Content uploaded by Manjula Ranagalage
Author content
All content in this area was uploaded by Manjula Ranagalage on Jul 10, 2019
Content may be subject to copyright.
Content uploaded by Manjula Ranagalage
Author content
All content in this area was uploaded by Manjula Ranagalage on Jul 10, 2019
Content may be subject to copyright.
remote sensing
Article
Spatial Analysis of Surface Urban Heat Islands in
Four Rapidly Growing African Cities
Matamyo Simwanda 1, 2, *, Manjula Ranagalage 1,3 , Ronald C. Estoque 4and Yuji Murayama 1
1Faculty of Life and Environmental Sciences, University of Tsukuba, 1-1-1 Tennodai,
Tsukuba City 305-8572, Ibaraki, Japan
2Department of Plant and Environmental Sciences, School of Natural Resources, Copperbelt University,
P.O. Box 21692, Kitwe 10101, Zambia
3
Department of Environmental Management, Faculty of Social Sciences and Humanities, Rajarata University
of Sri Lanka, Mihintale 50300, Sri Lanka
4National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba City, Ibaraki 305-8506, Japan
*Correspondence: matamyo.simwanda@cbu.ac.zm; Tel.: +260-978-652214
Received: 23 May 2019; Accepted: 7 July 2019; Published: 10 July 2019
Abstract:
Africa’s unprecedented, uncontrolled and unplanned urbanization has put many African
cities under constant ecological and environmental threat. One of the critical ecological impacts of
urbanization likely to adversely affect Africa’s urban dwellers is the urban heat island (UHI) effect.
However, UHI studies in African cities remain uncommon. Therefore, this study attempts to examine
the relationship between land surface temperature (LST) and the spatial patterns, composition and
configuration of impervious surfaces/green spaces in four African cities, Lagos (Nigeria), Nairobi
(Kenya), Addis Ababa (Ethiopia) and Lusaka (Zambia). Landsat OLI/TIRS data and various geospatial
approaches, including urban–rural gradient, urban heat island intensity, statistics and urban landscape
metrics-based techniques, were used to facilitate the analysis. The results show significantly strong
correlation between mean LST and the density of impervious surface (positive) and green space
(negative) along the urban–rural gradients of the four African cities. The study also found high urban
heat island intensities in the urban zones close (0 to 10 km) to the city center for all cities. Generally,
cities with a higher percentage of the impervious surface were warmer by 3–4
◦
C and vice visa. This
highlights the crucial mitigating effect of green spaces. We also found significant correlations between
the mean LST and urban landscape metrics (patch density, size, shape, complexity and aggregation)
of impervious surfaces (positive) and green spaces (negative). The study revealed that, although most
African cities have relatively larger green space to impervious surface ratio with most green spaces
located beyond the urban footprint, the UHI effect is still evident. We recommend that urban planners
and policy makers should consider mitigating the UHI effect by restoring the urban ecosystems
in the remaining open spaces in the urban area and further incorporate strategic combinations of
impervious surfaces and green spaces in future urban and landscape planning.
Keywords:
urban heat island; land surface temperature; impervious surface; green space; African
cities; Landsat data
1. Introduction
Despite Africa being the least urbanized continent, its urbanization is arguably one of the fastest
in the world [
1
]. Africa’s urban population has been growing at a very high rate, i.e., from an estimated
28% in 1980 [
2
] to 43% in 2018 and projected to be about 60% by 2050 [
3
]. Much of the urbanization
in Africa has been unplanned and unregulated, exacerbated by the legacy of colonialism, structural
adjustment and neo-liberalism that has continuously spawned weak urban planning institutions [
4
].
Remote Sens. 2019,11, 1645; doi:10.3390/rs11141645 www.mdpi.com/journal/remotesensing
Remote Sens. 2019,11, 1645 2 of 20
Most of the African cities have thus emerged as unplanned cities dominated by overcrowded informal
settlements haphazardly located close to urban growth centers such as the central business district and
other industrial and commercial areas [5]. Consequently, ecological and environmental conditions in
African cities are under constant threat.
