- Email: [email protected]
Reconstruction of historical datasets for analyzing spatiotemporal influence of built environment on urban microclimates across a compact city
Accepted Manuscript Reconstruction of historical datasets for analyzing spatiotemporal influence of built environment on urban microclimates across a compact city Fen Peng, Man Sing Wong, Hung Chak Ho, Janet Nichol, Pak Wai Chan PII:
S0360-1323(17)30337-2
DOI:
10.1016/j.buildenv.2017.07.038
Reference:
BAE 5017
To appear in:
Building and Environment
Received Date: 24 June 2017 Revised Date:
24 July 2017
Accepted Date: 25 July 2017
Please cite this article as: Peng F, Wong MS, Ho HC, Nichol J, Chan PW, Reconstruction of historical datasets for analyzing spatiotemporal influence of built environment on urban microclimates across a compact city, Building and Environment (2017), doi: 10.1016/j.buildenv.2017.07.038. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT 1
Title: Reconstruction of Historical Datasets for Analyzing Spatiotemporal Influence
2
of Built Environment on Urban Microclimates across a Compact City
3
Authors: Fen Penga, Man Sing Wonga,*, Hung Chak Hoa, Janet Nichola, Pak Wai Chanb
RI PT
4 5 6
Affiliations:
7
a
8
University, Kowloon, Hong Kong
9
b
M AN U
SC
Department of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic
Hong Kong Observatory, Hong Kong
10
*Corresponding Author: Man Sing Wong, Department of Land Surveying and
12
Geo-Informatics, The Hong Kong Polytechnic University, Kowloon, Hong Kong.
13
Email: [email protected]
EP
15
AC C
14
TE D
11
1
ACCEPTED MANUSCRIPT 1
Abstract The high-rise/high-density environment of a compact city can influence the
3
microclimate resulting in lower living quality. Previous studies have analyzed the
4
relationships between high-rise/high-density environment and microclimates, by
5
either a temporal study or a spatial approach, while a strategy for investigating the
6
spatiotemporal relationship has yet to be developed. This study initiated a set of
7
innovative strategies to map the historical built environment/microclimates of a
8
compact city, with a spatiotemporal approach to analyze the relationships between
9
building structures and urban climates, for developing a sustainable protocol for
10
future urban planning. Three major components were reconstructed, including 1) the
11
annually averaged Land Surface Temperature (LST) for determining the relative
12
temperature across a compact city; 2) 3D building datasets for representing the
13
building morphology; and 3) sets of urban morphological data derived from building
14
datasets for analyzing microclimate and thermal distress.
TE D
M AN U
SC
RI PT
2
There are high correlations between observed and predicted LSTs (R = 0.64 to
16
0.89), with mean absolute error (MAE) of annually averaged LST ranging 0.49°C to
17
2.60°C, and root mean square error (RMSE) ranging 0.62°C to 2.98°C. There are low
18
errors for reconstructing building data, in which MAEs and RMSEs of an open space
19
are 0.41m - 1.23m and 0.78m - 1.46m; and for an area with buildings are 0.81m -
20
3.25m and 1.06m - 5.92m. The spatiotemporal estimation indicated areas with
21
improved air ventilation through years can significantly reduce an additional 0.12°C -
22
1.09°C than the areas without improvement, while areas with an increase in shades
AC C
EP
15
2
ACCEPTED MANUSCRIPT 1
through years have 0.6°C to 0.76°C higher reduction of relative temperature.
2
Keywords: historical built environment; temperature; shading effect; air ventilation;
4
urban design; spatial analytics
RI PT
3
AC C
EP
TE D
M AN U
SC
5
3
ACCEPTED MANUSCRIPT 1 2
1. Introduction Population growth and limited natural resources are two global challenges since
4
the twentieth century. As a result, an urban form with high-rise and high-density
5
urban settings, named as “compact city”, has since been in practice in the
6
developing countries for minimizing population stress [1]. Although the idea of a
7
“compact city” was originally an urban design for developing a more sustainable
8
environment, the high-rise and high-density settings of these cities have also formed
9
a unique building morphology that can influence the microclimate across urban
10
areas [2,3]. Previous studies for sustainable planning have focused on applying
11
temporal analyses to study the relationship between urban climates and building
12
structures [4-7] for overcoming the planning issues from a problematic urban design
13
(e.g. building with higher thermal distress, living environment with poor air
14
ventilation).
TE D
M AN U
SC
RI PT
3
The temporal analysis can partially provide fundamental understanding for
16
sustainable urban planning. For example, Lau and Ng [5] used only five weather
17
stations of Hong Kong to examine the interaction between built environments and
18
air temperature, and found a cooling effect of high-density environment due to the
19
decline of urbanization. However, there is a limitation of ordinal temporal study, in
20
which these studies only used several locations of a city as demonstration, leading to
21
a potential bias of not being able to represent the holistic urban built environment
22
across a compact city. Thus, an alternative of using remotely sensed images has then
AC C
EP
15
4
ACCEPTED MANUSCRIPT been applied, for mapping built environment and for analyzing urban microclimates
2
[8-10]. This technique is shown to be promising, since satellite images or aerial
3
photographs increase the spatial coverage of an urban climatic analysis. A common
4
example is the application of thermal remote sensing for effectively mapping land
5
surface temperature (LST) across a city, resulting to improve the results of analysis of
6
urban heat island effect in many small areas over a large region [11-17], while the
7
traditional method is limited by sparse distribution of weather stations across a
8
compact city. This advantage has led to a wide-use of remote sensing techniques for
9
monitoring the urban environment, for sustainable policy making, and for urban
SC
M AN U
10
RI PT
1
planning [18, 19].
However, the use of remote sensing techniques may also be a trade-off for
12
temporal resolution [20]. For example, using satellite images to retrieve land features
13
are restricted by the satellite overpass and cloud coverage, and the applications of
14
aerial photographs are restricted by the number of flights across the city. It is
15
necessary to develop a set of techniques to reconstruct the historical datasets that
16
can monitor the changing urban environment with high spatial and temporal quality,
17
in order to investigate the relationship between building morphology and
18
microclimate for developing sustainable planning protocols.
EP
AC C
19
TE D
11
This study hereby develops a set of spatiotemporal models that can accurately
20
map the urban built environment with high spatial resolution for each year, including
21
1) annually averaged LST indicating the spatial variability of urban heat across a
22
high-density environment, 2) geodatabase with 3-dimensional (3D) building data that 5
ACCEPTED MANUSCRIPT can fully represent the spatial patterns of historical urban morphology, and 3) a set of
2
spatial data derived from the 3D building data that can be used as building
3
morphological parameters for urban climatic study. A spatiotemporal analysis was
4
also conducted to estimate the impact of building morphology on urban
5
microclimate between 1988 and 2010 across Kowloon Peninsula, Hong Kong. Based
6
on previous studies, which has suggested that high-rise and high-density
7
environments have significantly contributed to urban microclimates [5, 21] Kowloon
8
Peninsula was selected as the site for this study. It is expected that a data-driven
9
analysis of Kowloon Peninsula can be a global example of urban climatic studies for
SC
M AN U
10
RI PT
1
compact cities.
11
2. Study Area
TE D
12
The Kowloon Peninsula is an extensively urbanized area in Hong Kong (Figure 1),
14
with approximately 1 million people living in a relatively small region (~ 26 km2) [22].
15
Kowloon Peninsula has a typical setting of a compact city, where the majority of
16
lands consist of high-rises and is a high-density environment with very few urban
17
greenspaces to be found across the study site. A “wall effect” controlled by building
18
morphology has been recognized as one of the main effects that can block sea
19
breezes for reducing thermal distress during a hot summer day [23]. Influencing by
20
the subtropical climate, there are multiple Very Hot Days (>= 33°C) and Hot Night
21
(>=28°C) as observed by the weather station located at the Hong Kong Observatory
22
(HKO) in Kowloon Peninsula [24]. During these extremely hot days and nights, urban
AC C
EP
13
6
ACCEPTED MANUSCRIPT heat islands with extremely high temperatures can occur across the high-density
2
environment and, as a result, these can induce the significance of thermal
3
discomfort, morbidity risks, and even mortality risks for its inhabitants [9, 21, 25].
4
However, a study focusing on intra-urban temperature difference has also observed
5
a phenomenon of “urban cool island” across this high-density environment [26].
6
Thus it is necessary for the spatial pattern of microclimate across Kowloon Peninsula
7
to be further explored, especially as the site area contains a high percentage of
8
poorly maintained aged buildings (>= 50 years) that can influence the microclimatic
9
variations as a result of urban renewal. A spatiotemporal analysis to study the
10
influence of building morphology on urban climate across Kowloon Peninsula is
11
necessary since it is important to develop sustainable protocols prior to future urban
12
planning.
13
3. Data and Methods
15
3.1 Data
EP
14
TE D
M AN U
SC
RI PT
1
There are three major analyses to reconstruct historical datasets for this urban
17
climatic study (Figure 2), including 1) annually averaged LST estimation, 2) 3D
18
reconstruction of building data, and 3) retrieval of building morphology data based
19
on results from the aforementioned analyses.
AC C
16
20
In this study, fifteen Landsat 5 TM satellite images from the years 1988, 1996,
21
2000, 2005 and 2010, with three corresponding images per year, of cloud-free scene
22
across the study area were used to estimate the annually averaged LST (Table 1). The 7
ACCEPTED MANUSCRIPT hourly air temperature data between 1885 and 2010 from the weather station
2
located at the HKO were used to improve the LST estimation. Data of hourly wind
3
direction of the HKO station between 1988 and 2010 were also retrieved for
4
estimating the ventilation paths. In addition, seventy-seven aerial photographs taken
5
between 1964 and 2010 were used to reconstruct the built environment of the
6
Kowloon Peninsula (Table 2). Acquired from the Hong Kong Lands Department, these
7
aerial photographs contain at least 83 control points with 3D information that can be
8
used to accurately reconstruct the building structures across the high-density
9
environment. Airborne light detection and ranging (LiDAR) data of 2010 and 2011
10
[27] were applied for generation of a digital surface model (DSM) and a digital
11
elevation model (DEM) with 1-meter resolution, which can be used to validate the
12
3D historical building data reconstructed by aerial photographs.
14
SC
M AN U
TE D
13
RI PT
1
3.2 Reconstruction of Annually Averaged LST A mono-window algorithm was applied to estimate the LST, based on thermal
16
infrared (TIR) band of each Landsat 5 TM image [28], with additional parameters of
17
upwelling and downwelling atmospheric radiances, atmospheric transmittance, and
18
surface emissivity retrieved from the MODTRAN 4.0 code [29]. By using these LST
19
images, a quadratic function was applied to estimate the daily LST from 1988 to 2010.
20
The concept of using a quadratic function for estimation was first developed based
21
on a quick comparison of LST and the day of year using four pixels (1 km x 1 km)
22
randomly selected on all MODerate resolution Imaging Spectroradiometer (MODIS)
AC C
EP
15
8
ACCEPTED MANUSCRIPT 1
LST products in 2010 (Figure 3), with results showing a quadratic trend between daily
2
LSTs and day of a year. Two of three Landsat images of each year were used to
3
construct a quadratic function: Y = −a ∗
+
+
5
Y = −a ∗
− ⁄ 2 ∗ )) +
+4∗
RI PT
∗
(1)
4
∗ )⁄ 4 ∗ )
(2)
where
is the day of the year, Y is the LST at pixel i, and a, b, c are the
7
coefficients of the quadratic function. Based on the quadratic distribution of air
8
temperatures at the HKO station through a year (Figure 4),
9
defined as the hottest day of a year. Therefore, a day of the year with the hottest
10
temperature was substituted in the Eq. (2) for estimating the daily LST with the
11
information from selected Landsat images.
