Content uploaded by Alfredo Mahar Lagmay
Author content
All content in this area was uploaded by Alfredo Mahar Lagmay on Sep 23, 2014
Content may be subject to copyright.
Downloaded By: [Stanford University] At: 17:47 4 April 2008
ASTER-based study of the night-time urban heat island effect in Metro
Manila
M. TIANGCO*{, A. M. F. LAGMAY{ and J. ARGETE{
{Institute of Environmental Science and Meteorology, University of the Philippines,
Diliman, 1101 Quezon City, Philippines
{National Institute of Geological Sciences, University of the Philippines, Diliman, 1101
Quezon City, Philippines
(Received 18 July 2006; in final form 18 April 2007 )
The Advanced Spaceborne Thermal Emission and Reflection Radiometer
(ASTER) was used to derive land surface temperatures to quantify the night-
time urban heat island (UHI) effect in Metro Manila. Temperature differences
between Metro Manila and its adjacent rural towns were compared to determine
heat island intensity and analyse spatial variation of surface temperature.
Transects were drawn across from the rural to the urban region to characterize
the UHI profile and the Normalized Difference Vegetation Index (NDVI) was
used to examine the relationship between amount of vegetation and temperature.
The thermal images revealed the highest UHI intensity to be 2.96uC with the
presence of a heat island existing in the central part of the city. The transects
described the cross-sectional heat island profile characterized by gradients of
‘cliffs’, ‘plateaus’ and a ‘peak’ occurring in the city centre. The study also showed
an inverse relationship between NDVI and temperature, which suggests that
increasing the amount of plants in cities can reduce the UHI effect.
1. Introduction
The urban heat island (UHI) effect is the observed higher temperature that cities or
urban areas experience compared to the surrounding rural areas. It is caused by
urbanization, a process of land cover change wherein natural land and vegetation
are replaced by built surface materials such as asphalt, cement, brick and stones.
These materials absorb and store a significant amount of solar heat during the day,
resulting in the subsequent slow release of heat at night. In the absence of
vegetation, evapotranspiration, which has a cooling effect on the Earth’s surface, is
reduced. The presence of buildings in cities not only prevents the surface heat from
escaping into the upper atmosphere but also causes friction, hampering heat loss
through advection or movement of cool air from rural areas. Increased human
activity results in a wide range of heat sources (power generation, the use of
motorized vehicles, air conditioners, etc.) that also contribute to the excess heat of
the atmosphere. These activities also produce pollutants that warm the city by
enhancing the greenhouse effect. All these factors result in higher temperatures in
cities compared to the surrounding rural areas.
*Corresponding author. Email: mgtiangco@up.edu.ph
INT. J. REMOTE SENSING
2008, iFirst Article, 1–20
International Journal of Remote Sensing
ISSN 0143-1161 print/ISSN 1366-5901 online # 2008 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/01431160701408360
Downloaded By: [Stanford University] At: 17:47 4 April 2008
The UHI effect is considered the best example among the manifestations of the
impact of human activity on local climate (Ja¨ger 1983, Hinkel et al. 2003). In this
study, satellite remote sensin g with the Advanced Spaceborne Thermal Emission
and Reflection Radiometer (ASTER) was used to examine the effect of urbanization
on the micr oclimate of Metro Manila. Previous investigations on the UHI effect
through remote sensing have been conducted by Gallo et al. (1993), Lo et al. (1997),
Streutker (2002, 2003), Dousset and Gourmelon (2003), Fukui (2003), Lo and
Quattrochi (2003), Weng (2003), Weng and Yang (2004), and Weng et al. (2004), to
name a few. This paper quantifies and documents the degree of the night-time UHI
phenomenon in the study area and analyses the spatial variation of its surface
temperature. The UHI profile is also characterized and the cooling effect of
vegetation determined by examining the relationship between the Normalized
Difference Vegetation Index (NDVI) and temperature.
The inverse relationship of vegetation and temperature is widely documented
in remote sensing literature (Gallo et al. 1993, Lambin and Ehrlich 1996, Weng
2001, Dousset and Gourmelon 2003, Lo and Quattrochi 2003). In the study of
the UHI of Tel-Aviv, Israel, Saaroni et al. (2000) concluded that parks, open spaces
and other vegetated areas appear relatively colder compared to other urban
components. In Kumamoto City, Saito et al. (1990/91) showed that 30 m660 m
small green spaces could be 3uC cooler than their surrounding built-up areas in
cities. Yang and Wang (1989) studied the effect of the g reening and tree planting
project of Guangzhou City in South China and concluded that in terms of daily
mean temperatures, daily high temperatures and duration of temperatures >
30uC, vegetated areas consistently generated cooler air temperatures than non-
vegetated streets. Moreover, the study demonstrated a decrease of 0.9uC and 0.5uC
in the daily mean temperature of afforested streets and residential areas, res-
pectively, and found that trees in the parks can lower under-canopy temperatures by
2.1uC.
Saaroni et al. (2004) described this microclimatic change in vegetated or green
areas as the ‘oasis effect’, which is exhibited by cooler temperatures and higher
relative humidity as a result of evapotranspiration (Oke 1987). In cities,
evapotranspiration rates are decreased by urbanization (Owen et al. 1998), which
then leads to lower humidity. Cities therefore have less cooling effe ct from the
reduced evaporative process (Yannas 2001).
2. Study site
Metro Manila (figure 1) is the capital, and the largest city, of the Philippines.
Located at 14u359 N, 121u E, it covers an area of 636 km
2
and is bounded on the west
by Manila Bay and on the southeast by Laguna Lake.
Based on the census conducted in 2000, the population of Metro Manila was
estimated at 9 932 560 (NSCB 2005). Migration to urban areas is identified as the
major cause of population increase in Metro Manila, which brought about half the
city’s growth between 1970 and 1980 (Icamina 1995).
The Philippines is a tropical country characterized by two seasons: the rainy
season that occurs from June to November and the dry season from December to
May. According to the Philippine Atmospheric, Geophysical and Astronomical
Services Administration (PAGASA), it has a mean annual temperature of 26.6uC
with the coolest month in January and the warmest in May. The average
temperatures for January and May are 25.5uC and 28.3uC, respectively.