One of the ecological consequences of urbanization is the urban heat island (UHI) effect,
a phenomenon that refers to the occurrence of higher temperatures in urban areas than the surrounding
rural areas [
6
–
10
]. UHI occurs as a result of land cover transformations, mainly the replacement of
natural vegetation and agricultural lands by impervious surfaces (concrete, asphalt, rooftops and
building walls) associated with urban land use [
11
]. Some of the negative impacts of the UHI include
increased energy consumption, elevated emissions of air pollutants and greenhouse gases, impaired
water quality as well as causing compromised environmental conditions that affect human health and
comfortability [
12
,
13
]. It is for this reason that the UHI phenomenon has become a key research focus
in various disciplines such as urban geography, urban planning, urban ecology and urban climatology.
Generally, there are two types of UHIs: Atmospheric UHI (AUHI) and surface UHI (SUHI) [
13
].
AUHIs are measured using air temperature while SUHIs are measured using surface temperature [
8
,
10
,
13
].
The high temporal resolution of air temperature makes AUHIs effective in describing the temporal
variation of UHIs. However, AUHIs have a drawback of failing to depict the spatial variation of UHIs [
14
].
Conversely, surface temperature patterns can exhibit both the spatial and temporal variation of SUHIs of
entire cities [
14
,
15
]. The use of land surface temperature (LST) retrieved from remotely sensed thermal
infrared data has since become widely recognized as an effective tool for examining spatial patterns of
UHIs in relation to urban landscape patterns [
14
–
20
]. This study focuses on SUHIs based on LST retrieved
from Landsat data.
Many studies have shown that LST can be related to land cover, mainly impervious surfaces [
6
,
8
,
21
,
22
] and green spaces [
14
,
23
–
25
], to comprehend the SUHI effect in urbanized landscapes. Researchers
have consistently demonstrated that increasing green space or vegetation cover in urban areas has a
mitigating effect on UHIs, while the growth of impervious surfaces increases urban heating [
17
,
24
,
26
,
27
].
Recently, techniques such as urban–rural gradient and statistical analysis [
8
,
28
,
29
] as well as UHI
intensity analysis [
23
,
30
,
31
] have been familiar in understanding the effect of landscape patterns
on LST (i.e., the UHI effect). There has also been increasing interest in the spatial composition and
configuration of impervious surface and green spaces owing to the different mix or complexity of
different urban environments. A proliferation of studies has applied urban landscape metrics-based
techniques to show that the spatial composition and configuration of impervious surfaces and green
spaces (e.g., size, patch density and complexity) affect the magnitude of LST [7,8,14,24,26].
It is evident from the vast literature that the UHI phenomenon has been extensively studied
in cities worldwide irrespective of their sizes and locations. Several studies have examined the
relationship between LST and the composition and configuration of impervious surfaces and green
spaces. While some recent studies (e.g., [
32
] in Durban (South Africa), [
33
] in Lagos (Nigeria) and [
34
]
in Addis Ababa (Ethiopia)) have been conducted, UHI studies are still very uncommon in Africa.
Moreover, previous studies have been conducted on individual cities based on the specific conditions
of their urban environments. The uncontrolled and unplanned urbanization that has been experienced
in African cities in recent decades makes them interesting case studies for a comparative study of UHIs.
Therefore, this study conducts a comparative analysis to examine the relationship between LST
and the spatial patterns, composition and configuration of impervious surfaces and green spaces in
four African cities, Lagos (Nigeria), Nairobi (Kenya), Addis Ababa (Ethiopia) and Lusaka (Zambia).
Landsat OLI/TIRS data and various geospatial approaches, including urban–rural gradient, UHI
intensity, statistics and urban landscape metrics-based techniques, were used to facilitate the analysis.