⁄ 2 ∗ ) can be
TE D
M AN U
SC
6
In order to validate the quadratic function, the estimated LSTs at the day of
13
another Landsat image were compared with the retrieved LST [30]. All pixels on the
14
images were compared, with measures of root mean square error (RMSE) and mean
15
absolute error (MAE), for determining the accuracy of the quadratic function. After
16
validation, the results of each year were summarized to be an annually averaged LST
17
image. Each annually averaged LST map was converted to relative temperature map
18
using Kowloon Park as a reference. The use of Kowloon Park as a reference is due to
19
its extensive greenspace, while a comparison of surface temperature between
20
Kowloon and other areas can help estimate the potential thermal discomfort
21
influenced by urban morphology. Based on these relative temperature maps, the
AC C
EP
12
9
ACCEPTED MANUSCRIPT 1
spatial variability of temperature across the study area based on the temperature
2
difference from Kowloon Park can be depicted.
3
3.3 Reconstruction of 3D building data
RI PT
4
ERDAS Leica Photogrammetry Suite module with a stereo image matching
6
technique was applied to 1) identify dense matching points between all aerial
7
photographs; 2) estimate the 3D coordinates of each matching point; and 3)
8
interpolate all points for building a DSM that can represent the building environment
9
across the Kowloon Peninsula [31].
M AN U
SC
5
DSM generated by the stereo image matching technique (DSMPHG) was then
11
subtracted by a DSM generated from the airborne LiDAR data (DSMLiDAR) for
12
validation. A resultant pixel closer to zero from this subtraction indicated less
13
differences between DSMs and a higher accuracy of DSMPHG. In order to validate the
14
DSMPHG comprehensively, four sets of randomly distributed samples, with 100
15
samples of each set, were applied to determine the accuracy of DSMPHG across open
16
spaces and areas with buildings.
EP
AC C
17
TE D
10
To reconstruct the 3D building data, orthophotos and DSMs were used to
18
digitize the building footprints of the Kowloon Peninsula. Building footprints of 2010
19
from the Hong Kong Lands Department were used as the reference for digitization.
20
Building heights were estimated by subtracting the DSM from the LiDAR DEM and
21
were used as a reference to identify the possible areas with buildings during
10
ACCEPTED MANUSCRIPT 1
digitization. Lastly, the 3D building models were developed using ESRI ArcScene by
2
combining building footprints and building heights.
3
3.4. Reconstruction of urban morphological data
RI PT
4
Sky view factor (SVF) and ventilation paths were two major components that can
6
influence microclimate of urban areas. SVF is a ratio measuring obstructed sky of the
7
surrounding environment, which can indicate the amount of solar radiation received
8
by a planar surface and the entire hemispheric environment [32]. The ventilation
9
path is a ventilation corridor controlled by wind direction, wind speed, and building
10
structure. In this study, SVF was estimated based on a plugin module [33] of
11
ENVI+IDL [34], by a 90m searching radius on each pixel of each DSM between 1990
12
and 2010. Finally, ventilation paths were estimated based on 3D building data
13
between 1988 and 2010 by the method noted on Peng et al. [35] with the use of
14
wind data from HKO.
17 18
3.5. Spatiotemporal analyses for estimating influences of urban morphology on
AC C
16
EP
15
TE D
M AN U
SC
5
microclimate
Based on previous studies, shading effect and the air ventilation are two major
19
components controlled by the urban morphology that can influence microclimate
20
[36 – 38]. Therefore, two case studies for spatiotemporally analyzing impacts of air
21
ventilation and shades on surface temperature have been conducted in this study.
22
Annually average LSTs at the year with cloud-free Landsat images (1988, 1996, 2000, 11
ACCEPTED MANUSCRIPT 2005, and 2010) were used for the spatiotemporal comparison. These Landsat
2
images represent the cloud-free data with the closest year to the building datasets
3
retrieved from aerial photographs. To estimate the impact of urban ventilation on
4
temperature, a rule-based model was first used to identify open areas with and
5
without changes in air ventilation between two known years of the study.
6
Normalized wind speeds of 1988, 1996, 2000, 2005, and 2010 were estimated for
7
demonstrating air ventilation across the Kowloon Peninsula. In order to map the
8
area without significant change in air ventilation, the following rules (Figure 5) have
9
been used to a) identify areas with change of building height <= 5 meters in between
10
two known years; b) identify areas with SVF >= 0.9 for two known years for
11
hypothesizing an environment with more unchanged open spaces; c) map areas with
12
a change of SVF <= 0.05 between years to further identify the unchanged
13
environment; and d) identify an area with a change of normalized wind speed <=0.05
14
m/s between years to map the open area without changes in air ventilation. These
15
areas without changes in air ventilation were compared with the areas with
16
significant changes determined by the followings: i) map of unchanged open area
17
using a), b), and c) above, and ii) identification of the area with significant change of
18
air ventilation by considering all areas with at least 0.2 m/s increase of normalized
19
wind speed.
AC C
EP
TE D
M AN U
SC
RI PT
1
20
These results of rule-based model were further applied to an equation for
21
quantifying the spatiotemporal relationship between wind speed and relative
22
temperature. 12
ACCEPTED MANUSCRIPT 1
=
−
)−
−
)
(3)
where Temp represents the spatiotemporal difference in relative temperature (°C)
3
between areas with and without observed changes; t represents the year of latter
4
Landsat scene; t-1 represents year of earlier Landsat scene, SCt represents the
5
averaged relative temperature (°C) of all areas with observed changes at t, SCt-1
6
represents the averaged relative temperature (°C) of all areas with observed changes
7
at t-1; SUt represents the averaged relative temperature (°C) of areas without
8
observed change at t; and SUt-1 represents the averaged relative temperature (°C) of
9
all areas without observed change at t-1. In brief, this equation can estimate the
10
difference in relative temperature change across the Kowloon Peninsula due to the
11
observed increase in wind speed.
M AN U
SC
RI PT
2
For estimating the shading effects on surface temperature, two new sets of the
13
rule-based model (Figure 5) were used for identifying areas with and without
14
changes of shades. The areas without changes of shades were mapped by the
15
followings: a) identification of areas with the change of building height <= 5m
16
between years; b) identification of areas with the change of normalized wind speed
17
<= 0.05; and c) map of areas with an absolute change of SVF <=0.05. These areas
18
were compared with the area with a significant increase in shades, determined by
19
the rule a) and b) above, and an additional rule of mapping areas with SVF at least
20
0.1 lower after years. The rules a) and b) are to set control for the rule-based model,
21
thus the datasets for comparison only have the difference in shading effect. This
AC C
EP
TE D
12
13
ACCEPTED MANUSCRIPT method can minimize the bias from other environmental influences such as wind.
2
For comparing the areas with and without changes of shade, Eq. (3) was applied to
3
quantify the influence of shading effect, where the results can be expressed as a
4
difference in relative temperature change across the Kowloon Peninsula due to the
5
observed increase of shading.
6
4. Result
8
4.1. Validation of spatial data reconstruction
M AN U
7
SC
RI PT
1
LST estimation of this data reconstruction model is fairly accurate, with MAE
10
ranging between 0.49°C and 2.60°C and RMSE ranging between 0.62°C and 2.98°C
11
(Table 3). There is also a higher correlation between the estimated LST and the
12
observed LST (R = 0.64 to 0.89) for the results of 1988 and 2010. In addition, the
13
spatiotemporal trend of annually average LSTs was similar to the temporal change of
14
air temperature in the Kowloon Peninsula. Among the five years with retrieved
15
satellite images, the yearly average of maximum air temperature of 2005 was the
16
highest (35.4°C), followed by 1996 (34.3°C), 2000 (34.2°C), 2010 (34.1°C) and 1988
17
(33.8°C). In comparison, the results of annually average LSTs also showed that the
18
highest temperature occurred in 2005, with an average of 31.3°C for the whole study
19
area, compared to the lowest temperature (28.4°C) in 1988. A high accuracy of
20
reconstructing 3D building environment between 1964 and 2010 was also found in
21
this study (Figure 6). For an open space, MAE ranged from 0.41m to 1.23m and
22
RSME ranged from 0.78m to 1.46m (Table 4). For an area with buildings, MAE ranged
AC C
EP
TE D
9
14
ACCEPTED MANUSCRIPT from 0.81m to 3.25m, and RMSE ranged from 1.06m to 5.92m. Considering the
2
high-rise buildings (>100m) that can be found in the Kowloon Peninsula, high
3
accuracy of this dataset can be used as data input for mapping urban morphology
4
such as SVF and ventilation paths (Figure 7).
RI PT
1
5 6
4.2. Spatiotemporal influences of building morphologies on urban microclimate
The “old” airport in the study area, namely the “Kai Tak International Airport”,
8
was the region with the highest surface temperature between 1988 and 2010 (Figure
9
8). Impervious surface of Kai Tak has been expanding since 1998. This expansion of
10
impervious surface has increased the areas with higher surface temperature across
11
the Kowloon Peninsula. Areas with higher surface temperature have expanded more
12
rapidly after 2000, due to the closure of the Kai Tak International Airport at 1998,
13
changing this area to bare land. A trend of increasing building heights across the
14
Kowloon Peninsula has also been observed in this study, in which the average
15
building height has increased by 1.95m for every ten years since 1964. There were
16
only 5% of buildings above 70m before 2000, while 16% of buildings were above 70m
17
in 2010. Increases in building coverage and heights from 1964 have also significantly
18
influenced the urban morphology. This change of urban morphology to a high-rise,
19
high-density environment has reduced the paths of air ventilation since 1988. It is
20
important to note that the areas with less air ventilation paths are usually of higher
21
temperature (Figure 9), as wind has a significant contribution to reduce the urban
22
temperature. Multiple spots of the Kowloon Peninsula have observed a significant
AC C
EP
TE D
M AN U
SC
7
15
ACCEPTED MANUSCRIPT change in temperature through the years as air ventilation paths across these areas
2
have changed. Under this changing environment, E-W street canyons were the
3
significant wind corridors with higher wind speed across the Kowloon Peninsula
4
through the years.
RI PT
1
In addition, the variation of the urban building not only developed a complex
6
system for air ventilation, but also enhanced the shading effect on the high-density
7
environment. This shading effect significantly reduced the relative temperature
8
(Figure 10).
M AN U
SC
5
The cooling effect of air ventilation and shading is more significant based on the
10
quantitative analyses with Eq. (3). Areas with an improvement of air ventilation
11
determined by increased wind speed can reduce the relative temperature of 0.12°C
12
to 1.09°C more than the areas without improvement of air ventilation (Table 5).
13
Areas with an increase in shading also have a 0.6°C to 0.76°C temperature reduction
14
more than the areas without changes in shading effect. These findings showed that
15
the effect of air ventilation has dropped after 1996 most likely due to a major change
16
of ventilation paths, while the spatiotemporal effect of shades on microclimate
17
across the Kowloon Peninsula has been more stable.
EP
AC C
18
TE D
9
19
5. Discussion
20
5.1. Sustainable design and planning for urban heat mitigation
21
In this study, a reconstruction of historical building and environmental datasets
22
was conducted for microclimate assessment. These data were applied to analyze 16
ACCEPTED MANUSCRIPT how urban morphology could influence urban heat. It is found that availability of
2
shading and air ventilation are the major elements for reducing intra-urban
3
temperature. These results provided an implication that a high variation of building
4
height is needed for maintaining both significant mitigating effects of air ventilation
5
and shades on urban heat [39].