2 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
The geographic configuration of Metro Manila should benefit from the regulating
effect of the two bodies of water adjoining it, which generate sea and land breezes.
During the day, the land heats up faster than the water, inducing the warm air ov er
it to rise and create a pressure gradient. This process draws cooler air from the sea to
flow inland, widely known as a sea breeze. At night-time, the land becomes cooler
than the water. Thus, air circulati on is reversed, causing a land breeze, with air
flowing from the land towards the sea. A land breeze is much weaker than a sea
breeze. However, because of the nocturnal UHI effect, especially found in the
equatorial tropics, the temperature over the land does not cool much during the
night, making the temperature of the land similar to that over the sea or lake,
Figure 1. Metro Manila study area and the rural towns of Bulacan and Cavite/Laguna with
the 200-km
2
circular regions of interest sampled for surface temperature and the four transects
drawn over to characterize the UHI profile. The centres of the rural towns of Bulacan and
Cavite are at 25 km and 30 km, respectively, away from the centre of Metro Manila while the
transects each cover approximately 60–70 km.
The night-time UHI effect in Metro Manila 3
Downloaded By: [Stanford University] At: 17:47 4 April 2008
further weakening any night-time land breeze that would otherwise cool the city
(Emmanuel 1993).
3. Background on ASTER, data acquisition and image processing
ASTER is an optical imager or sensor with a high spatial resolution and multi-
spectral capability (Yamaguchi et al. 1998). The instrument is on board Terra,
the initiating satellite of the Nation al Aeronautics and Space Admi nistration’s
(NASA) Earth Observing System (EOS) Project launched on 18 December 1999
(ERSDAC 2003). ASTER has a five-band thermal infrared (TIR) radiometer
subsystem that records wavelength regions covering 8.125–8.475, 8.475–8.825,
8.925–9.275, 10.25–10.95 and 10.95–11.65 mm, respectively, at 90 m resolution
(Yamaguchi et al. 1998).
ASTER surface kinetic temperature data were used in this study. The image s
are the Level 2 standard data products containing surface temperatures in Kelvin
(610 units), which have been corrected for atmospheric transmission, absorption
and path radiance (Gillespie et al. 1999). The product is also corrected for
emissivity, the ratio of thermal radiation emitted by a surface to that of a blackbody
at the same temperature, a value important in determining the Earth’s surface
temperature with variable land cover (Becker and Li 1990a,b, Kealy and Gabell
1990, Kahle and Alley 1992, Watson 1992a,b, Kealy and Hook 1993, Prata et al.
1995, Schmugge et al. 1998, Snyder et al. 1998). ASTER surface kinetic tempera-
ture data products, which have a published absolute accuracy of 1–4uC (Jet
Propulsion Laboratory 2001, Sobrino et al. 2007), have their corresponding land
cover emissivities estimated using the temperature-emissivity separation (TES)
algorithm, which was developed by the ASTER science team (Gillespie et al. 1998,
1999).
Emissivities for the five TIR bands of the ASTER are initially extracted from the
land-leaving TIR radiance, L 9, through the normalized emissivity method (NEM).
These emissivities, used with Kirchoff’s law, are used to account for reflected sky
irradiance. This value is subtracted from L9 iteratively to estimate emitted radiance,
R, from which an initial temperature, T
i
, is calculated. T
i
and R allow for the
computation of normalized emissivities through the spectral ratio algorithm. The
min-max difference (MMD) of the normalized emissivity spectrum is then calculated
and minimum emissivity is derived through regression analysis that relates MMD
and minimum emissivity. The minimum emis sivity is used to scale the normalized
emissivities. Together with R, the refined emissivities are used to recalculate the
‘actual’ surface temperature using the Planck equation (Gillespie et al. 1998, 1999,
Jet Propulsion Laboratory 2001).
Data product images were ordered and downloaded from the internet via the EOS
Data Gateway (EDG) webs ite. To determine UHI for Metro Manila, seven night-
time ASTER thermal images wer e used. The images were acquired between 22:14
and 22:22 local time. A separate set of 73 night-time thermal images, which include
areas beyond Metro Manila, was also acquired to validate ASTER-derived surface
temperature with air temperature from PAGASA.
To remove the influence of clouds in the data analysis, the quality assurance (QA)
planes provided with each image were used to mask the clouds. QA planes are pixel -
based graphical displays of bad data that show por tions of the image with poor
quality data due to cloud cover. Pixel areas that are of poor quality as defined by the
QA data were then app lied as a mask to the temperature image by assigning a value
4 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
of 0, thereby removing portions of bad data in the imagery. The images were also
converted to degrees centigrade.
4. Methods
4.1 Validating ASTER-derived surface temperature
The set of 7 3 night-time thermal images acquired to validate surface temperatures
derived from ASTER were used together with corresponding air temperature
measurements from 14 of PAGASA’s synoptic weather stations. This was to ensure
a statistically valid analysis for the accuracy of the surface kinetic temperature
product of ASTER. The tally of the temperature measurements for each station is
given in table 1. The time when PAGA SA air temperatures were measured was the
nearest available record to the time of the satellite pass (approximately within
45 min). ASTER surface temperature was derived from the pixel that corresponds to
the geographic coordinates of a particular weather station as given by PAGASA.
Rainfall data 5 days before the recorded satellite scene dates were also obtained.
4.2 Deriving urban and rural surface temperatures
Figure 1 shows the 200-km
2
circular regions of interest drawn over Metro Manila,
Bulacan and Cavite to represent the urban and rural areas of the study sites and to be
sampled for their surface temperatures. Bulacan and Cavite are the rural towns north
and southwest of Metro Manila, respectively. The centre of the circular areas drawn
is about 25 km (for Bulacan) and 30 km (for Cavite) away from Metro Manila.
UHI intensity was estimated by computing the difference between the average
surface temperatures of the circles over Metro Manila and Bulacan/Cavite. The
absolute urban–rural temperature difference (UHI
max
) was also computed. The
output values represented the difference between the highest urban temperature and
lowest rural temperature.
Table 1. Tally of temperature measurements for each PAGASA
station to validate ASTER-derived temperature.
