124 1.423 0.0005 0.337
Middle zone 1,146 0.012 0.476 1.340 0.126 1.423 0.0011 0.453
Outer zone 1,561 0.017 0.626 1.372 0.126 1.426 0.0015 0.448
Study area 3,416 0.013 0.521 1.348 0.125 1.424 0.0012 0.423
Note: n.a. is not available
Table 2. The distribution of the number and average landscape characteristics of greenspace
5. Results
We identified 3,416 greenspace patches covering an area of 44.62 km2 (or roughly 28.97%)
within the study area.
More than half of all greenspace areas (57.98%) were found in the outer zone consistent
with most landscape metrics revealing dense greenspace patterns, indicating a stronger effect
on UHI reduction than what is observed in other urban zones.
5.1) Overall characteristics of greenspace and corresponding LSTs
11
12. 12
Fig. 2 Spatial distribution of landscape metric values calculated for urban
greenspaces at the patch level.
13. 13
5.2) Characteristics of greenspace cool islands and their relationship to cooling effectiveness
Fig. 3 (a) Distribution of land surface temperatures (LSTs) retrieved from Landsat-8 thermal infrared
sensor (TIRS) data used to identify (b) urban cool islands (UCIs) in the study area
14. 14
Fig. 4 (a) Greenspace cool islands (GCIs) and (b) differentiation in the mean cooling
potential at the patch level
15. Urban zones
Greenspace cool islands (GCIs) (N=1,155) Not GCIs (N=2,261)
Count %
Mean
LST
Mean
cooling
potential
Count %
Mean
LST
Mean
cooling
potential
Core zone 1 0.09 32.36 0.37 38 1.68 34.79 2.41
Inner zone 100 8.66 32.10 -0.07 570 25.21 34.06 1.66
Middle zone 324 28.05 31.79 -0.34 822 36.36 33.67 1.36
Outer zone 730 63.20 31.33 -0.69 831 36.75 33.34 1.10
Remark: Cooling potential is defined as the reduction in temperature achieved relative to the mean LST for the study area
Table 3. Mean LST results for different types of greenspace and corresponding cooling effects (Unit: °C)
1,155 GCIs (or roughly 33.81% of the greenspace patches studied) covering a mean patch
area of 23.14 km2 were found in the study area. The mean GCI cooling potential for the study
area was recorded as 1.03 °C.
Approximately 63.20% of GCIs in the study area (with a mean patch area of 16.07 km2) are
located in peripheral urban zones. Only one GCI is found in the core urban zone. The GCI
includes the Buak Hard Public Park positioned at the southwestern corner of the core urban zone. 15
16. 16
Fig. 5 Boxplot showing the mean LST values of greenspaces for
different urban zones in the study area
(a) (b)
17. Table 4. Pearson correlation between the mean LST and landscape metrics of greenspace
All GCI pattern metrics were found to be significantly correlated with surface
temperature reduction for GCIs of all urban zones.
Negative correlations with GCI size and core metrics are especially strong while shape
metrics for SI, PAR, and FD correlate significantly with decreasing cooling potential in
outer urban zones.
These results indicate that the cooling potential of GCIs serves as configuration of the
spatial patterns of greenspace. However, fragmented greenspaces also provide cooling
potential in different urban zones.
17
PA TE SI PAR FD CA CI
Core zone -0.408** -0.404* -0.306 0.363* 0.361 n.a. n.a.
Inner zone -0.212** -0.217** -0.271** -0.061 -0.082* -0.082** -0.180**
Middle zone -0.186** -0.210** -0.198** -0.057 -0.072* -0.140** -0.195**
Outer zone -0.142** -0.146** -0.119** -0.119 -0.030 -0.115** -0.142**
Study area -0.162** -0.173** -0.174** -0.037* -0.050** -0.124** -0.157**
Note: **p<0.01, *p<0.05.
18. 18
Fig. 6 (a) Location of GCIs of the study area displayed as false-color composites and (b)
patterns of UCIs at the patch level
19. 19
Fig. 7 (a) Spatial distribution of cooling effects of GCIs with (b) examples that identify
effective and ineffective GCIs
20. 20
Urban zones
(The number of samples)
The number of predictors
in the best fit model
(Adjusted R2)
Metrics entered
(or Independent
variables)
Standardized
coefficients
(Beta)
Core zone (N=1) n.a n.a. n.a.
Inner zone (N=100) 2 (0.107) TE 0.070*
SI 0.279*
Middle zone (N=324) 2 (0.053) SI 0.374**
PA -1.174*
Outer zone (N=730) 2 (0.014) SI 0.211**
CA -4.115*
Study area (N=1,155) 3 (0.027) SI 0.262**
CA -4.406**
FD 0.117*
Note: **p<0.01, *p<0.05.
Table 5. Stepwise multiple regression of cooling potential and pattern metrics of GCIs (N = 1155)
Where: Total Edge (TE), Shape Index (SI), Patch Area (PA), Core Area (CA), and Fractal Dimension (FD)
21. 6. Discussion and Conclusion
21
We found patterns of shape and cores, including Shape Index (SI), Core
Area (CA), and Fractal Dimension (FD) metrics, to be valuable in making
urban cool island (UCI) predictions of GCIs in the study area.
Our results show that Patch Area (PA) and Core Area (CA) metrics are
correlated negatively with greenspace cool islands (GCIs), providing that
the presence of more greenspace cools GCIs, echoing the findings of other
patch-level studies (Chen et al. 2014a, b; Gioia et al. 2014; Feyisa et al.
2014).
Our analysis revealed that size of GCI patches has a significantly negative
relationship with the mean cooling potential, suggesting the increase of
patch size of greenspace may further lower the temperatures, and this
suggestion has also been confirmed in previous studies (Chang et al. 2007; Li
et al. 2013; Kong et al. 2014).
22. 7. Further research is needed
In this study, we did not consider greenspace structures that significantly affect
UCI effects (Hamada and Ohta 2010; Lin et al. 2015; Akbari and Kolokotsa 2016; Wang
and Akbari 2016), including;
o Types of greenspace,
o Species compositions and distributions,
o Green coverage rates,
o Horizontal and vertical structures of vegetation,
o Cooling extents, and
o Seasonal variations in green area
Furthermore, as previous studies on surface UCIs of urban greenspaces have focused
on the cooling effects of water bodies (Sun et al. 2012; Kong et al. 2014; Du et al.
2016), more work determining how “Water Cooling Islands (WCIs)” mitigate UHI
effects and especially UCI effects may be needed to further investigate WCI impacts.
22
23. END
Thank You
For Your Attention !
Any Questions?
2019 2nd International Conference on Civil Engineering and Architecture
September 21-23, 2019, Seoul National University, South Korea
Acknowledgements Special thanks are due to the U.S.
Geological Survey (USGS) for sharing Landsat-8 satellite
images. This study was supported by Thammasat University
Research Fund, Contact No.23/2553.