The four cities were selected to get a good representation of African cities based on the following
criteria: (i) Being the largest or capital city; (ii) being the main economic and commercial center of
their country; and (iii) experiencing rapid urbanization with the highest population in their respective
Remote Sens. 2019,11, 1645 3 of 20
countries. In 2016, the population of Lagos was estimated at 13.7 million, Nairobi at 4.2 million, Addis
Ababa at 3.3 million, while Lusaka was at 2.3 million [35].
2. Data and Methods
2.1. Study Areas
The study areas include the city cores of Lagos (Nigeria) located in West Africa, Nairobi (Kenya)
and Addis Ababa (Ethiopia) located in East Africa and Lusaka (Zambia) located in central-southern
Africa (Figure 1). For comparison, we used a 40 km
×
40 km subset with a 20 km radius from the
city center of each city as a common unit of analysis (Figure 1). All the study areas are located in the
tropical climate zones of sub-Saharan Africa.
Remote Sens. 2019, 11, x FOR PEER REVIEW 3 of 21
their country; and (iii) experiencing rapid urbanization with the highest population in their respective
countries. In 2016, the population of Lagos was estimated at 13.7 million, Nairobi at 4.2 million, Addis
Ababa at 3.3 million, while Lusaka was at 2.3 million [35].
2. Data and Methods
2.1. Study Areas
The study areas include the city cores of Lagos (Nigeria) located in West Africa, Nairobi (Kenya)
and Addis Ababa (Ethiopia) located in East Africa and Lusaka (Zambia) located in central-southern
Africa (Figure 1). For comparison, we used a 40 km × 40 km subset with a 20 km radius from the city
center of each city as a common unit of analysis (Figure 1). All the study areas are located in the
tropical climate zones of sub-Saharan Africa.
Figure 1. Location of study areas in Africa. Study areas are displayed using a false color composite of
Landsat 8 images (band 5—red, band 4—green and band 3—blue).
According to [36], the climate in Lagos is tropical with two distinct seasons, i.e., a pronounced
dry season in the low-sun months and a wet season is in the high-sun months. The annual mean
temperature in Lagos is approximately 26.5 °C. The climate in Addis Ababa, Nairobi and Lusaka is
generally sub-tropical with moderate seasonality, although there are variations across the cities.
Addis Ababa and Lusaka have a climate characterized by dry winters and mild rainy and hot humid
summers with annual mean temperatures of 15.9 °C and 19.9 °C, respectively. Nairobi has a marine
west-coast climate that is mild with no dry season, warm summers and an annual mean temperature
of 17.7 °C.
The land cover features in the four cities are typical of those in rapidly urbanizing African cities
with built-up lands (impervious surfaces) characterized by various land uses including commercial,
industrial, public institutions and residential areas dominated by informal settlements located close
to urban growth centers, especially the central business district [5,37]. Other land cover features
include forests, woodlands, grasslands, croplands and water surfaces such as the sea, lakes, rivers
and dams [38].
Figure 1.
Location of study areas in Africa. Study areas are displayed using a false color composite of
Landsat 8 images (band 5—red, band 4—green and band 3—blue).
According to [
36
], the climate in Lagos is tropical with two distinct seasons, i.e., a pronounced
dry season in the low-sun months and a wet season is in the high-sun months. The annual mean
temperature in Lagos is approximately 26.5
◦
C. The climate in Addis Ababa, Nairobi and Lusaka is
generally sub-tropical with moderate seasonality, although there are variations across the cities. Addis
Ababa and Lusaka have a climate characterized by dry winters and mild rainy and hot humid summers
with annual mean temperatures of 15.9
◦
C and 19.9
◦
C, respectively. Nairobi has a marine west-coast
climate that is mild with no dry season, warm summers and an annual mean temperature of 17.7 ◦C.
The land cover features in the four cities are typical of those in rapidly urbanizing African cities
with built-up lands (impervious surfaces) characterized by various land uses including commercial,
industrial, public institutions and residential areas dominated by informal settlements located close
to urban growth centers, especially the central business district [
5
,
37
]. Other land cover features
include forests, woodlands, grasslands, croplands and water surfaces such as the sea, lakes, rivers
and dams [38].