PAGASA station
No. of temperature
measurements
Science Garden, Metro Manila 9
Port Area, Metro Manila 9
NAIA, Metro Manila 9
ADMU, Metro Manila* 8
Dagupan City, Pangasinan 3
Cabanatuan City, Nueva Ecija 1
Iba, Zambales 3
Clark Airport, Pampanga 6
Sangley Point, Cavite 10
Ambulong, Batangas 7
Tayabas, Quezon 2
Alabat, Quezon 1
Calapan, Oriental Mindoro 1
Legazpi City, Albay 4
Total 73
*Not a PAGASA station.
The night-time UHI effect in Metro Manila 5
Downloaded By: [Stanford University] At: 17:47 4 April 2008
4.3 Characterizing the UHI profile
Four transects, each approximately 60–70 km in length, were drawn across each
image (figure 1) to characterize the UHI profile. Each transect starts from the rural
town of Bulacan in the north, passing through the city, down to the opposite
southwestern and southern rural towns of Cavite and Laguna, respectively.
4.4 Correlating NDVI and temperature
The NDVI was derived from an ASTER Surface Reflectance–VNIR (visible and
near-infrared) daytime image acquired on 10 November 2001, 10:39 local time. The
ASTER surface reflectance data product used was already corrected for influences
of satellite-sun geometry and the effects of atmospheric conditions (Jet Propulsion
Laboratory 2001). NDVI was computed through the formula NDVI5(B32B2)/
(B3 + B2), where B3 is the near-infrared band of ASTER and B2 is the visible band.
High values of NDVI suggest the presence or abundance of vegetation while lower
values suggest poor vegetative cover. The formula is based on the knowledge that
visible light is absorbed by the leaves of plants while near-infrared light is reflected.
Healthy, or a large amount of, vegetation absorbs more visible light and reflects a
large amount of near-infrared radiation, thereby producing high NDVI values.
Conversely, sparse or unhealthy vegetation reflects more visible light and less near-
infrared radiation. The NDVI and temperature image were overlain and examined
by linear regression analysis.
5. Results and discussion
5.1 ASTER surface temperature validation
A root mean square error (RMSE) of 2.87 was computed for the night-time
measurements between ASTER surface temperature and PAGASA air temperature.
Pearson correlation analysis yielded a coefficient (r) of 0.74 with p,0.0001 while
linear regression analysis (figure 2) showed a positive relationship with a coefficient
of determination (R
2
) of 0.55. The statistical analysis indicates that 55% of the
variability of air temperatur e can be explained by the changes in surface temperature
during the night.
Previous studies suggest a positive correlation between surface and air
temperature (Stoll and Brazel 1992, Nichol 1996, Ben-Dor and Saaroni 1997,
Nichol 1998). Although surface temperature influences air temperature, the
relationship is governed by several factors such as atmospheric conditions (Ben-
Dor and Saaroni 1997), processes and surface properties (Voogt and Oke 2003), and
is highly complicated (Saaroni et al. 2004) and not exact (Arnfield 2003). Detailed
models of surface and atmospheric processes are necessary to understand the
relationship (Voogt and Oke 2003). The improvement and development of
algorithms and models to more accurately derive surface temperature from
satellites, as well as to estimate air temperature from surface temperatures and
establish a direct surface–air temperature relationship, are currently being actively
researched.
In this study, the RMSE of the measurements between ASTE R and PAGASA
temperatures is within the publis hed absolute accuracy of the ASTER thermal
product. It should be noted that ASTER measures the average of a 90 m690 m area
per pixel while PAGASA measures the temperature of a single point in space. The
6 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
RMSE or the temperature difference between the ASTER surface temperature and
PAGASA air data (measured 1.5 m above the ground) are in good agreement with
each other. Similar correlation of surface temperature da ta with in situ ambient air
temperature measur ements have been demonstrated by Takemata et al. (2004) for
night-time ASTER (2B) imagery and by Stathopoulou et al. (2004) for National
Oceanic and Atmospheric Administration/Advance Very High Resolution
Radiometer (NOAA/AVHRR) images.
5.2 UHI intensity
The spatial variation of surface temperature for the 4 May 2002 ASTER scene is
presented in figure 3. It shows the central part of Metro Manila, which is around the
major highway EDSA, exhibi ting a circular heat island. Table 2 presents the urban–
rural temperature statistics of the seven thermal images used in this study of UHI.
The mean, minimum and maximum values are within the circles drawn over Metro
Manila, Bulacan and Cavite as illustrated in figure 1.
Mean UHI intensity was computed as the difference between the mean
temperatures of the representative areas of Metro Manila and Bul acan/Cavite
(UHI5T
mean-urban
2T
mean-rural
). Based on the ASTER sample data set, the 4 May
2002 scene resulted to the highest mean UHI intensity of 2.96uC in Metro Manila.
The date was during su mmer and it was also the time when Metro Manila
experienced the greatest temperature variation of 13.9uC. The temperature ranged
from a minimum of 21.6uC to a maxi mum of 35.5uC, the highest surface
temperature also recorded for the given data set. The maximum UHI intensit y,
which is the difference between the maximum temperature in Metro Manila and the
minimum temperature in either Bulacan or Cavite (UHI
max
5T
max-urban
2T
min-rural
)
Figure 2. Night-time scatter plot of surface vs. air temperature.
The night-time UHI effect in Metro Manila 7
Downloaded By: [Stanford University] At: 17:47 4 April 2008
and which shows the highest degree of urban heating, was 14.3uC for the same
satellite scene date.
The lowest UHI intensity was 0.87uC recorded from the 7 December 2002 satellite
scene. Although this decrease in UHI intensity can be explained by the fact that the
measurement was during the cold season, there was a recorded 9 mm of rainfall at
the Port Area station of PAGASA on 6 Decem ber 2002 (table 3), the day before the
Figure 3. Night-time colour-coded thermal image of 4 May 2002 ASTER scene showing the
spatial pattern of surface temperature throughout the study areas. The commercial and
business district of Metro Manila clearly exhibits a circular heat island. Water bodies have
been masked out.