RI PT
1
A compact city is occupied by a high-rise and high-density environment that can
7
continue to provide a shading effect for reducing incoming solar radiation. However,
8
some high-rise buildings of this high-density environment can directly block the air
9
ventilation resulting in a parcel of high-temperature trapped in the high-density
10
spots. Therefore, in order to maintain the air ventilation and at the same time
11
providing a high-density environment that can sustain the large population in urban
12
areas, it is necessary to build up the core of urban centers with both high and low
13
buildings; this difference in building heights can generate a fluid dynamic path that
14
allows more cold air parcels to disperse into the urban center, at the same time
15
maintains the shading effects from building heights on ground-level heat. One
16
strategy that has been proposed for practicing in high-rise buildings is the use of
17
multiple horizontal wind corridors at the higher-floors of the building [40], which
18
allows better air ventilation in one vertical profile of a high-rise building and at the
19
same time this strategy does not need to reduce the building height. Combining
20
these strategies of improving shading effects and air ventilation in a high-density
21
environment, the relative temperature of urban areas should be reduced. Suggested
22
by the results of this study, the daytime thermal distress of the study area can be
AC C
EP
TE D
M AN U
SC
6
17
ACCEPTED MANUSCRIPT reduced by enhancing the shading effects, and by improving air ventilation based on
2
more E-W horizon corridors in a vertical profile. These combined strategies can also
3
be practiced with other common sustainable planning protocols such as urban
4
greening, while the latter has found to be useful for reducing thermal discomfort [41].
5
In addition, a future study can be conducted to forecast the change of perceived
6
temperature (e.g. physiological equivalent temperature, apparent temperature)
7
based on the historical datasets created by this study, in order to target the land use
8
policies and sustainable urban designs that should be applied to reduce thermal
9
discomfort across urban areas in the coming decades. This study is important,
10
because temporally-static perceived temperature maps have found to be useful to
11
determine thermal discomfort and heat health risk across urban areas [42 – 45],
12
while a more comprehensive spatiotemporal modelling with historical data should
13
enhance the assessment of thermal issues and for sustainable urban planning in the
14
future.
17
5.2. Limitations
AC C
16
EP
15
TE D
M AN U
SC
RI PT
1
In this study, the diurnal effect of daytime/nighttime urban temperature has
18
not been discussed. It is due to the limitation of data of nighttime satellite images for
19
reconstructing
20
spatiotemporal analysis. Previous study has attempted to use ASTER satellite image
21
to estimate the spatial distribution of nighttime urban temperature [9]. However,
22
lack of cloud-free images and lack of acquisition in the earlier past years have limited
historical
dataset
and
for
conducting
a
comprehensive
18
ACCEPTED MANUSCRIPT the ability of using ASTER images to reconstruct historical datasets. A future study
2
combining machine learning with limited data of nighttime satellite images can be
3
conducted for testing the ability to spatiotemporally reconstruct the intra-urban
4
temperature data in a compact city.
RI PT
1
Another limitation of this study is that the annually averaged LSTs were used
6
instead of the summer average. Annually averaged LST was used because of the shift
7
of summer in Hong Kong due to climate change, as there were no fixed months of
8
summer in Hong Kong since 2000, as the number of hot days has been increased
9
rapidly, although the hottest days are still commonly found in July and August. For
10
reducing statistical bias, annually averaged LST can minimize the random cutoff of
11
season for study. In the perspective of thermal comfort study, relative temperature
12
predicted by annually averaged LST is better matched to the scenario that
13
temperature has been rising exponentially due to climate change and hot days will
14
eventually cover the whole year in a sub-tropical city in the near future.
15
17
6. Conclusions
AC C
16
EP
TE D
M AN U
SC
5
This study has developed a comprehensive approach to reconstruct historical
18
temperature maps based on Landsat images and aerial photographs. In the past
19
years, there have also been innovative techniques to reconstruct the building
20
morphology, with 3D information for clearly representing the change of building
21
environment in a compact city. A spatiotemporal analysis was also conducted for
22
evaluating the impacts of building morphology on microclimate across a compact city. 19
ACCEPTED MANUSCRIPT The results showed that shading effects and influence of air ventilation should be the
2
keys to reducing surface temperature; while in the study area, more air ventilation
3
should be built along with the continuing shading effects for better improving the
4
urban thermal environment.
RI PT
1
5 6
Acknowledgements
This work was supported in part by the grant Early Career Scheme (project id:
8
25201614) from the Research Grants Council of Hong Kong, and grant G-YM85 from
9
the Hong Kong Polytechnic University. The authors thank the Hong Kong Lands
10
Department for the historical aerial photographs, ground control points, and existing
11
building GIS data; the Hong Kong Observatory for the meteorological data; NASA LP
12
DAAC for the Landsat satellite imagery.
M AN U
TE D EP
14
AC C
13
SC
7
20
ACCEPTED MANUSCRIPT References
2
[1] X.Q. Zhang. (2000). High-rise and high-density compact urban form: The
3
development of Hong Kong, Compact Cities: Sustainable urban forms for
4
developing countries. 245254.
5 6
RI PT
1
[2] J.A. Voogt, T.R. Oke. (2003). Thermal remote sensing of urban climates, Remote Sens. Environ. 86(3) 370-384.
[3] K.S. Lam, D.W.T. Chan, E.H.W. Chan, C.T. Tai, W.Y. Fung, K.C. Law. (2006). Infiltration
8
of outdoor air in two newly constructed high rise residential buildings, Int. J.
9
Vent. 5(2) 249-258.
M AN U
SC
7
[4] R. Giridharan, S.S.Y. Lau, S. Ganesan, B. Givoni. (2007). Urban design factors
11
influencing heat island intensity in high-rise high-density environments of
12
Hong Kong, Build. Environ. 42(10) 3669-3684.
TE D
10
[5] K.L. Lau, E. Ng. (2013). An investigation of urbanization effect on urban and rural
14
Hong Kong using a 40-year extended temperature record, Landscape Urban
15
Plan. 11442-52.
17 18
[6] A.P. Kamoutsis, A.S. Matsoukis, K.I. Chronopoulos. (2013). Bioclimatic conditions
AC C
16
EP
13
under different ground cover types in the greater Athens area, Greece, Global Nest Journal, 15(2) (2013) 254-260.
19
[7] L. Liu, Y. Lin, J. Liu, L. Wang, D. Wang, T. Shui, X. Chen, Q. Wu. (2017). Analysis of
20
local-scale urban heat island characteristics using an integrated method of
21
mobile measurement and GIS-based spatial interpolation, Building and
22
Environment 117 191-207. 21
ACCEPTED MANUSCRIPT 1 2 3
[8] M. Hodul, A. Knudby, H.C. Ho. (2016). Estimation of Continuous Urban Sky View Factor from Landsat Data Using Shadow Detection, Remote Sens. 8(7) 568. [9] J.E. Nichol, P.H. To. (2012). Temporal characteristics of thermal satellite images for
5
urban heat stress and heat island mapping, ISPRS J. Photogramm. 74,
6
153-162.
RI PT
4
[10] J. Yang, M.S. Wong, M. Menenti, J. Nichol. (2015). Modeling the effective
8
emissivity of the urban canopy using sky view factor, ISPRS J.
9
Photogramm.105 211-219.
M AN U
SC
7
[11] J.P. Connors, C.S. Galletti, W.T. Chow. (2013). Landscape configuration and urban
11
heat island effects: assessing the relationship between landscape
12
characteristics and land surface temperature in Phoenix, Arizona, Landscape
13
ecol. 28(2) 271-283.
16 17
thermal analysis, Remote Sens. Environ. 104(2) 123-132.
EP
15
[12] R. Gluch, D.A. Quattrochi, J.C. Luvall. (2006). A multi-scale approach to urban
[13] L. Liu, Y. Zhang. (2011). Urban Heat Island Analysis Using the Landsat TM Data
AC C
14
TE D
10
and ASTER Data: A Case Study in Hong Kong, Remote Sens. 3(12) 1535-1552.
18
[14] J.E. Nichol, W.Y. Fung, K.-s. Lam, M.S. Wong. (2009). Urban heat island diagnosis
19
using ASTER satellite images and ‘in situ’ air temperature, Atmos. Res. 94(2)
20
276-284.
21
[15] D.A. Quattrochi, J.C. Luvall. (1999). Thermal infrared remote sensing for analysis
22
of landscape ecological processes: Methods and applications, Landscape Ecol. 22
ACCEPTED MANUSCRIPT 1
14(6) 577-598. [16] Q. Weng, D. Lu, J. Schubring. (2004) Estimation of land surface temperature–
3
vegetation abundance relationship for urban heat island studies, Remote
4
Sens. Environ. 89(4) 467-483.
RI PT
2
[17] M.S. Wong, F. Peng, B. Zou, W.Z. Shi, G.J. Wilson. (2016) Spatially Analyzing the
6
Inequity of the Hong Kong Urban Heat Island by Socio-Demographic
7
Characteristics, Int. J. Environ. Res. Public Health 13, 317.
SC
5
[18] L. Chen, E. Ng. (2011). Quantitative urban climate mapping based on a
9
geographical database: A simulation approach using Hong Kong as a case
10
M AN U
8
study, Int. J. Appl. Earth OBS 13(4) 586-594.
[19] M. Aminipouri, A. Knudby, H.C. Ho. (2016) Using multiple disparate data sources
12
to map heat vulnerability: Vancouver case stud, The Canadian Geographer/Le
13
Géographe canadien 60(3) 356-368.
TE D
11
[20] W. Zhan, Y. Chen, J. Zhou, J. Wang, W. Liu, J. Voogt, X. Zhu, J. Quan, J. Li. (2013).
15
Disaggregation of remotely sensed land surface temperature: literature
16
survey, taxonomy, issues, and caveats, Remote Sens. Environ. 131, 119-139.
AC C
EP
14
17
[21] R. Giridharan, S. Ganesan, S.S.Y. Lau. (2004). Daytime urban heat island effect in
18
high-rise and high-density residential developments in Hong Kong, Energy
19 20 21 22
Buildings 36(6), 525-534.
[22] Population Census. (2011). 2011 Population Census Office Census and Statistics Department [23] M.S. Wong, J. Nichol, E. Ng. (2011). A study of the “wall effect” caused by 23
ACCEPTED MANUSCRIPT 1
proliferation of high-rise buildings using GIS techniques, Landscape Urban
2
Plan. 102(4) 245-253. [24] H.C. Ho, K.K.-L. Lau, C. Ren, E. Ng. (2017). Characterizing prolonged heat effects
4
on mortality in a sub-tropical high-density city, Hong Kong, Int. J. Biometer.
5
DOI: 10.1007/s00484-00017-01383-00484.
RI PT
3
[25] W.B. Goggins, E.Y. Chan, E. Ng, C. Ren, L. Chen. (2012). Effect modification of the
7
association between short-term meteorological factors and mortality by
8
urban heat islands in Hong Kong, PLoS One 7(6) e38551.
10
M AN U
9
SC
6
[26] X. Yang, Y. Li, Z. Luo, P.W. Chan. (2017). The urban cool island phenomenon in a high-rise high-density city and its mechanisms, Int. J. Climatol. 37(2) 890-904. [27] A.C.S. Lai, A.C. So, S.K.C. Ng, D. Jonas. (2012). The territory-wide airborne light
12
detection and ranging survey for the Hong Kong Special Administrative Region,
13
In The 33RD Asian Conference on Remote Sensing (pp 26-30)
TE D
11
[28] Z. Qin, A. Karnieli, P. Berliner. (2001). A mono-window algorithm for retrieving
15
land surface temperature from Landsat TM data and its application to the
16
Israel-Egypt border region, Int. J. Remote Sens. 22(18) 3719-3746.