Remote Sens. 2019,11, 1645 4 of 20
2.2. Satellite Data and Pre-Processing
The satellite data used in this study were six cloud-free (<10%) Landsat-8 OLI/TIRS images
obtained from the US Geological Survey website for each of the study areas. All the Landsat-8 OLI/TIRS
data obtained were acquired during the dry season (Table 1). LST in tropical cities is better derived for
the dry season due to cloud-coverage problems [
39
]. Other studies have also shown that the SUHI
phenomenon is more prominent during the dry season [
8
,
33
,
34
]. The dry season was also chosen
to eliminate non-permanent green spaces that only exist during the wet season. The pre-processing
for each image included radiometric calibration and atmospheric correction (dark-object subtraction)
carried out using the TerrSet Geospatial Monitoring and Modeling Software. The purpose of this
pre-processing was to convert the digital number (DN) values of the multispectral bands (bands 1–7
and 9) into surface reflectance values and convert the DN values of the thermal bands (bands 10 and
11) into at-satellite brightness temperature (T
B
) expressed in degrees Kelvin [
8
]. The pre-processed
data were then used to extract the green spaces, impervious surfaces and LST.
Table 1. Details of the Landsat-8 imagery used.
City Sensor Scene ID Acquisition Date Time (GMT) Season
Addis Ababa Landsat-8
OLI/TIRS LC81680542017026LGN00 26-January 2017 7:40:29 Dry
Lagos Landsat-8
OLI/TIRS LC81910552015342LGN00 8-December 2015 10:03:03 Dry
Nairobi Landsat-8
OLI/TIRS LC81680612016216LGN00 3-August 2016 7:43:12 Dry
Lusaka Landsat-8
OLI/TIRS LC81720712016212LGN00 20-July 2016 8:11:54 Dry
2.3. Retrieval of LST
The methods for retrieving LST from Landsat data have been extensively documented in the
literature. The process involves, first, converting DNs of the thermal bands (i.e., bands 10 and 11 in
Landsat-8) to absolute units of at-sensor spectral radiance [
6
,
15
,
40
]. Second, under the assumption that
the Earth’s surface is a black body (i.e., spectral emissivity =1), the thermal band data is converted from
at-sensor spectral radiance to effective at-sensor brightness temperature using Equation (1) [40,41].
TB=K2
In K2
Lλ+1(1)
where T
B
is the effective at-sensor brightness temperature in degrees Kelvin, L
λ
is the spectral radiance
at the sensor’s aperture in W/(m2 sr
µ
m) and K
1
and K
2
are pre-launch calibration constants (i.e.,
thermal conversion constants for the bands 10 or 11 provided in the Landsat-8 metadata [
41
] in this
study). Finally, the at-sensor brightness temperatures are corrected for varied spectral emissivity
depending on the nature of the land cover and LST is retrieved [6,8,17].
In this study, prior to LST retrieval, we used the method from [
42
] that takes into account standard
deviation (m), a combined mean value of the soil and vegetation emissivities cd (n) and the vegetation
proportion (Pv), calculated by Equations (2)–(4), respectively, to obtain the land surface emissivity(ε)
(Equation (5)) for each study area.