8 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
satellite scene was acquired. Rainfall may have had a cooling effect on the
temperature in Metro Manila and might explain this result. There were also trace
precipitation or a very light rain shower and another rainfall occurrence on 3
December and 2 December 2002, respectively. By contrast, it was on the 7 December
2002 image scene that UHI
max
was highest at 16.3uC. On this date, the maximum
temperature reached at Metro Manila was 33.3uC while the minimum was 20.3uC,
that is a range of 13uC.
The mean UHI intensities for 24 January 2003, 31 January 2003, 16 February
2003, 22 March 2004 and 29 January 2005 were 0.97, 0.96, 2.15, 1.79 and 1.23uC,
Table 2. Surface temperature statistics of the 200-km
2
study areas (uC).
Scene date
(time)/area Mean Min Max Range SD UHI* UHI
max
{
4 May 2002 (22:22)
Bulacan 26.11 21.20 36.30 15.10 1.03
Metro
Manila
30.22 21.60 35.50 13.90 1.27 2.96 14.30
Cavite 28.41 23.40 33.80 10.40 1.37
7 December 2002 (22:16)
Bulacan 24.81 19.80 34.50 14.70 1.65
Metro
Manila
26.27 20.30 33.30 13.00 1.66 0.87 16.30
Cavite 25.98 17.00 32.50 15.50 0.98
24 January 2003 (22:16)
Bulacan 24.13 19.60 29.40 9.80 1.40
Metro
Manila
25.48 19.20 30.00 10.80 1.03 0.97 10.40
Cavite 24.88 19.60 31.30 11.70 0.66
31 January 2003 (22:22)
Bulacan{ –––––
Metro
Manila
23.09 20.10 28.30 8.20 1.34 0.96 13.00
Cavite 22.13 15.30 26.10 10.80 1.03
16 February 2003 (22:22)
Bulacan 21.65 19.70 32.80 13.10 1.03
Metro
Manila
24.69 17.70 30.50 12.80 1.56 2.15 11.90
Cavite 23.42 18.60 27.20 8.60 1.03
22 March 2004 (22:21)
Bulacan 24.81 22.40 40.80 18.40 1.05
Metro
Manila
26.70 23.60 31.50 7.90 1.13 1.79 12.00
Cavite 25.01 19.50 44.10 24.60 0.99
29 January 2005 (22:14)
Bulacan 25.51 20.00 37.60 17.60 1.26
Metro
Manila
27.16 20.20 32.60 12.40 1.52 1.23 12.60
Cavite 26.34 21.60 31.40 9.80 0.96
*UHI5T
mean-urban
2T
mean-rural
.
{UHI
max
5T
max-urban
2T
min-rural
.
{Data not available for Bulacan because of cloud cover.
The night-time UHI effect in Metro Manila 9
Downloaded By: [Stanford University] At: 17:47 4 April 2008
respectively. The corresponding values for UHI
max
were 10.4, 13, 11.9, 12 and
12.6uC, respectively. There was also an occurrence of rain on 11 February 2003.
UHI intensity for the 16 February 2003 scene did not, however, show signs of
Table 3. Rainfall data prior to acquisition dates of ASTER images (mm).
Date
Station
UHI (uC)Science Garden Port Area
4 May 2002 0 0 2.96
3 May 2002 0 0
2 May 2002 0 0
1 May 2002 0 0
30 April 2002 0 0
29 April 2002 0 0
7 December 2002 0 0 0.87
6 December 2002 0 9.0 (06:00)
5 December 2002 0 0
4 December 2002 0 0
3 December 2002 Trace (06:00) 0
2 December 2002 0.8 (06:00) 0.4 (12:00)
24 January 2003 0 0 0.97
23 January 2003 0 0
22 January 2003 0 0
21 January 2003 0 0
20 January 2003 0 0
19 January 2003 0 0
31 January 2003 0 0 0.96
30 January 2003 0 0
29 January 2003 0 0
28 January 2003 0 0
27 January 2003 0 0
26 January 2003 0 0
16 February 2003 0 0 2.15
15 February 2003 0 0
14 February 2003 0 0
13 February 2003 0 0
12 February 2003 0 0
11 February 2003 3.8 (12:00) 7.9*
22 March 2004 –{ 0 1.79
21 March 2004 – 0
20 March 2004 – 0
19 March 2004 – 0
18 March 2004 – 0
17 March 2004 – 0
29 January 2005 0 0 1.23
28 January 2005 0 0
27 January 2005 0 0
26 January 2005 0 0
25 January 2005 0 0
24 January 2005 0 0
*PAGASA data on time of rainfall missing.
{PAGASA data missing.
10 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
cooling. In fact, UHI intensity for that date was the next highest among the sample
data set. The rainfall may not have had a significant effect as it oc curred 5 days
before the satellite scene date.
The highest temperatur es for the rural towns were recorded on 22 March 2004.
Bulacan reached its highest temperature of 40.8uC and Cavite reached 44.1uC, both
exceeding the 31.5uC maximum temperature at Metro Manila on that day.
The coldest temperature recorded was 15.3uC in Cavite on 31 January 2003. The
lowest temperature that Metro Manila reached was 17.7uC on 16 February 2003.
For Bulacan, the lowest temperature was 19.6uC on 24 January 2003.
The data show that the maximum temperatures recorded in Bulacan were higher
than the maximum temperatures in Metro Manila except for 24 January 2003. There
were also scene dates when the maximum temperature in Cavite exceeded that of
Metro Manila (24 January 2003 and 22 March 2004). The case of Bulacan may be
considered an isolated one because the pixels of these maximum temperatures poi nt
to the same geographic location in four of the satellite scenes available while the rest
are just within one vicinity. This means that the spot consistently experiences a high
temperature. This area is thought to correspond to part of a cement manufacturing
plant in Bulacan.
5.2.1 Elevation and temperature. Altitude influences temperature measurements
because temperature within the troposphere decreases with height. To look into this
effect, the topography of the study area was examined through a Shuttle Radar
Topographic Mission (SRTM) Digital Elevation Model (DEM). The elevation data
showed that the minimum and maximum heights of the study area have a difference
of 189 m. Metro Manila is lowest along the coast and has an average altitude of
23.5 m. Bulacan and Cavite have maximum altitudes of 189 and 143 m, respectively.