18
AC C
17
EP
14
[29] L.W. Abreu, G.P. Anderson. (1996). The MODTRAN 2/3 report and LOWTRAN 7 model, Contract 19628(91-C)(0132)
19
[30] Z. Qin, A. Karnieli, P. Berliner. (2001). A mono-window algorithm for retrieving
20
land surface temperature from Landsat TM data and its application to the
21
Israel-Egypt border region, Int. J. Remote Sens. 223719-3746.
22
[31] L. Geosystems. (2002). Stereo analyst user’s guide, Leica Geosystems GIS & 24
ACCEPTED MANUSCRIPT
4
urban environments, J. Climatol. 7(2) 193–197. [33] ZRC SAZU (2010) Institute of Anthropological and Spatial Studies ZRC SAZU.
5
Sky-View
6
Accessed 12.20 13
7
Factor.
.
RI PT
3
[32] I.D. Watson, G.T. Johnson. (1987). Graphical estimation of sky view-factors in
http://iaps.zrc-sazu.si/index.php?q=en//node/138.
[34] ITT Visual Information Solutions (2010) ENVI Software: Image processing &
8
analysis
solutions.
9
Accessed 11.09 13
SC
2
Mapping Division, Atlanta, GA, USA
http://www.ittvis.com/ProductServices/ENVI.aspx.
M AN U
1
[35] F. Peng, M.S. Wong, Y. Wan, J.E. Nichol. (2017). Modeling of urban wind
11
ventilation using high resolution airborne LiDAR data, Comput. Environ.
12
Urban Syst. 64, 81-90.
TE D
10
[36] R.L. Hwang, T.P. Lin, A. Matzarakis. (2011). Seasonal effects of urban street
14
shading on long-term outdoor thermal comfort, Build. Environ. 46(4)
15
863-870.
17
[37] T.P. Lin, A. Matzarakis, R.L. Hwang. (2010). Shading effect on long-term outdoor
AC C
16
EP
13
thermal comfort, Build. Environ. 45(1) 213-221.
18
[38] E. Ng. (2009). Policies and technical guidelines for urban planning of high-density
19
cities–air ventilation assessment (AVA) of Hong Kong, Build. Environ. 44(7)
20
1478-1488.
21
[39] Hang, J., Li, Y., Sandberg, M., Buccolieri, R., & Di Sabatino, S. (2012). The
22
influence of building height variability on pollutant dispersion and pedestrian 25
ACCEPTED MANUSCRIPT 1
ventilation in idealized high-rise urban areas. Building and Environment, 56,
2
346-360. [40] Blocken, B., Carmeliet, J., & Stathopoulos, T. (2007). CFD evaluation of wind
4
speed conditions in passages between parallel buildings—effect of
5
wall-function roughness modifications for the atmospheric boundary layer
6
flow. Journal of Wind Engineering and Industrial Aerodynamics, 95(9),
7
941-962.
SC
RI PT
3
[41] T.E. Morakinyo, L. Kong, K.K.L. Lau, C. Yuan, E. Ng. (2017). A study on the impact
9
of shadow-cast and tree species on in-canyon and neighborhood's thermal
10
M AN U
8
comfort, Build. Environ. 115(2017) 1-17.
[42] H. C. Ho, A. Knudby, B.B. Walker, Henderson, S. B. (2017). Delineation of spatial
12
variability in the temperature–mortality relationship on extremely hot days in
13
greater Vancouver, Canada. Environmental health perspectives, 125(1), 66.
TE D
11
[43] H. C. Ho, A. Knudby, Y. Xu, M. Hodul, & M. Aminipouri. (2016). A comparison of
15
urban heat islands mapped using skin temperature, air temperature, and
16
apparent temperature (Humidex), for the greater Vancouver area. Science of
AC C
17
EP
14
the Total Environment, 544, 929-938.
18
[44] P. C. Lai, C.C. Choi, P.P. Wong, T.Q. Thach, M.S. Wong, W. Cheng, A. Krämerd,C.M.
19
Wong (2016). Spatial analytical methods for deriving a historical map of
20
physiological equivalent temperature of Hong Kong. Building and
21
Environment, 99, 22-28.
22
[45] T. Q. Thach, Q. Zheng, P. C. Lai, P. P. Y. Wong, P. Y. K.Chau, H.J. Jahn, D. Plass, L. 26
ACCEPTED MANUSCRIPT 1
Katzschner, A. Kraemer, C.M. Wong (2015). Assessing spatial associations
2
between thermal stress and mortality in Hong Kong: A small-area ecological
3
study. Science of the Total Environment, 502, 666-672.
AC C
EP
TE D
M AN U
SC
RI PT
4 5 6
27
ACCEPTED MANUSCRIPT 1
Table 1. Landsat 5 TM images used in this study. These Landsat 5 TM images were
2
acquired on cloud-free or semi-cloud-free dates across the study site.
3 4
7% 40% 0% 0% 2% 24% 1% 70% 35% 11% 3% 12% 38% 28% 34%
AC C
EP
TE D
5
Cloud Cover
RI PT
Sun Elevation (degree) 38.75 38.14 36.10 41.00 55.09 36.48 36.15 40.22 46.13 38.55 49.30 41.18 56.34 48.75 37.20
SC
Acquired Time (Greenwich Mean Time, GMT) 02:22 02:16 02:16 01:58 02:02 02:13 02:21 02:26 02:25 02:30 02:40 02:40 02:43 02:36 02:42
M AN U
Acquired Date 24 Nov 1988 3 Dec 1988 19 Dec 1988 3 Mar 1996 19 Aug 1996 30 Nov 1996 3 Jan 2000 12 Feb 2000 2 Nov 2000 16 Jan 2005 22 Oct 2005 23 Nov 2005 26 Mar 2010 29 Oct 2010 23 Dec 2010
28
ACCEPTED MANUSCRIPT 1
Table 2. Aerial photographs used in this study. Number of Aerial Year
1964
6
1975
8
1980
8
1985
8
1990
8
1995
8
2000
8
2005 2010
8 9
AC C
EP
TE D
2 3
SC
6
M AN U
1945
RI PT
Photographs
29
ACCEPTED MANUSCRIPT 1 2 3
Table 3. Validation of daily LST estimation. Results indicated a comparison estimated LST and observed LST, at the date with a cloud-free scene across the study site.
RI PT
4
Pearson correlation coefficient (R)
RMSE (°C)
MAE (°C)
3 Dec 1988
0.89
0.62
0.49
30 Nov 1996
0.64
2.71
2.47
3 Jan 2000
0.68
1.40
1.02
22 Oct 2005
0.89
1.56
1.32
29 Oct 2010
0.81
2.98
2.60
AC C
EP
TE D
M AN U
SC
Date
30
ACCEPTED MANUSCRIPT Table 4. Validation of 3D building data. Results indicated 1) a comparison between estimated building data and the building height predicted by LiDAR across open spaces, and 2) a comparison between estimated building data and the building height predicted by LiDAR across areas with buildings. Areas with Buildings RMSE (m) MAE (m) 3.65 2.83 3.12 2.18 2.59 1.92 3.81 2.52 3.63 2.21 2.20 1.66 5.92 3.25 1.06 0.83 1.19 0.81
RI PT
1964 1975 1980 1985 1990 1995 2000 2005 2010
Open Spaces RMSE (m) MAE (m) 1.46 1.23 0.93 0.71 0.82 0.74 1.39 1.13 1.01 0.79 1.13 0.85 0.78 0.41 0.86 0.60 1.06 0.82
SC
Year
M AN U
1 2 3 4 5
AC C
EP
TE D
6 7 8 9
31
ACCEPTED MANUSCRIPT Table 5. Spatiotemporal assessment of building morphology and urban microclimate. Results indicate the additional reduction of relative temperature by air ventilation and shading effects, compared to the areas with improvement of ventilation and shading through years.
Cooling
Averaged relative
Effects
temperature (°C)
Comparison (1988 - 1996) Comparison (1996- 2000) Comparison (2000- 2005) Comparison (2005- 2010)
1988
1996
1996
2000
2000
2005
2005
2010
3.56
2.47
2.76
4.66
4.25
4.26
4.62
5.75
3.96
1.78
2.32
4.32
3.52
4.08
Identified areas without a change of wind speed between
Air Ventilation comparison Identified areas with
speed between two years of comparison Additional reduction of relative
1.09
temperature from air ventilation (°C)
Effects
temperature (°C)
Identified areas
4.43
0.27
0.12
0.57
Comparison (1988 - 1996) Comparison (1996- 2000) Comparison (2000- 2005) Comparison (2005- 2010)
TE D
Averaged relative
3.95
M AN U
an increase of wind
SC
two years of
Cooling
RI PT
1 2 3 4 5
1988
1996
1996
2000
2000
2005
2005
2010
3.63
2.67
2.56
4.24
3.9
3.78
4.01
5.36
3.5
1.78
2.05
3.23
3.68
2.92
3.2
3.81
without a change of
shades between two
EP
years of comparison Shading
Identified areas with
AC C
an increase of shades
between two years of comparison
Additional reduction of relative 0.76
0.6
0.64
0.74
temperature from shading (°C)
6 7
32
ACCEPTED MANUSCRIPT 1 2
Captions of Figures
3
Figure 1. Study area: Kowloon Peninsula, Hong Kong.
5
Figure 2. Flow diagram of the model design of this study.
6
RI PT
4
Figure 3. Trend of daily LST from MODIS LST products in 2010. Blue, red, bright blue
8
and yellow represent four selected grids on the MODIS LST product. X-axis is the day
9
of a year and y-axis is the LST of a selected grid.
M AN U
SC
7
10
Figure 4. Quadratic function to estimate daily air temperature. X-axis is the day of a
12
year and y-axis is the air temperature at HKO station in 2010.
13
TE D
11
Figure 5. Flow chart of the rule-based model. This rule-based model is designed for
15
estimating the additional reduction of relative temperature contributed by air
16
ventilation and shades.
AC C
17
EP
14
18
Figure 6. 3D built environment of Kowloon Peninsula between 1964 and 2010. The
19
followings are the years of each map of built environment represents: (a) 1964, (b)
20
1975, (c) 1980, (d) 1985, (e) 1990, (f) 1995, (g) 2000, (h) 2005, and (i) 2010.
21 22
Figure 7. SVF map retrieved from 3D building data, including the year of (a) 1990, (b)
33
ACCEPTED MANUSCRIPT 1
1995, (c) 2000, (d) 2005, and (e) 2010.
2
Figure 8. Relative temperature maps of Kowloon Peninsula, including the year of (a)
4
1988, (b) 1996, (c) 2000, (d) 2005, and (e) 2010.
RI PT
3
5
Figure 9. Maps of major air ventilation paths across the Kowloon Peninsula at the
7
year of (a) 1988, (b) 1996, (c) 2000, (d) 2005, and (e) 2010. The base map of these
8
figures are the relative temperature maps of Kowloon Peninsula in each year, for
9
comparing the effects of wind on intra-urban heat.
M AN U
SC
6
10
Figure 10. Examples of shading effects across the study site. Areas with relatively
12
taller buildings across the study site in 2010 have been selected and marked.
13
Selected areas with shades indicate a lower daytime temperature compared to areas
14
without a shade.