m=(εv−εs)−(1−εs)Fεv4 (2)
n=εs+(1−εs)Fεv(3)
PV= NDVI −NDVImin
NDVImax −NDVImin !2
(4)
ε=mPV+n(5)
Remote Sens. 2019,11, 1645 5 of 20
where,
εs
is the soil emissivity,
εv
is the vegetation emissivity and Fis a shape factor whose mean
value, assuming different geometrical distributions, is 0.55 (Sobrino et al., 1990 in [
42
]). NDVI is the
normalized difference vegetation index derived using the surface reflectance of bands 4 (
ρred
) and 5
(ρNIR) of Landsat-8 (Equation (6)) [8]:
NDVI =(ρred −ρNIR)
(ρred +ρNIR)(6)
We applied the values 0.004 for m and 0.986 for n based on the findings of Sobrino et al. (2004) to
calculate
ε
. Finally, we converted the brightness temperatures (T
B
) obtained through pre-processing
band 10 of Landsat-8 (see Section 2.2) to degrees Celsius (
◦
C) [
8
] and calculated the emissivity-corrected
LST using Equation (7) [6,17,43]:
LST(◦C)=TB
1+(λ×TB/ρ)Inε(7)
where
λ
=wave- length of emitted radiance (
λ
=10.8
µ
m, for Landsat-8 band 10 [
8
]);
ρ
=h
×
c/
σ
(1.438
×
10
−
2 m K),
σ
=Boltzmann constant (1.38
×
10
−
23 J/K), h =Planck’s constant (6.626
×
10
−
34 Js) and
c=velocity of light (2.998 ×108 m/s); and εis the land surface emissivity.
2.4. Extraction of Land Cover
Many studies have demonstrated that LST can be related to land cover, mainly impervious
surfaces [
6
,
8
,
21
] and green spaces [
14
,
17
,
24
], to comprehend the SUHI effect in urbanized landscapes.
In this study, we used the pre-processed Landsat-8 images to extract impervious surfaces and green
spaces using spectral indices. Several studies have shown the aptness of the spectral-based approach
in land cover extraction [
8
,
9
,
28
]. Our land cover extraction process was as follows. First, we used
the modified normalized difference water index (MNDWI) to extract water bodies and exclude them
from the images. The MNDWI has been proven to accurately discriminate water from non-water
features [44]. Equation (6) was used to compute the MNDWI for each study area [44]:
MNDWI =(ρGreen −ρSWIR1)
(ρGreen +ρSWIR1)(8)
where
ρGreen and ρSWIR1
are the surface reflectance values of bands 3 and 6 of the Landsat-8
images, respectively.
Afterwards, we used the visible red and NIR-based built-up index (VrNIR-BI) to extract impervious
surfaces. One of the most noted spectral confusions in the land cover classification of African landscapes
is between the impervious surface (IS) and bare lands usually characterized by dry grasslands and
abandoned croplands. The VrNIR-BI can accurately separate impervious surfaces from bare lands [
8
].
The VrNIR-BI was recommended by [
45
] after comparing the index to six other spectral built-up
indices, including the commonly applied normalized difference built-up index (NDBI) [
46
] based on
Landsat ETM+and Landsat OLI/TIRS images. Equation (9) was used to compute the VrNIR-BI for
each study area:
VrNIR-BI =(ρRed −ρNIR)
(ρRed +ρNIR)(9)
where
ρRed and ρNIR
are the surface reflectance values of bands 4 and 5 of the Landsat-8 images,
respectively. To extract the green spaces for each study area, we used the NDVI expressed in Equation
(6) above. NDVI is one of the extensively applied indices when relating LST to green spaces in
SUHI studies [
17
]. Manual thresholding was applied to extract VrNIR-BI and NDVI after several
tests through visual assessments of the index maps with close reference to the Landsat-8 images and
high-resolution Google earth imagery in each study area. The thresholds applied to extract VrNIR-BI
for Lagos, Nairobi, Addis Ababa and Lusaka were 0.45, 0.565, 0.352 and 0.485, respectively. To extract
Remote Sens. 2019,11, 1645 6 of 20
NDVI, the thresholds applied were
−
0.425,
−
0.245,
−
0.169 and
−
0.315 for Lagos, Nairobi, Addis Ababa
and Lusaka, respectively (see Figure 2for a zoomed-in sample of VrNIR-BI and Landsat 8 imagery in
each study area).