Considering that the temperature lapse rate decreases by 6.5 uC for every km increase
in elevation (Ahrens 1998), the entire study area would have a maximum difference
in temperature of 1.23uC due to altitude changes. On this basis, it was assumed that
elevation had a relatively minimal influence on surface temperature measurements
for this study.
5.3 Spatial variation of surface temperature
The spatial variation of surface temperature in Metro Manila follows the UHI
description that city centres exhibit the highest temperature within an urban area.
This is clearly shown in figure 3, where the area around the circumferential highway,
EDSA, is displ ayed as a heat island. In this area the majority of commercial and
business establishments are concentrated. The figure also shows that the
temperature starts to decrease away from the urban core and continues to do so
up to the surrounding rural towns. Within this major heat island are ‘hot spots’ that
can also be identi fied in figure 4.
The area beside the North Avenue Station of the Metro Rail Transit (MRT) in
Quezon City, which is where the highest temperature of 35.5uC occurred on 4 May
2002, can be seen as a ‘hot spot’ in the image. Other examples where maximum
temperatures have occurred within them include the Makati commercial district (16
February 2003), Port Area (7 December 2002), and the Chinese Cemetery in
Caloocan (29 January 2005). Ortigas Centre can also be identified as a warm spot.
Major highways such as EDSA, South Expressway, Ortigas Avenue, Aurora
Boulevard and Quezon Avenue can clearly be seen in the image to be warm as well.
The night-time UHI effect in Metro Manila 11
Downloaded By: [Stanford University] At: 17:47 4 April 2008
This may be attributed to the cemented material of the roads, which stores
significant amounts of solar heat, and to the large vehicular concentration during
the daytime. The airport and its runway located below and outside the major heat
island are also notably warm.
Within the heat island, cold spots or isolated areas with lower temperatures can be
identified, including the Sta. Ana racetrack in Makati, the Wack-Wack Golf and
Country Club in Mandaluyong, Camp Aguinaldo, and the Manila Seedling Bank
Foundation, Inc. in Quezon City, which is located beside the North Avenue MRT
station. These are mostly grass-covered areas. The Villamor golf course adjacent to
the airport can also be seen as an area with a relatively lower temperature.
In general, warmer temperatures are exhibited by major roads and highways,
commercial or highly built-up areas and areas of high anthropogenic activity, while
parks, grasslands and open spaces are characterized by lower temperatures.
Figure 4. The commercial district of Metro Manila is revealed as a heat island with smaller
heat islands or ‘hot spots’ (in black) as well as ‘cool islands’ (in white) that make
them thermally distinct from their surrounding areas. The ‘hot spots’ are areas of high
building concentration and anthropogenic activity while the cool zones are predominantly
vegetated.
12 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
5.4 UHI profile
The temperature profile of the BB9 transect drawn over the 4 May 2002 ASTER
thermal image (figure 5) clearly shows a warmer temperature as the gradient passes
through Metro Manila. The statistics of each transect profile for all the satellite
scenes used in this study are shown in table 4. The data are different from the
average temperature values given in section 5.2 to define UHI and are simply
presented as reference information on the temperature limits along a single transect.
The transect illustrates and identifies the sections of the UHI profile and
conforms to the general cross-section of the urban–rural temperature profile, which,
according to Oke (1987), is characterized by a steep temperature gradient at the
rural–urban boundary (‘cliff’), followed by a ‘plateau’ of warm air with a steady but
weaker horizontal gradient of increasing temperature as it reaches the ‘peak’ at the
city centre. Interruptions of ‘valleys’ as well as minor ‘peaks’ can be observed as the
gradient crosses distinct land-use features such as parks, lakes, open areas, and
commercial, industrial and residential developments as well as other topographical
features of the land.
Figure 5 also shows these UHI profile sections across the study area. From the
end point in Bulacan, the low temperature is fairly constant within that rural part of
the transect. The ‘cliff’ begins just as it enters Metro Manila while the ‘plateau’ run s
almost entirely across the major heat island identified earlier. The ‘peak’ zone is in
the Makati central business district (CBD), which is notably bright in colour.
Makati is the commercial, communication, cultural and financ ial centre of the
country and is a densely built-up area because of the location of the offices of many
Figure 5. Sections of the UHI profile and their corresponding areas in Metro Manila.
The night-time UHI effect in Metro Manila 13
Downloaded By: [Stanford University] At: 17:47 4 April 2008
companies and corporations. The temperature then begins to decrease as it leaves
the region of the heat island towards the south in the area of Laguna.
Spikes or minor peaks within the ‘plateau’ that signify an abrupt increase in
temperature are identified to be major highways. The first and highest peak
immediately after the ‘cliff’ is found where the transect crosses EDSA. The second
spike intersects with Quezon Avenue. After the Villamor golf course ‘valley’, the
subsequent ‘peak’ is identified as the airpor t, coinciding with the runway. The low-
temperature zones or ‘valleys’ just before and after the ‘peak’ are identified as the
Sta. Ana racetrack and the Villamor golf course, respectively, both in Makati.
The jagged lines of the profile show the complex behaviour of an urban surface as
the transect passes through areas of different morphology. This is also indicated by
the larger standard deviation, which further suggests wider variation in temperature.
These variations demonstrate that microclimatic attributes of the urban areas are
Table 4. Surface temperature statistics of transects (uC).