EP
16
AC C
15
TE D
11
34
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
RI PT
SC M AN U TE D
-
EP
-
Reconstructed the historical temperature maps across Kowloon Peninsula based on Landsat images Reconstructed 3D built environment with aerial photos Areas with more shades through years have greater contribution to reduce local heat Areas with more ventilation through years have greater contribution to reduce local temperature Larger variation of building heights is suggested to integrate both shading effects and air ventilation for developing a compact city
AC C
-
S0360-1323(17)30337-2
DOI:
10.1016/j.buildenv.2017.07.038
Reference:
BAE 5017
To appear in:
Building and Environment
Received Date: 24 June 2017 Revised Date:
24 July 2017
Accepted Date: 25 July 2017
Please cite this article as: Peng F, Wong MS, Ho HC, Nichol J, Chan PW, Reconstruction of historical datasets for analyzing spatiotemporal influence of built environment on urban microclimates across a compact city, Building and Environment (2017), doi: 10.1016/j.buildenv.2017.07.038. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT 1
Title: Reconstruction of Historical Datasets for Analyzing Spatiotemporal Influence
2
of Built Environment on Urban Microclimates across a Compact City
3
Authors: Fen Penga, Man Sing Wonga,*, Hung Chak Hoa, Janet Nichola, Pak Wai Chanb
RI PT
4 5 6
Affiliations:
7
a
8
University, Kowloon, Hong Kong
9
b
M AN U
SC
Department of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic
Hong Kong Observatory, Hong Kong
10
*Corresponding Author: Man Sing Wong, Department of Land Surveying and
12
Geo-Informatics, The Hong Kong Polytechnic University, Kowloon, Hong Kong.
13
Email: [email protected]
EP
15
AC C
14
TE D
11
1
ACCEPTED MANUSCRIPT 1
Abstract The high-rise/high-density environment of a compact city can influence the
3
microclimate resulting in lower living quality. Previous studies have analyzed the
4
relationships between high-rise/high-density environment and microclimates, by
5
either a temporal study or a spatial approach, while a strategy for investigating the
6
spatiotemporal relationship has yet to be developed. This study initiated a set of
7
innovative strategies to map the historical built environment/microclimates of a
8
compact city, with a spatiotemporal approach to analyze the relationships between
9
building structures and urban climates, for developing a sustainable protocol for
10
future urban planning. Three major components were reconstructed, including 1) the
11
annually averaged Land Surface Temperature (LST) for determining the relative
12
temperature across a compact city; 2) 3D building datasets for representing the
13
building morphology; and 3) sets of urban morphological data derived from building
14
datasets for analyzing microclimate and thermal distress.
TE D
M AN U
SC
RI PT
2
There are high correlations between observed and predicted LSTs (R = 0.64 to
16
0.89), with mean absolute error (MAE) of annually averaged LST ranging 0.49°C to
17
2.60°C, and root mean square error (RMSE) ranging 0.62°C to 2.98°C. There are low
18
errors for reconstructing building data, in which MAEs and RMSEs of an open space
19
are 0.41m - 1.23m and 0.78m - 1.46m; and for an area with buildings are 0.81m -
20
3.25m and 1.06m - 5.92m. The spatiotemporal estimation indicated areas with
21
improved air ventilation through years can significantly reduce an additional 0.12°C -
22
1.09°C than the areas without improvement, while areas with an increase in shades
AC C
EP
15
2
ACCEPTED MANUSCRIPT 1
through years have 0.6°C to 0.76°C higher reduction of relative temperature.
2
Keywords: historical built environment; temperature; shading effect; air ventilation;
4
urban design; spatial analytics
RI PT
3
AC C
EP
TE D
M AN U
SC
5
3
ACCEPTED MANUSCRIPT 1 2
1. Introduction Population growth and limited natural resources are two global challenges since
4
the twentieth century. As a result, an urban form with high-rise and high-density
5
urban settings, named as “compact city”, has since been in practice in the
6
developing countries for minimizing population stress [1]. Although the idea of a
7
“compact city” was originally an urban design for developing a more sustainable
8
environment, the high-rise and high-density settings of these cities have also formed
9
a unique building morphology that can influence the microclimate across urban
10
areas [2,3]. Previous studies for sustainable planning have focused on applying
11
temporal analyses to study the relationship between urban climates and building
12
structures [4-7] for overcoming the planning issues from a problematic urban design
13
(e.g. building with higher thermal distress, living environment with poor air
14
ventilation).
TE D
M AN U
SC
RI PT
3
The temporal analysis can partially provide fundamental understanding for
16
sustainable urban planning. For example, Lau and Ng [5] used only five weather
17
stations of Hong Kong to examine the interaction between built environments and
18
air temperature, and found a cooling effect of high-density environment due to the
19
decline of urbanization. However, there is a limitation of ordinal temporal study, in
20
which these studies only used several locations of a city as demonstration, leading to
21
a potential bias of not being able to represent the holistic urban built environment
22
across a compact city. Thus, an alternative of using remotely sensed images has then
AC C
EP
15
4
ACCEPTED MANUSCRIPT been applied, for mapping built environment and for analyzing urban microclimates
2
[8-10]. This technique is shown to be promising, since satellite images or aerial
3
photographs increase the spatial coverage of an urban climatic analysis. A common
4
example is the application of thermal remote sensing for effectively mapping land
5
surface temperature (LST) across a city, resulting to improve the results of analysis of
6
urban heat island effect in many small areas over a large region [11-17], while the
7
traditional method is limited by sparse distribution of weather stations across a
8
compact city. This advantage has led to a wide-use of remote sensing techniques for
9
monitoring the urban environment, for sustainable policy making, and for urban
SC
M AN U
10
RI PT
1
planning [18, 19].
However, the use of remote sensing techniques may also be a trade-off for
12
temporal resolution [20]. For example, using satellite images to retrieve land features
13
are restricted by the satellite overpass and cloud coverage, and the applications of
14
aerial photographs are restricted by the number of flights across the city. It is
15
necessary to develop a set of techniques to reconstruct the historical datasets that
16
can monitor the changing urban environment with high spatial and temporal quality,
17
in order to investigate the relationship between building morphology and
18
microclimate for developing sustainable planning protocols.
EP
AC C
19
TE D
11
This study hereby develops a set of spatiotemporal models that can accurately
20
map the urban built environment with high spatial resolution for each year, including
21
1) annually averaged LST indicating the spatial variability of urban heat across a
22
high-density environment, 2) geodatabase with 3-dimensional (3D) building data that 5
ACCEPTED MANUSCRIPT can fully represent the spatial patterns of historical urban morphology, and 3) a set of
2
spatial data derived from the 3D building data that can be used as building
3
morphological parameters for urban climatic study. A spatiotemporal analysis was
4
also conducted to estimate the impact of building morphology on urban
5
microclimate between 1988 and 2010 across Kowloon Peninsula, Hong Kong. Based
6
on previous studies, which has suggested that high-rise and high-density
7
environments have significantly contributed to urban microclimates [5, 21] Kowloon
8
Peninsula was selected as the site for this study. It is expected that a data-driven
9
analysis of Kowloon Peninsula can be a global example of urban climatic studies for
SC
M AN U
10
RI PT
1
compact cities.
11
2. Study Area
TE D
12
The Kowloon Peninsula is an extensively urbanized area in Hong Kong (Figure 1),
14
with approximately 1 million people living in a relatively small region (~ 26 km2) [22].
15
Kowloon Peninsula has a typical setting of a compact city, where the majority of
16
lands consist of high-rises and is a high-density environment with very few urban
17
greenspaces to be found across the study site. A “wall effect” controlled by building
18
morphology has been recognized as one of the main effects that can block sea
19
breezes for reducing thermal distress during a hot summer day [23]. Influencing by
20
the subtropical climate, there are multiple Very Hot Days (>= 33°C) and Hot Night
21
(>=28°C) as observed by the weather station located at the Hong Kong Observatory
22
(HKO) in Kowloon Peninsula [24]. During these extremely hot days and nights, urban
AC C
EP
13
6
ACCEPTED MANUSCRIPT heat islands with extremely high temperatures can occur across the high-density
2
environment and, as a result, these can induce the significance of thermal
3
discomfort, morbidity risks, and even mortality risks for its inhabitants [9, 21, 25].
4
However, a study focusing on intra-urban temperature difference has also observed
5
a phenomenon of “urban cool island” across this high-density environment [26].
6
Thus it is necessary for the spatial pattern of microclimate across Kowloon Peninsula
7
to be further explored, especially as the site area contains a high percentage of
8
poorly maintained aged buildings (>= 50 years) that can influence the microclimatic
9
variations as a result of urban renewal. A spatiotemporal analysis to study the
10
influence of building morphology on urban climate across Kowloon Peninsula is
11
necessary since it is important to develop sustainable protocols prior to future urban
12
planning.
13
3. Data and Methods
15
3.1 Data
EP
14
TE D
M AN U
SC
RI PT
1
There are three major analyses to reconstruct historical datasets for this urban
17
climatic study (Figure 2), including 1) annually averaged LST estimation, 2) 3D
18
reconstruction of building data, and 3) retrieval of building morphology data based
19
on results from the aforementioned analyses.
AC C
16
20
In this study, fifteen Landsat 5 TM satellite images from the years 1988, 1996,
21
2000, 2005 and 2010, with three corresponding images per year, of cloud-free scene
22
across the study area were used to estimate the annually averaged LST (Table 1). The 7
ACCEPTED MANUSCRIPT hourly air temperature data between 1885 and 2010 from the weather station
2
located at the HKO were used to improve the LST estimation. Data of hourly wind
3
direction of the HKO station between 1988 and 2010 were also retrieved for
4
estimating the ventilation paths. In addition, seventy-seven aerial photographs taken
5
between 1964 and 2010 were used to reconstruct the built environment of the
6
Kowloon Peninsula (Table 2). Acquired from the Hong Kong Lands Department, these
7
aerial photographs contain at least 83 control points with 3D information that can be
8
used to accurately reconstruct the building structures across the high-density
9
environment. Airborne light detection and ranging (LiDAR) data of 2010 and 2011
10
[27] were applied for generation of a digital surface model (DSM) and a digital
11
elevation model (DEM) with 1-meter resolution, which can be used to validate the
12
3D historical building data reconstructed by aerial photographs.
14
SC
M AN U
TE D
13
RI PT
1
3.2 Reconstruction of Annually Averaged LST A mono-window algorithm was applied to estimate the LST, based on thermal
16
infrared (TIR) band of each Landsat 5 TM image [28], with additional parameters of
17
upwelling and downwelling atmospheric radiances, atmospheric transmittance, and
18
surface emissivity retrieved from the MODTRAN 4.0 code [29]. By using these LST
19
images, a quadratic function was applied to estimate the daily LST from 1988 to 2010.
20
The concept of using a quadratic function for estimation was first developed based
21
on a quick comparison of LST and the day of year using four pixels (1 km x 1 km)
22
randomly selected on all MODerate resolution Imaging Spectroradiometer (MODIS)
AC C
EP
15
8
ACCEPTED MANUSCRIPT 1
LST products in 2010 (Figure 3), with results showing a quadratic trend between daily
2
LSTs and day of a year. Two of three Landsat images of each year were used to
3
construct a quadratic function: Y = −a ∗
+
+
5
Y = −a ∗
− ⁄ 2 ∗ )) +
+4∗
RI PT
∗
(1)
4
∗ )⁄ 4 ∗ )
(2)
where
is the day of the year, Y is the LST at pixel i, and a, b, c are the
7
coefficients of the quadratic function. Based on the quadratic distribution of air
8
temperatures at the HKO station through a year (Figure 4),
9
defined as the hottest day of a year. Therefore, a day of the year with the hottest
10
temperature was substituted in the Eq. (2) for estimating the daily LST with the
11
information from selected Landsat images.