Remote Sens. 2019, 11, x FOR PEER REVIEW 6 of 21
through visual assessments of the index maps with close reference to the Landsat-8 images and high-
resolution Google earth imagery in each study area. The thresholds applied to extract VrNIR-BI for
Lagos, Nairobi, Addis Ababa and Lusaka were 0.45, 0.565, 0.352 and 0.485, respectively. To extract
NDVI, the thresholds applied were −0.425, −0.245, −0.169 and −0.315 for Lagos, Nairobi, Addis Ababa
and Lusaka, respectively (see Figure 2 for a zoomed-in sample of VrNIR-BI and Landsat 8 imagery
in each study area).
Figure 2. Zoomed-in sample of visible red and NIR-based built-up index (VrNIR-BI) and Landsat 8
imagery in each study area.
Finally, we produced a land cover map for each study area containing four categories,
impervious surfaces, green spaces, other and water. Impervious surfaces included buildings,
transport utilities and all other impervious areas. Green spaces comprised forests, grass and all
healthy green vegetation cover, while other comprised all land cover features excluding impervious
surface, green space and water. Water included the sea, lakes, rivers, streams, dams, swamps,
reservoirs and ponds. The other and water categories were excluded in all further analyses.
2.4. Analysis of Spatial Patterns
2.4.1. Urban–Rural Gradient Analysis
The aptness of the gradient analysis approach in revealing the distribution and spatial variations
of LST along the urban–rural areas has been shown in recent studies [8,28,29]. There are two main
urban–rural gradient analysis methods that have been developed and applied in the literature. The
first one is the use of directional transects running across the city center with their ends both
extending to the rural areas [47,48]. The second one applies concentric rings or zones around the city
center with standard distance intervals extending to the rural areas [8,48]. The concentric ring
gradient analysis method is effective in cities exhibiting single-core urban growth patterns around
the city center such as the four African cities in this study [8,28,49].
Therefore, we used the concentric ring gradient analysis method to study the spatial patterns
and influences of impervious surfaces, and green spaces on LST along the urban–rural landscape of
each city. Considering that the urban development patterns of all the African cites in this study are
Figure 2.
Zoomed-in sample of visible red and NIR-based built-up index (VrNIR-BI) and Landsat 8
imagery in each study area.
Finally, we produced a land cover map for each study area containing four categories, impervious
surfaces, green spaces, other and water. Impervious surfaces included buildings, transport utilities
and all other impervious areas. Green spaces comprised forests, grass and all healthy green vegetation
cover, while other comprised all land cover features excluding impervious surface, green space and
water. Water included the sea, lakes, rivers, streams, dams, swamps, reservoirs and ponds. The other
and water categories were excluded in all further analyses.
2.5. Analysis of Spatial Patterns
2.5.1. Urban–Rural Gradient Analysis
The aptness of the gradient analysis approach in revealing the distribution and spatial variations
of LST along the urban–rural areas has been shown in recent studies [
8
,
28
,
29
]. There are two main
urban–rural gradient analysis methods that have been developed and applied in the literature. The first
one is the use of directional transects running across the city center with their ends both extending to
the rural areas [
47
,
48
]. The second one applies concentric rings or zones around the city center with
standard distance intervals extending to the rural areas [
8
,
48
]. The concentric ring gradient analysis
method is effective in cities exhibiting single-core urban growth patterns around the city center such as
the four African cities in this study [8,28,49].
Therefore, we used the concentric ring gradient analysis method to study the spatial patterns and
influences of impervious surfaces, and green spaces on LST along the urban–rural landscape of each
city. Considering that the urban development patterns of all the African cites in this study are based on
the single-core concept, we selected the city center by identifying the oldest building around the city
Remote Sens. 2019,11, 1645 7 of 20
center area in each city. We then created multiple concentric rings around the city center of each study
area with distance intervals of 200 m. Subsequently, the densities of impervious surfaces and green
spaces were determined in each zone and plotted across the urban–rural gradient for each study area.
2.5.2. SUHI Intensity Analysis
The SUHI intensity is a well-known measure of the SUHI effect across the urban–rural landscape.