Scene date/transect Mean Min Max Range SD
4 May 2002
AA9 28.60 23.30 34.30 11.00 1.95
BB9 28.22 23.20 34.10 10.90 2.16
CC9 28.04 22.40 34.30 11.90 2.31
DD9 29.11 23.20 34.00 10.80 1.73
7 December 2002
AA9 26.66 23.00 32.00 9.00 1.44
BB9 25.99 21.50 32.30 10.80 1.74
CC9 25.34 20.50 31.80 11.30 1.97
DD9 25.55 21.10 32.30 11.20 1.94
24 January 2003
AA9 25.18 19.90 29.20 9.30 1.37
BB9 24.84 21.20 29.30 8.10 1.35
CC9 24.87 20.60 29.30 8.70 1.34
DD9 25.21 22.50 29.30 6.80 1.04
31 January 2003
AA9 22.95 17.10 28.40 11.30 1.81
BB9 22.50 16.80 27.90 11.10 1.74
CC9 22.36 16.50 27.70 11.20 1.78
DD9 22.51 18.30 28.40 10.10 1.55
16 February 2003
AA9 24.30 18.90 30.20 11.30 1.93
BB9 23.76 18.10 28.90 10.80 1.87
CC9 23.37 16.90 29.40 12.50 2.17
DD9 23.12 16.60 30.00 13.40 2.29
22 March 2004
AA9 26.34 22.80 30.90 8.10 1.25
BB9 25.90 22.70 30.50 7.80 1.39
CC9 25.23 19.60 31.00 11.40 1.81
DD9 24.60 18.30 30.40 12.10 2.05
29 January 2005
AA9 27.22 21.70 32.30 10.60 1.74
BB9 26.76 22.00 32.30 10.30 1.57
CC9 26.45 21.40 31.80 10.40 1.64
DD9 26.75 23.00 32.30 9.30 1.52
14 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
not homogeneous but can vary widely even between adjacent spots (Yannas 2001)
because of the different material characteristic an d anthropogenic activities (Ben-
Dor and Saaroni 1997) within an urban area. Because of the varying radiant load,
surface temperature differences may vary considerably, up to tens of degrees on a
microscale (Chudnovsky et al. 2004). The existence of these temperature differences
indicates that isolated areas of improved microclimate (e.g. parks) can survive and
start a reversal of the UHI effect (Yannas 2001).
5.5 NDVI–temperature relationship
Figure 6 shows the NDVI image of the study area transformed from a daytime
ASTER VNIR surface reflectance imag e. The NDVI roughly estimates the amount
of vegeta tion over an area based on the reflectance of its visible and near-infrared
wavelengths. In figure 6, built-up areas that are poorly vegetated and therefore have
a low NDVI appear dark in colour while those areas that are highly vegetated are
bright.
Figure 6. NDVI image of the study area transformed from an ASTER VNIR daytime
image acquired on 10 November 2001, 10:39 local time. Areas with a small amount of
vegetation (built-up areas) appear dark in colour while those that are highly vegetated are
bright.
The night-time UHI effect in Metro Manila 15
Downloaded By: [Stanford University] At: 17:47 4 April 2008
Linear regression analysis shows a negative correlation between NDVI and
temperature (figure 7) with a computed determination coefficient (R
2
) of 0.22 and a
regression equation of y534.92213.18x. The correlation is weak but nevertheless
demonstrates the well-documented (Gallo et al. 1993, Lambin and Ehrlich 1996,
Weng 2001, Dousset and Gourmelon 2003, Lo and Quattrochi 2003) inverse
relationship between temperature and NDVI, wherein vege tation can lower or have
a cooling effect on the temperature of an area.
The weak correlation in the analysis may be attributed to several factors: (1) the
edges of the clouds may not have been completely masked; (2) the shadows of the
clouds may have covered an area with little or no vegetation but, because of the
shadow, may have given a low temperature; (3) streams and rivers within the city
might also have not been masked out completely yielding erroneous NDVI values ;
and (4) accurate aerosol-free NDVI was not obtained.
Furthermore, the analysis was based on only a single image data set. At the time
the study was conducted, there were no other ASTER VNIR daytime images
with minimal cloud cover available from which to extract the NDVI. The relation-
ship between NDVI and temperature might have been stronger had other images
been used. In addition, the relationship may have varied over the season. The
analysis is, however, limited only for the wet season, which is still another factor
that may explain the result of a weak vegetation–temperature relationship in this
study.
To further understand the effect of land co ver on the UHI effect in Metro Manila,
further work is needed, including the measurement of aerosol in the atmosphere.
Atmospheric aerosol is reported to influence the reflectance of the red and near-
infrared bands (Takemata et al. 2004), thereby affecting NDVI measurements.
6. Conclusions
Urbanization has caused significant changes in the climate of cities. The most
common of these is the UHI effect manifested by the warming of urban areas
Figure 7. Scatter plot of the negative correlation between surface temperature and NDVI.
16 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
relative to adjacent rural areas caused by the replacement of natural land and
vegetation with building materials such as asphalt, concrete, brick and stone.
Advances in satellite remote sensing have given rise to improved climatic studies,
especially those on heat islands, as synoptic data coverage is provided. ASTER has
been demonstrated to be useful for this study because of its high spatial and spectral
resolution. Night-time surface kinetic temperature data products, which were derived
from the five TIR bands of ASTER, were used to quantify and describe the thermal
characteristics of the surface heat island effect in Metro Manila. The study demon-
strated the potential of such an instrument to analyse the thermal behaviour of the
Earth’s surface. It is particularly useful in urban surface studies because it can distin-
guish thermal differences between small and adjacent areas of different characteristics
with high accuracy considering the heterogeneity and disarray of the urban surface.
The sparse temporal resolution of the images used was mainly due to the small
number of available satellite scenes with clear skies. However, we do not consider
that this limits the results of the study, nor the overall capacity of the method in
depicting the night-time UHI effect, because the objective of the study was to
determine the degree of the UHI effect, which required, and was in fact provided
with, high spatial resolution data. In contrast to in situ-based studies, where point
source data bias temperature measur ements from areas only where weather stations
are located, the present study gave a synoptic view of the behaviour of surface
temperature over the entire study area. However, the sparse temporal resolution of
the data set did not allow for the determination of how the UHI effect varies
throughout the year or over the different seasons, as well as its development or
changes in the given period.
From the given data set, the warmest UHI effect recorded for Metro Manila was
an intensity of 2.96uC from the 4 May 2002 ASTER image. The highest temperature
the city reached was 35.5uC, which was recorded for the same satellite scene date.
The spatial patte rn of temperatures exhibited by the thermal images shows the
presence of a heat island in Metro Manila developed around the commercial and
business district of the city. Built-up areas formed ‘hot spots’ that were identified in
the images, as well as isolated cold spo ts characterized by open and vegetated areas
and that can be regarded as ‘cool islands’. It was also shown that temperature
becomes lower away from the city centre towards the rural areas of Bulacan and
Cavite/Laguna.
Temperature profiles over the rural communities of Bulacan and Cavite/Laguna
towards Metro Manila follow the general description of the UHI profile, which is
characterized by ‘cliffs’, ‘plateaus’ and a ‘peak’. Variations in the temperature
profile correspond to different land cover features depicting their thermal
characteristics as the gradient went across from the rural to the urban area.