⁄ 2 ∗ ) can be
TE D
M AN U
SC
6
In order to validate the quadratic function, the estimated LSTs at the day of
13
another Landsat image were compared with the retrieved LST [30]. All pixels on the
14
images were compared, with measures of root mean square error (RMSE) and mean
15
absolute error (MAE), for determining the accuracy of the quadratic function. After
16
validation, the results of each year were summarized to be an annually averaged LST
17
image. Each annually averaged LST map was converted to relative temperature map
18
using Kowloon Park as a reference. The use of Kowloon Park as a reference is due to
19
its extensive greenspace, while a comparison of surface temperature between
20
Kowloon and other areas can help estimate the potential thermal discomfort
21
influenced by urban morphology. Based on these relative temperature maps, the
AC C
EP
12
9
ACCEPTED MANUSCRIPT 1
spatial variability of temperature across the study area based on the temperature
2
difference from Kowloon Park can be depicted.
3
3.3 Reconstruction of 3D building data
RI PT
4
ERDAS Leica Photogrammetry Suite module with a stereo image matching
6
technique was applied to 1) identify dense matching points between all aerial
7
photographs; 2) estimate the 3D coordinates of each matching point; and 3)
8
interpolate all points for building a DSM that can represent the building environment
9
across the Kowloon Peninsula [31].
M AN U
SC
5
DSM generated by the stereo image matching technique (DSMPHG) was then
11
subtracted by a DSM generated from the airborne LiDAR data (DSMLiDAR) for
12
validation. A resultant pixel closer to zero from this subtraction indicated less
13
differences between DSMs and a higher accuracy of DSMPHG. In order to validate the
14
DSMPHG comprehensively, four sets of randomly distributed samples, with 100
15
samples of each set, were applied to determine the accuracy of DSMPHG across open
16
spaces and areas with buildings.
EP
AC C
17
TE D
10
To reconstruct the 3D building data, orthophotos and DSMs were used to
18
digitize the building footprints of the Kowloon Peninsula. Building footprints of 2010
19
from the Hong Kong Lands Department were used as the reference for digitization.
20
Building heights were estimated by subtracting the DSM from the LiDAR DEM and
21
were used as a reference to identify the possible areas with buildings during
10
ACCEPTED MANUSCRIPT 1
digitization. Lastly, the 3D building models were developed using ESRI ArcScene by
2
combining building footprints and building heights.
3
3.4. Reconstruction of urban morphological data
RI PT
4
Sky view factor (SVF) and ventilation paths were two major components that can
6
influence microclimate of urban areas. SVF is a ratio measuring obstructed sky of the
7
surrounding environment, which can indicate the amount of solar radiation received
8
by a planar surface and the entire hemispheric environment [32]. The ventilation
9
path is a ventilation corridor controlled by wind direction, wind speed, and building
10
structure. In this study, SVF was estimated based on a plugin module [33] of
11
ENVI+IDL [34], by a 90m searching radius on each pixel of each DSM between 1990
12
and 2010. Finally, ventilation paths were estimated based on 3D building data
13
between 1988 and 2010 by the method noted on Peng et al. [35] with the use of
14
wind data from HKO.
17 18
3.5. Spatiotemporal analyses for estimating influences of urban morphology on
AC C
16
EP
15
TE D
M AN U
SC
5
microclimate
Based on previous studies, shading effect and the air ventilation are two major
19
components controlled by the urban morphology that can influence microclimate
20
[36 – 38]. Therefore, two case studies for spatiotemporally analyzing impacts of air
21
ventilation and shades on surface temperature have been conducted in this study.
22
Annually average LSTs at the year with cloud-free Landsat images (1988, 1996, 2000, 11
ACCEPTED MANUSCRIPT 2005, and 2010) were used for the spatiotemporal comparison. These Landsat
2
images represent the cloud-free data with the closest year to the building datasets
3
retrieved from aerial photographs. To estimate the impact of urban ventilation on
4
temperature, a rule-based model was first used to identify open areas with and
5
without changes in air ventilation between two known years of the study.
6
Normalized wind speeds of 1988, 1996, 2000, 2005, and 2010 were estimated for
7
demonstrating air ventilation across the Kowloon Peninsula. In order to map the
8
area without significant change in air ventilation, the following rules (Figure 5) have
9
been used to a) identify areas with change of building height <= 5 meters in between
10
two known years; b) identify areas with SVF >= 0.9 for two known years for
11
hypothesizing an environment with more unchanged open spaces; c) map areas with
12
a change of SVF <= 0.05 between years to further identify the unchanged
13
environment; and d) identify an area with a change of normalized wind speed <=0.05
14
m/s between years to map the open area without changes in air ventilation. These
15
areas without changes in air ventilation were compared with the areas with
16
significant changes determined by the followings: i) map of unchanged open area
17
using a), b), and c) above, and ii) identification of the area with significant change of
18
air ventilation by considering all areas with at least 0.2 m/s increase of normalized
19
wind speed.
AC C
EP
TE D
M AN U
SC
RI PT
1
20
These results of rule-based model were further applied to an equation for
21
quantifying the spatiotemporal relationship between wind speed and relative
22
temperature. 12
ACCEPTED MANUSCRIPT 1
=
−
)−
−
)
(3)
where Temp represents the spatiotemporal difference in relative temperature (°C)
3
between areas with and without observed changes; t represents the year of latter
4
Landsat scene; t-1 represents year of earlier Landsat scene, SCt represents the
5
averaged relative temperature (°C) of all areas with observed changes at t, SCt-1
6
represents the averaged relative temperature (°C) of all areas with observed changes
7
at t-1; SUt represents the averaged relative temperature (°C) of areas without
8
observed change at t; and SUt-1 represents the averaged relative temperature (°C) of
9
all areas without observed change at t-1. In brief, this equation can estimate the
10
difference in relative temperature change across the Kowloon Peninsula due to the
11
observed increase in wind speed.
M AN U
SC
RI PT
2
For estimating the shading effects on surface temperature, two new sets of the
13
rule-based model (Figure 5) were used for identifying areas with and without
14
changes of shades. The areas without changes of shades were mapped by the
15
followings: a) identification of areas with the change of building height <= 5m
16
between years; b) identification of areas with the change of normalized wind speed
17
<= 0.05; and c) map of areas with an absolute change of SVF <=0.05. These areas
18
were compared with the area with a significant increase in shades, determined by
19
the rule a) and b) above, and an additional rule of mapping areas with SVF at least
20
0.1 lower after years. The rules a) and b) are to set control for the rule-based model,
21
thus the datasets for comparison only have the difference in shading effect. This
AC C
EP
TE D
12
13
ACCEPTED MANUSCRIPT method can minimize the bias from other environmental influences such as wind.
2
For comparing the areas with and without changes of shade, Eq. (3) was applied to
3
quantify the influence of shading effect, where the results can be expressed as a
4
difference in relative temperature change across the Kowloon Peninsula due to the
5
observed increase of shading.
6
4. Result
8
4.1. Validation of spatial data reconstruction
M AN U
7
SC
RI PT
1
LST estimation of this data reconstruction model is fairly accurate, with MAE
10
ranging between 0.49°C and 2.60°C and RMSE ranging between 0.62°C and 2.98°C
11
(Table 3). There is also a higher correlation between the estimated LST and the
12
observed LST (R = 0.64 to 0.89) for the results of 1988 and 2010. In addition, the
13
spatiotemporal trend of annually average LSTs was similar to the temporal change of
14
air temperature in the Kowloon Peninsula. Among the five years with retrieved
15
satellite images, the yearly average of maximum air temperature of 2005 was the
16
highest (35.4°C), followed by 1996 (34.3°C), 2000 (34.2°C), 2010 (34.1°C) and 1988
17
(33.8°C). In comparison, the results of annually average LSTs also showed that the
18
highest temperature occurred in 2005, with an average of 31.3°C for the whole study
19
area, compared to the lowest temperature (28.4°C) in 1988. A high accuracy of
20
reconstructing 3D building environment between 1964 and 2010 was also found in
21
this study (Figure 6). For an open space, MAE ranged from 0.41m to 1.23m and
22
RSME ranged from 0.78m to 1.46m (Table 4). For an area with buildings, MAE ranged
AC C
EP
TE D
9
14
ACCEPTED MANUSCRIPT from 0.81m to 3.25m, and RMSE ranged from 1.06m to 5.92m. Considering the
2
high-rise buildings (>100m) that can be found in the Kowloon Peninsula, high
3
accuracy of this dataset can be used as data input for mapping urban morphology
4
such as SVF and ventilation paths (Figure 7).
RI PT
1
5 6
4.2. Spatiotemporal influences of building morphologies on urban microclimate
The “old” airport in the study area, namely the “Kai Tak International Airport”,
8
was the region with the highest surface temperature between 1988 and 2010 (Figure
9
8). Impervious surface of Kai Tak has been expanding since 1998. This expansion of
10
impervious surface has increased the areas with higher surface temperature across
11
the Kowloon Peninsula. Areas with higher surface temperature have expanded more
12
rapidly after 2000, due to the closure of the Kai Tak International Airport at 1998,
13
changing this area to bare land. A trend of increasing building heights across the
14
Kowloon Peninsula has also been observed in this study, in which the average
15
building height has increased by 1.95m for every ten years since 1964. There were
16
only 5% of buildings above 70m before 2000, while 16% of buildings were above 70m
17
in 2010. Increases in building coverage and heights from 1964 have also significantly
18
influenced the urban morphology. This change of urban morphology to a high-rise,
19
high-density environment has reduced the paths of air ventilation since 1988. It is
20
important to note that the areas with less air ventilation paths are usually of higher
21
temperature (Figure 9), as wind has a significant contribution to reduce the urban
22
temperature. Multiple spots of the Kowloon Peninsula have observed a significant
AC C
EP
TE D
M AN U
SC
7
15
ACCEPTED MANUSCRIPT change in temperature through the years as air ventilation paths across these areas
2
have changed. Under this changing environment, E-W street canyons were the
3
significant wind corridors with higher wind speed across the Kowloon Peninsula
4
through the years.
RI PT
1
In addition, the variation of the urban building not only developed a complex
6
system for air ventilation, but also enhanced the shading effect on the high-density
7
environment. This shading effect significantly reduced the relative temperature
8
(Figure 10).
M AN U
SC
5
The cooling effect of air ventilation and shading is more significant based on the
10
quantitative analyses with Eq. (3). Areas with an improvement of air ventilation
11
determined by increased wind speed can reduce the relative temperature of 0.12°C
12
to 1.09°C more than the areas without improvement of air ventilation (Table 5).
13
Areas with an increase in shading also have a 0.6°C to 0.76°C temperature reduction
14
more than the areas without changes in shading effect. These findings showed that
15
the effect of air ventilation has dropped after 1996 most likely due to a major change
16
of ventilation paths, while the spatiotemporal effect of shades on microclimate
17
across the Kowloon Peninsula has been more stable.
EP
AC C
18
TE D
9
19
5. Discussion
20
5.1. Sustainable design and planning for urban heat mitigation
21
In this study, a reconstruction of historical building and environmental datasets
22
was conducted for microclimate assessment. These data were applied to analyze 16
ACCEPTED MANUSCRIPT how urban morphology could influence urban heat. It is found that availability of
2
shading and air ventilation are the major elements for reducing intra-urban
3
temperature. These results provided an implication that a high variation of building
4
height is needed for maintaining both significant mitigating effects of air ventilation
5
and shades on urban heat [39].