There was also a weak but evident negative correlation between NDVI and
temperature. This inverse relationship is well documented in remote sensing
literature. As vegetation can have a cooling effect on the temperature of the
atmosphere, it is understandable that to alleviate the UHI effect, the most efficient
way is to increase the amount of vegetation in cities by planting more trees. Planting
programmes should, however, determine the exact locations, arrangements as well
as the type or kind of plants that would be most beneficial to the city and would best
give the concerted effects of minimizing the warming caused by heat islands. An
example of how this can be achieved is for buildi ng owners to set up plants on
rooftops or roof gardens to increase the amount of vegetation within their territory.
The night-time UHI effect in Metro Manila 17
Downloaded By: [Stanford University] At: 17:47 4 April 2008
To be truly effective, the measure should be implemented on a wide spatial scale (e.g.
the entire city). Creating several small green open spaces or parks spread evenly
around a city is also more effective in cooling the urban atmosphere than developing
large isolated tracts of vegetated areas. Incentive programmes can also be promoted
to encourage planting activity.
Acknowledgements
The data were distributed by the Land Processes Distributed Active Archive Centre
(LP DAAC), located at the U.S. Geological Survey (USGS) Centre for Earth
Resources Observation and Science (EROS) (http://lpdaac.usgs.gov).
References
AHRENS, C.D., 1998, Meteorology Today: An Introduction to Weather, Climate, and the
Environment, 3rd edn (MN: West Publishing Company).
ARNFIELD, A., 2003, Two decades of urban climate research: a review of turbulence,
exchanges of energy and water, and the urban heat island. International Journal of
Climatology, 23, pp. 1–26.
BECKER, F. and LI, Z.-L., 1990a, Temperature-independent spectral indices in thermal
infrared bands. Remote Sensing of Environment, 32, pp. 17–33.
BECKER, F. and LI, Z.-L., 1990b, Towards a local split window method over land surfaces.
International Journal of Remote Sensing, 11, pp. 369–393.
BEN-DOR, E. and SAARONI, H., 1997, Airborne video thermal radiometry as a tool for
monitoring microscale structures of the urban heat island. International Journal of
Remote Sensing, 18, pp. 3039–3053.
CHUDNOVSKY, A., BEN-DOR, E. and SAARONI, H., 2004, Diurnal thermal behavior of selected
urban objects using remote sensing measurements. Energy and Buildings, 36, pp.
1063–1074.
DOUSSET, B. and GOURMELON, F., 2003, Satellite multi-sensor data analysis of urban surface
temperatures and landcover. ISPRS Journal of Photogrammetry and Remote Sensing,
58, pp. 43–54.
EMMANUEL, R., 1993, A hypothetical ‘shadow umbrella’ for thermal comfort enhancement in
the equatorial urban outdoors. Architectural Science Review, 36, pp. 25–36.
ERSDAC, 2003, ASTER Reference Guide, Version 1.0 (Japan: Earth Remote Sensing Data
Analysis Center).
FUKUI, Y., 2003, A study on surface temperature patterns in the Tokyo Metropolitan area
using ASTER data. Geosciences Journal, 7, pp. 343–346.
GALLO, K., MCNAB, A., KARL, T., BROWN, J., HOOD, J. and TARPLEY, J., 1993, The use of
NOAA AVHRR data for assessment of the urban heat island effect. Journal of
Applied Meteorology, 32, pp. 899–908.
GILLESPIE, A., ROKUGAWA, S., MATSUNAGA, T., COTHERN, J.S., HOOK, S. and KAHLE, A.,
1998, A temperature and emissivity separation algorithm for Advanced Spaceborne
Thermal Emission and Reflection radiometer (ASTER) images. IEEE Transactions on
Geoscience and Remote Sensing, 36, pp. 1113–1126.
GILLESPIE, A., ROKUGAWA, S., HOOK , S.J., MATSUNAGA, T. and KAHLE, A.B., 1999,
Temperature/emissivity separation algorithm theoretical basis document, version 2.4.
Available online at: http://eospso.gsfc.nasa.gov/eos_homepage/for_scientists/atbd/
docs/ASTER/atbd-ast-05-08.pdf (accessed 21 August 2005).
HINKEL, K.M., NELSON, F.E., KLENE, A.E. and BELL, J.H., 2003, The urban heat island in
winter at Barrow, Alaska. International Journal of Climatology, 23, pp. 1889–1905.
ICAMINA, P., 1995, The boiling point. In In the Shadow of the First World: The Environment as
Seen from Developing Nations, W.P. Carty and E. Lee (Eds), pp. – (Chicago: Chicago
Review Press).
18 M. Tiangco et al .
Downloaded By: [Stanford University] At: 17:47 4 April 2008
JA
¨
GER, J., 1983, Climate and Energy Systems: A Review of Their Interactions (New York: John
Wiley & Sons).
JET PROPULSION LABORATORY, 2001, ASTER Higher-Level Product User Guide, Version 2.0,
Available online at: http://asterweb.jpl.nasa.gov/content/03_data/04_Documents/
ASTERHigherLevelUserGuideVer2May01.pdf (accessed 21 August 2005).
KAHLE, A.B. and ALLEY, R.E., 1992, Separation of temperature and emittance in remotely
sensed radiance measurements. Remote Sensing of Environment, 42, pp. 107–111.
KEALY, P.S. and HOOK, S.J., 1993, Separating temperature and emissivity in thermal infrared
multispectral scanner data: implications for recovering land surface temperatures.
IEEE Transactions on Geoscience and Remote Sensing, 31, pp. 1155–1164.
KEALY, P.S. and GABELL, A.R., 1990, Estimation of emissivity and temperature using alpha
coefficients. In Proceedings of the 2nd TIMS Workshop, Pasadena, CA (Pasadena:
JPL Publication), pp. 90–55.
LAMBIN, E. and EHRLICH, D., 1996, The surface temperature-vegetation index space for land
cover and land-cover change analysis. International Journal of Remote Sensing, 17, pp.
463–487.
LO, C.P. and QUATTROCHI, D.A., 2003, Land-use and land-cover change, urban heat island
phenomenon, and health implications: a remote sensing approach. Photogrammetric
Engineering and Remote Sensing, 69, pp. 1053–1063.
LO, C.P., QUATTROCHI, D.A. and LUVALL, J.C., 1997, Application of high-resolution thermal
infrared remote sensing and GIS to assess the urban heat island effect. International
Journal of Remote Sensing, 18, pp. 287–304.
NICHOL, J.E., 1996, High-resolution surface temperature patterns related to urban mor-
phology in a tropical city: a satellite-based study. Journal of Applied Meteorology, 35,
pp. 135–146.
NICHOL, J.E., 1998, Visualisation of urban surface temperatures derived from satellite images.
International Journal of Remote Sensing, 19, pp. 1639–1649.
NSCB, 2005, National Statistical Coordination Board – Statistics – Population and Housing.
Available online at: www.nscb.gov.ph/secstat/d_popn.asp (accessed 12 October 2006).
OKE, T.R., 1987, Boundary Layer Climates (New York: Methuen Press).
OWEN, T.W., CARLSON, T.N. and GILLIES, R.R., 1998, An assessment of satellite remotely-
sensed land cover parameters in quantitatively describing the climatic effect of
urbanization. International Journal of Remote Sensing, 19, pp. 1663–1681.
PRATA, A.J., CASELLES, V., COLL, C., SOBRINO, J.A. and OTTLE, C., 1995, Thermal remote
sensing of land surface temperature from satellites: current status and future
prospects. Remote Sensing Reviews, 12, pp. 175–124.
SAARONI, H., BEN-DOR, E., BITAN, A. and POTCHTER, O., 2000, Spatial distribution and
microscale characteristics of the urban heat island in Tel-Aviv, Israel. Landscape and
Urban Planning, 48, pp. 1–18.
SAARONI, H., BITAN , A., BEN-DOR, E. and FELLER, N., 2004, The mixed results concerning
the ‘oasis effect’ in a rural settlement in the Negev Desert, Israel. Journal of Arid
Environments, 58, pp. 235–248.
SAITO, I., ISHIHARA, O. and KATAYAMA, T., 1990/91, Study of the effect of green area
on the thermal environment in an urban area. Energy and Buildings, 15/16, pp.
493–498.
SCHMUGGE, T., HOOK, S.J. and COLL, C., 1998, Recovering surface temperature and
emissivity from thermal infrared multispectral data. Remote Sensing of Environment,
65, pp. 121–131.
SNYDER, W.C., WAN, Z., ZHANG, Y. and FENG, Y.-Z., 1998, Classification-based emissivity
for land surface temperature measurement from space. International Journal of
Remote Sensing, 19, pp. 2753–2774.
SOBRINO, J.A., JIME
´
NEZ-MUN
˜
OZ, J.C., BALICK, L., GILLESPIE, A.R., SABOL, D.A. and
GUSTAFSON, W.T., 2007, Accuracy of ASTER Level-2 thermal-infrared standard
The night-time UHI effect in Metro Manila 19
Downloaded By: [Stanford University] At: 17:47 4 April 2008
products of an agricultural area in Spain. Remote Sensing of Environment, 106, pp.
146–153.
STATHOPOULOU, M., CARTALIS, C. and KERAMITSOGLOU, I., 2004, Mapping micro-urban heat
islands using NOAA/AVHRR images and CORINE Land Cover: an application to
coastal cities of Greece. International Journal of Remote Sensing, 25, pp. 2301–2316.
STOLL, M. and BRAZEL, A., 1992, Surface–air temperature relationships in the urban
environment of Phoenix, Arizona. Physical Geography, 13, pp. 160–179.
S
TREUTKER, D.R., 2002, A remote sensing study of the urban heat island of Houston, Texas.
International Journal of Remote Sensing, 23, pp. 2595–2608.
STREUTKER, D.R., 2003, Satellite-measured growth of the urban heat island of Houston,
Texas. Remote Sensing of Environment, 85, pp. 282–289.
TAKEMATA, K., KAWATA, Y. and NAOE, N., 2004, The correlation between NDVI value and
surface temperature measured by ASTER. In Proceedings of the XXth ISPRS
Congress, 12–23 July 2004, Istanbul, Turkey. Available online at: www.isprs.org/
istanbul2004/comm7/papers/104.pdf (accessed 28 May 2007).
V
OOGT, J.A. and O KE, T.R., 2003, Thermal remote sensing of urban climates. Remote Sensing
of Environment, 86, pp. 370–384.
WATSON, K., 1992a, Spectral ratio method for measuring emissivity. Remote Sensing of
Environment, 42, pp. 113–116.
WATSON, K., 1992b, Two-temperature method for measuring emissivity. Remote Sensing of
Environment, 42, pp. 117–121.
WENG, Q., 2001, A remote sensing-GIS evaluation of urban expansion and its impact on
surface temperature in the Zhujiang Delta, China. International Journal of Remote
Sensing, 22, pp. 1999–2014.
WENG, Q., 2003, Fractal analysis of satellite-detected urban heat island effect.
Photogrammetric Engineering and Remote Sensing, 69, pp. 555–566.
WENG, Q. and YANG, S., 2004, Managing the adverse thermal effects of urban development in
a densely populated Chinese city. Journal of Environmental Management, 70, pp.
145–156.
WENG, Q., LU, D. and SCHUBRING, J., 2004, Estimation of land surface temperature–
vegetation abundance relationship for urban heat island studies. Remote Sensing of
Environment, 89, pp. 467–483.
YAMAGUCHI, Y., KAHLE, A.B., TSU, H., KAWAKAMI, T. and PNIEL, M., 1998, Overview of
Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER).
IEEE Transactions on Geoscience and Remote Sensing, 36, pp. 1062–1071.
Y
ANG, S. and WANG, S., 1989, The effect of the afforestation trees in lowering temperatures
and enhancing humidity of the air in Guangzhou [in Chinese]. Geographical Series
III, Journal of South China Normal University, pp. 41–46.
Y
ANNAS, S., 2001, Toward more sustainable cities. Solar Energy, 70, pp. 281–294.
20 The night-time UHI effect in Metro Manila