RI PT
1
A compact city is occupied by a high-rise and high-density environment that can
7
continue to provide a shading effect for reducing incoming solar radiation. However,
8
some high-rise buildings of this high-density environment can directly block the air
9
ventilation resulting in a parcel of high-temperature trapped in the high-density
10
spots. Therefore, in order to maintain the air ventilation and at the same time
11
providing a high-density environment that can sustain the large population in urban
12
areas, it is necessary to build up the core of urban centers with both high and low
13
buildings; this difference in building heights can generate a fluid dynamic path that
14
allows more cold air parcels to disperse into the urban center, at the same time
15
maintains the shading effects from building heights on ground-level heat. One
16
strategy that has been proposed for practicing in high-rise buildings is the use of
17
multiple horizontal wind corridors at the higher-floors of the building [40], which
18
allows better air ventilation in one vertical profile of a high-rise building and at the
19
same time this strategy does not need to reduce the building height. Combining
20
these strategies of improving shading effects and air ventilation in a high-density
21
environment, the relative temperature of urban areas should be reduced. Suggested
22
by the results of this study, the daytime thermal distress of the study area can be
AC C
EP
TE D
M AN U
SC
6
17
ACCEPTED MANUSCRIPT reduced by enhancing the shading effects, and by improving air ventilation based on
2
more E-W horizon corridors in a vertical profile. These combined strategies can also
3
be practiced with other common sustainable planning protocols such as urban
4
greening, while the latter has found to be useful for reducing thermal discomfort [41].
5
In addition, a future study can be conducted to forecast the change of perceived
6
temperature (e.g. physiological equivalent temperature, apparent temperature)
7
based on the historical datasets created by this study, in order to target the land use
8
policies and sustainable urban designs that should be applied to reduce thermal
9
discomfort across urban areas in the coming decades. This study is important,
10
because temporally-static perceived temperature maps have found to be useful to
11
determine thermal discomfort and heat health risk across urban areas [42 – 45],
12
while a more comprehensive spatiotemporal modelling with historical data should
13
enhance the assessment of thermal issues and for sustainable urban planning in the
14
future.
17
5.2. Limitations
AC C
16
EP
15
TE D
M AN U
SC
RI PT
1
In this study, the diurnal effect of daytime/nighttime urban temperature has
18
not been discussed. It is due to the limitation of data of nighttime satellite images for
19
reconstructing
20
spatiotemporal analysis. Previous study has attempted to use ASTER satellite image
21
to estimate the spatial distribution of nighttime urban temperature [9]. However,
22
lack of cloud-free images and lack of acquisition in the earlier past years have limited
historical
dataset
and
for
conducting
a
comprehensive
18
ACCEPTED MANUSCRIPT the ability of using ASTER images to reconstruct historical datasets. A future study
2
combining machine learning with limited data of nighttime satellite images can be
3
conducted for testing the ability to spatiotemporally reconstruct the intra-urban
4
temperature data in a compact city.
RI PT
1
Another limitation of this study is that the annually averaged LSTs were used
6
instead of the summer average. Annually averaged LST was used because of the shift
7
of summer in Hong Kong due to climate change, as there were no fixed months of
8
summer in Hong Kong since 2000, as the number of hot days has been increased
9
rapidly, although the hottest days are still commonly found in July and August. For
10
reducing statistical bias, annually averaged LST can minimize the random cutoff of
11
season for study. In the perspective of thermal comfort study, relative temperature
12
predicted by annually averaged LST is better matched to the scenario that
13
temperature has been rising exponentially due to climate change and hot days will
14
eventually cover the whole year in a sub-tropical city in the near future.
15
17
6. Conclusions
AC C
16
EP
TE D
M AN U
SC
5
This study has developed a comprehensive approach to reconstruct historical
18
temperature maps based on Landsat images and aerial photographs. In the past
19
years, there have also been innovative techniques to reconstruct the building
20
morphology, with 3D information for clearly representing the change of building
21
environment in a compact city. A spatiotemporal analysis was also conducted for
22
evaluating the impacts of building morphology on microclimate across a compact city. 19
ACCEPTED MANUSCRIPT The results showed that shading effects and influence of air ventilation should be the
2
keys to reducing surface temperature; while in the study area, more air ventilation
3
should be built along with the continuing shading effects for better improving the
4
urban thermal environment.
RI PT
1
5 6
Acknowledgements
This work was supported in part by the grant Early Career Scheme (project id:
8
25201614) from the Research Grants Council of Hong Kong, and grant G-YM85 from
9
the Hong Kong Polytechnic University. The authors thank the Hong Kong Lands
10
Department for the historical aerial photographs, ground control points, and existing
11
building GIS data; the Hong Kong Observatory for the meteorological data; NASA LP
12
DAAC for the Landsat satellite imagery.
M AN U
TE D EP
14
AC C
13
SC
7
20
ACCEPTED MANUSCRIPT References
2
[1] X.Q. Zhang. (2000). High-rise and high-density compact urban form: The
3
development of Hong Kong, Compact Cities: Sustainable urban forms for
4
developing countries. 245254.
5 6
RI PT
1
[2] J.A. Voogt, T.R. Oke. (2003). Thermal remote sensing of urban climates, Remote Sens. Environ. 86(3) 370-384.
[3] K.S. Lam, D.W.T. Chan, E.H.W. Chan, C.T. Tai, W.Y. Fung, K.C. Law. (2006). Infiltration
8
of outdoor air in two newly constructed high rise residential buildings, Int. J.
9
Vent. 5(2) 249-258.
M AN U
SC
7
[4] R. Giridharan, S.S.Y. Lau, S. Ganesan, B. Givoni. (2007). Urban design factors
11
influencing heat island intensity in high-rise high-density environments of
12
Hong Kong, Build. Environ. 42(10) 3669-3684.
TE D
10
[5] K.L. Lau, E. Ng. (2013). An investigation of urbanization effect on urban and rural
14
Hong Kong using a 40-year extended temperature record, Landscape Urban
15
Plan. 11442-52.
17 18
[6] A.P. Kamoutsis, A.S. Matsoukis, K.I. Chronopoulos. (2013). Bioclimatic conditions
AC C
16
EP
13
under different ground cover types in the greater Athens area, Greece, Global Nest Journal, 15(2) (2013) 254-260.
19
[7] L. Liu, Y. Lin, J. Liu, L. Wang, D. Wang, T. Shui, X. Chen, Q. Wu. (2017). Analysis of
20
local-scale urban heat island characteristics using an integrated method of
21
mobile measurement and GIS-based spatial interpolation, Building and
22
Environment 117 191-207. 21
ACCEPTED MANUSCRIPT 1 2 3
[8] M. Hodul, A. Knudby, H.C. Ho. (2016). Estimation of Continuous Urban Sky View Factor from Landsat Data Using Shadow Detection, Remote Sens. 8(7) 568. [9] J.E. Nichol, P.H. To. (2012). Temporal characteristics of thermal satellite images for
5
urban heat stress and heat island mapping, ISPRS J. Photogramm. 74,
6
153-162.
RI PT
4
[10] J. Yang, M.S. Wong, M. Menenti, J. Nichol. (2015). Modeling the effective
8
emissivity of the urban canopy using sky view factor, ISPRS J.
9
Photogramm.105 211-219.
M AN U
SC
7
[11] J.P. Connors, C.S. Galletti, W.T. Chow. (2013). Landscape configuration and urban
11
heat island effects: assessing the relationship between landscape
12
characteristics and land surface temperature in Phoenix, Arizona, Landscape
13
ecol. 28(2) 271-283.
16 17
thermal analysis, Remote Sens. Environ. 104(2) 123-132.
EP
15
[12] R. Gluch, D.A. Quattrochi, J.C. Luvall. (2006). A multi-scale approach to urban
[13] L. Liu, Y. Zhang. (2011). Urban Heat Island Analysis Using the Landsat TM Data
AC C
14
TE D
10
and ASTER Data: A Case Study in Hong Kong, Remote Sens. 3(12) 1535-1552.
18
[14] J.E. Nichol, W.Y. Fung, K.-s. Lam, M.S. Wong. (2009). Urban heat island diagnosis
19
using ASTER satellite images and ‘in situ’ air temperature, Atmos. Res. 94(2)
20
276-284.
21
[15] D.A. Quattrochi, J.C. Luvall. (1999). Thermal infrared remote sensing for analysis
22
of landscape ecological processes: Methods and applications, Landscape Ecol. 22
ACCEPTED MANUSCRIPT 1
14(6) 577-598. [16] Q. Weng, D. Lu, J. Schubring. (2004) Estimation of land surface temperature–
3
vegetation abundance relationship for urban heat island studies, Remote
4
Sens. Environ. 89(4) 467-483.
RI PT
2
[17] M.S. Wong, F. Peng, B. Zou, W.Z. Shi, G.J. Wilson. (2016) Spatially Analyzing the
6
Inequity of the Hong Kong Urban Heat Island by Socio-Demographic
7
Characteristics, Int. J. Environ. Res. Public Health 13, 317.
SC
5
[18] L. Chen, E. Ng. (2011). Quantitative urban climate mapping based on a
9
geographical database: A simulation approach using Hong Kong as a case
10
M AN U
8
study, Int. J. Appl. Earth OBS 13(4) 586-594.
[19] M. Aminipouri, A. Knudby, H.C. Ho. (2016) Using multiple disparate data sources
12
to map heat vulnerability: Vancouver case stud, The Canadian Geographer/Le
13
Géographe canadien 60(3) 356-368.
TE D
11
[20] W. Zhan, Y. Chen, J. Zhou, J. Wang, W. Liu, J. Voogt, X. Zhu, J. Quan, J. Li. (2013).
15
Disaggregation of remotely sensed land surface temperature: literature
16
survey, taxonomy, issues, and caveats, Remote Sens. Environ. 131, 119-139.
AC C
EP
14
17
[21] R. Giridharan, S. Ganesan, S.S.Y. Lau. (2004). Daytime urban heat island effect in
18
high-rise and high-density residential developments in Hong Kong, Energy
19 20 21 22
Buildings 36(6), 525-534.
[22] Population Census. (2011). 2011 Population Census Office Census and Statistics Department [23] M.S. Wong, J. Nichol, E. Ng. (2011). A study of the “wall effect” caused by 23
ACCEPTED MANUSCRIPT 1
proliferation of high-rise buildings using GIS techniques, Landscape Urban
2
Plan. 102(4) 245-253. [24] H.C. Ho, K.K.-L. Lau, C. Ren, E. Ng. (2017). Characterizing prolonged heat effects
4
on mortality in a sub-tropical high-density city, Hong Kong, Int. J. Biometer.
5
DOI: 10.1007/s00484-00017-01383-00484.
RI PT
3
[25] W.B. Goggins, E.Y. Chan, E. Ng, C. Ren, L. Chen. (2012). Effect modification of the
7
association between short-term meteorological factors and mortality by
8
urban heat islands in Hong Kong, PLoS One 7(6) e38551.
10
M AN U
9
SC
6
[26] X. Yang, Y. Li, Z. Luo, P.W. Chan. (2017). The urban cool island phenomenon in a high-rise high-density city and its mechanisms, Int. J. Climatol. 37(2) 890-904. [27] A.C.S. Lai, A.C. So, S.K.C. Ng, D. Jonas. (2012). The territory-wide airborne light
12
detection and ranging survey for the Hong Kong Special Administrative Region,
13
In The 33RD Asian Conference on Remote Sensing (pp 26-30)
TE D
11
[28] Z. Qin, A. Karnieli, P. Berliner. (2001). A mono-window algorithm for retrieving
15
land surface temperature from Landsat TM data and its application to the
16
Israel-Egypt border region, Int. J. Remote Sens. 22(18) 3719-3746.
18
AC C
17
EP
14
[29] L.W. Abreu, G.P. Anderson. (1996). The MODTRAN 2/3 report and LOWTRAN 7 model, Contract 19628(91-C)(0132)
19
[30] Z. Qin, A. Karnieli, P. Berliner. (2001). A mono-window algorithm for retrieving
20
land surface temperature from Landsat TM data and its application to the
21
Israel-Egypt border region, Int. J. Remote Sens. 223719-3746.
22
[31] L. Geosystems. (2002). Stereo analyst user’s guide, Leica Geosystems GIS & 24
ACCEPTED MANUSCRIPT
4
urban environments, J. Climatol. 7(2) 193–197. [33] ZRC SAZU (2010) Institute of Anthropological and Spatial Studies ZRC SAZU.
5
Sky-View
6
Accessed 12.20 13
7
Factor.
.
RI PT
3
[32] I.D. Watson, G.T. Johnson. (1987). Graphical estimation of sky view-factors in
http://iaps.zrc-sazu.si/index.php?q=en//node/138.
[34] ITT Visual Information Solutions (2010) ENVI Software: Image processing &
8
analysis
solutions.
9
Accessed 11.09 13
SC
2
Mapping Division, Atlanta, GA, USA
http://www.ittvis.com/ProductServices/ENVI.aspx.
M AN U
1
[35] F. Peng, M.S. Wong, Y. Wan, J.E. Nichol. (2017). Modeling of urban wind
11
ventilation using high resolution airborne LiDAR data, Comput. Environ.
12
Urban Syst. 64, 81-90.
TE D
10
[36] R.L. Hwang, T.P. Lin, A. Matzarakis. (2011). Seasonal effects of urban street
14
shading on long-term outdoor thermal comfort, Build. Environ. 46(4)
15
863-870.
17
[37] T.P. Lin, A. Matzarakis, R.L. Hwang. (2010). Shading effect on long-term outdoor
AC C
16
EP
13
thermal comfort, Build. Environ. 45(1) 213-221.
18
[38] E. Ng. (2009). Policies and technical guidelines for urban planning of high-density
19
cities–air ventilation assessment (AVA) of Hong Kong, Build. Environ. 44(7)
20
1478-1488.
21
[39] Hang, J., Li, Y., Sandberg, M., Buccolieri, R., & Di Sabatino, S. (2012). The
22
influence of building height variability on pollutant dispersion and pedestrian 25
ACCEPTED MANUSCRIPT 1
ventilation in idealized high-rise urban areas. Building and Environment, 56,
2
346-360. [40] Blocken, B., Carmeliet, J., & Stathopoulos, T. (2007). CFD evaluation of wind
4
speed conditions in passages between parallel buildings—effect of
5
wall-function roughness modifications for the atmospheric boundary layer
6
flow. Journal of Wind Engineering and Industrial Aerodynamics, 95(9),
7
941-962.
SC
RI PT
3
[41] T.E. Morakinyo, L. Kong, K.K.L. Lau, C. Yuan, E. Ng. (2017). A study on the impact
9
of shadow-cast and tree species on in-canyon and neighborhood's thermal
10
M AN U
8
comfort, Build. Environ. 115(2017) 1-17.
[42] H. C. Ho, A. Knudby, B.B. Walker, Henderson, S. B. (2017). Delineation of spatial
12
variability in the temperature–mortality relationship on extremely hot days in
13
greater Vancouver, Canada. Environmental health perspectives, 125(1), 66.
TE D
11
[43] H. C. Ho, A. Knudby, Y. Xu, M. Hodul, & M. Aminipouri. (2016). A comparison of
15
urban heat islands mapped using skin temperature, air temperature, and
16
apparent temperature (Humidex), for the greater Vancouver area. Science of
AC C
17
EP
14
the Total Environment, 544, 929-938.
18
[44] P. C. Lai, C.C. Choi, P.P. Wong, T.Q. Thach, M.S. Wong, W. Cheng, A. Krämerd,C.M.
19
Wong (2016). Spatial analytical methods for deriving a historical map of
20
physiological equivalent temperature of Hong Kong. Building and
21
Environment, 99, 22-28.
22
[45] T. Q. Thach, Q. Zheng, P. C. Lai, P. P. Y. Wong, P. Y. K.Chau, H.J. Jahn, D. Plass, L. 26
ACCEPTED MANUSCRIPT 1
Katzschner, A. Kraemer, C.M. Wong (2015). Assessing spatial associations
2
between thermal stress and mortality in Hong Kong: A small-area ecological
3
study. Science of the Total Environment, 502, 666-672.
AC C
EP
TE D
M AN U
SC
RI PT
4 5 6
27
ACCEPTED MANUSCRIPT 1
Table 1. Landsat 5 TM images used in this study. These Landsat 5 TM images were
2
acquired on cloud-free or semi-cloud-free dates across the study site.
3 4
7% 40% 0% 0% 2% 24% 1% 70% 35% 11% 3% 12% 38% 28% 34%
AC C
EP
TE D
5
Cloud Cover
RI PT
Sun Elevation (degree) 38.75 38.14 36.10 41.00 55.09 36.48 36.15 40.22 46.13 38.55 49.30 41.18 56.34 48.75 37.20
SC
Acquired Time (Greenwich Mean Time, GMT) 02:22 02:16 02:16 01:58 02:02 02:13 02:21 02:26 02:25 02:30 02:40 02:40 02:43 02:36 02:42
M AN U
Acquired Date 24 Nov 1988 3 Dec 1988 19 Dec 1988 3 Mar 1996 19 Aug 1996 30 Nov 1996 3 Jan 2000 12 Feb 2000 2 Nov 2000 16 Jan 2005 22 Oct 2005 23 Nov 2005 26 Mar 2010 29 Oct 2010 23 Dec 2010
28
ACCEPTED MANUSCRIPT 1
Table 2. Aerial photographs used in this study. Number of Aerial Year
1964
6
1975
8
1980
8
1985
8
1990
8
1995
8
2000
8
2005 2010
8 9
AC C
EP
TE D
2 3
SC
6
M AN U
1945
RI PT
Photographs
29
ACCEPTED MANUSCRIPT 1 2 3
Table 3. Validation of daily LST estimation. Results indicated a comparison estimated LST and observed LST, at the date with a cloud-free scene across the study site.
RI PT
4
Pearson correlation coefficient (R)
RMSE (°C)
MAE (°C)
3 Dec 1988
0.89
0.62
0.49
30 Nov 1996
0.64
2.71
2.47
3 Jan 2000
0.68
1.40
1.02
22 Oct 2005
0.89
1.56
1.32
29 Oct 2010
0.81
2.98
2.60
AC C
EP
TE D
M AN U
SC
Date
30
ACCEPTED MANUSCRIPT Table 4. Validation of 3D building data. Results indicated 1) a comparison between estimated building data and the building height predicted by LiDAR across open spaces, and 2) a comparison between estimated building data and the building height predicted by LiDAR across areas with buildings. Areas with Buildings RMSE (m) MAE (m) 3.65 2.83 3.12 2.18 2.59 1.92 3.81 2.52 3.63 2.21 2.20 1.66 5.92 3.25 1.06 0.83 1.19 0.81
RI PT
1964 1975 1980 1985 1990 1995 2000 2005 2010
Open Spaces RMSE (m) MAE (m) 1.46 1.23 0.93 0.71 0.82 0.74 1.39 1.13 1.01 0.79 1.13 0.85 0.78 0.41 0.86 0.60 1.06 0.82
SC
Year
M AN U
1 2 3 4 5
AC C
EP
TE D
6 7 8 9
31
ACCEPTED MANUSCRIPT Table 5. Spatiotemporal assessment of building morphology and urban microclimate. Results indicate the additional reduction of relative temperature by air ventilation and shading effects, compared to the areas with improvement of ventilation and shading through years.
Cooling
Averaged relative
Effects
temperature (°C)
Comparison (1988 - 1996) Comparison (1996- 2000) Comparison (2000- 2005) Comparison (2005- 2010)
1988
1996
1996
2000
2000
2005
2005
2010
3.56
2.47
2.76
4.66
4.25
4.26
4.62
5.75
3.96
1.78
2.32
4.32
3.52
4.08
Identified areas without a change of wind speed between
Air Ventilation comparison Identified areas with
speed between two years of comparison Additional reduction of relative
1.09
temperature from air ventilation (°C)
Effects
temperature (°C)
Identified areas
4.43
0.27
0.12
0.57
Comparison (1988 - 1996) Comparison (1996- 2000) Comparison (2000- 2005) Comparison (2005- 2010)
TE D
Averaged relative
3.95
M AN U
an increase of wind
SC
two years of
Cooling
RI PT
1 2 3 4 5
1988
1996
1996
2000
2000
2005
2005
2010
3.63
2.67
2.56
4.24
3.9
3.78
4.01
5.36
3.5
1.78
2.05
3.23
3.68
2.92
3.2
3.81
without a change of
shades between two
EP
years of comparison Shading
Identified areas with
AC C
an increase of shades
between two years of comparison
Additional reduction of relative 0.76
0.6
0.64
0.74
temperature from shading (°C)
6 7
32
ACCEPTED MANUSCRIPT 1 2
Captions of Figures
3
Figure 1. Study area: Kowloon Peninsula, Hong Kong.
5
Figure 2. Flow diagram of the model design of this study.
6
RI PT
4
Figure 3. Trend of daily LST from MODIS LST products in 2010. Blue, red, bright blue
8
and yellow represent four selected grids on the MODIS LST product. X-axis is the day
9
of a year and y-axis is the LST of a selected grid.
M AN U
SC
7
10
Figure 4. Quadratic function to estimate daily air temperature. X-axis is the day of a
12
year and y-axis is the air temperature at HKO station in 2010.
13
TE D
11
Figure 5. Flow chart of the rule-based model. This rule-based model is designed for
15
estimating the additional reduction of relative temperature contributed by air
16
ventilation and shades.
AC C
17
EP
14
18
Figure 6. 3D built environment of Kowloon Peninsula between 1964 and 2010. The
19
followings are the years of each map of built environment represents: (a) 1964, (b)
20
1975, (c) 1980, (d) 1985, (e) 1990, (f) 1995, (g) 2000, (h) 2005, and (i) 2010.
21 22
Figure 7. SVF map retrieved from 3D building data, including the year of (a) 1990, (b)
33
ACCEPTED MANUSCRIPT 1
1995, (c) 2000, (d) 2005, and (e) 2010.
2
Figure 8. Relative temperature maps of Kowloon Peninsula, including the year of (a)
4
1988, (b) 1996, (c) 2000, (d) 2005, and (e) 2010.
RI PT
3
5
Figure 9. Maps of major air ventilation paths across the Kowloon Peninsula at the
7
year of (a) 1988, (b) 1996, (c) 2000, (d) 2005, and (e) 2010. The base map of these
8
figures are the relative temperature maps of Kowloon Peninsula in each year, for
9
comparing the effects of wind on intra-urban heat.
M AN U
SC
6
10
Figure 10. Examples of shading effects across the study site. Areas with relatively
12
taller buildings across the study site in 2010 have been selected and marked.
13
Selected areas with shades indicate a lower daytime temperature compared to areas
14
without a shade.
EP
16
AC C
15
TE D
11
34
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
RI PT
SC M AN U TE D
-
EP
-
Reconstructed the historical temperature maps across Kowloon Peninsula based on Landsat images Reconstructed 3D built environment with aerial photos Areas with more shades through years have greater contribution to reduce local heat Areas with more ventilation through years have greater contribution to reduce local temperature Larger variation of building heights is suggested to integrate both shading effects and air ventilation for developing a compact city
AC C
-