It is generally defined as the difference in temperature between an urban and a rural area [
50
]. The SUHI
intensity is calculated using either air temperature from meteorological data (e.g., [
51
,
52
]) or mean
surface temperatures using satellite images [
23
]. Analyzing SUHI intensity patterns and their urban
and rural area variations has remained an imperative part of SUHI studies [30,53].
To analyze the SUHI intensity patterns, we divided the study areas into two major areas, urban
and rural. To delineate the urban and rural areas, we estimated the urban area (also referred to as the
‘built-up footprint’), based on the physical extent of the impervious surface in each city. By way of
justification, a wide range of social, economic, demographic, administrative or political indicators have
been used to define urban areas, but there is no consensus on how to construct a consistent definition
based on any single set of attributes [
54
]. For example, an administrative boundary of a city cannot
be relied on as a means of defining an urban area as boundaries frequently change over time, are not
comparable across cities and are usually over- or under-estimated [
55
]. The terms ‘urban area’ or
‘urban footprint’ are widely used to basically refer to the spatial extent of urbanized areas on a regional
scale; a definition which is both fuzzy and inconsistent [56,57].
As such, defining the urban area based on the physical extent of the built-up land (impervious
surface), as adopted in most remote sensing urban studies (e.g., [
58
,
59
]), is the best potion. We used
the concentric zones defined in Section 2.5.1 to determine the urban area, i.e., all concentric zones that
contained impervious surfaces in each city. Accordingly, all concentric zones beyond the maximum
radius of the urban or built-up footprint were considered as rural. We calculated the SUHI intensity by
calculating the difference between the mean LST at the city center of each study area (i.e., the kilometer
0) and the mean LST in each of the 200 m concentric zones created as outlined in Section 2.5.1 across
the urban–rural landscape. Equation (7) was used to calculate the SUHI intensity for each study area:
SUHI intensity =µLST0−µLSTi(10)
where
µ
LST
0
is the kilometer 0 mean LST at the city center of each study area and
µ
LST
i
is the mean
LST in each buffer zone (SZ), where i=1,2,3
. . .
.nand nis the total number of buffer zones in each
study area.
2.5.3. Urban Landscape Metrics Analysis
One of our other interests in this study was to comprehend how the composition, shape, complexity
and spatial arrangement of impervious surfaces and green spaces could have influenced the spatial
distribution of LST across the landscape of each study area. The use of urban landscape metrics
has been widely proven to enhance the understanding of LST spatial variability in relation to the
configuration of landscape features (e.g., impervious surface and green spaces) [
7
,
14
,
24
,
26
]. In this
study, we selected five class level spatial metrics, patch density (PD), mean patch area (AREA_MN),
mean shape index (SHAPE_MN), mean fractal dimension index (FRAC_MN) and aggregation index
(AI). The descriptions and equations for calculating each selected spatial metric are presented in Table 2.
To relate the spatial metrics to LST distribution, each study area was first divided into 100 polygon
grids (4 km x 4 km). Then, impervious surfaces and green spaces in each polygon grid were extracted
and used in the computation of spatial metrics in each study area. We computed the class level spatial
metrics using Fragstats software (version 4.2v) [
60
]. We defined the patch neighbor using the 8-cell rule.
Remote Sens. 2019,11, 1645 8 of 20
Table 2. Selected class level spatial metrics.
Metric (Abbreviation) Description Measure Equation
Mean Patch Area
(AREA_MN)
Average patch area—total impervious
surface or green space area divided by
number of their respective patches -
(unit: km2)
Composition 1
10,000×n×
n
P
i=1ai
Patch Density (PD)
The number of patches per unit area of
impervious surface or green space (unit:
number per km2).
Composition and
spatial arrangement n
A×106
Mean Shape Index
(SHAPE_MN)
Mean value of shape index—it is the
simplest and most straightforward
measure of shape complexity. MSI is
greater than one; MSI =1 would result
if all impervious surface or green space
patches were circular or square grids
(unit: none).
Shape and complexity 1
n×0.25 Pi
√ai
Mean Fractal Dimension
Index (FRAC_MN)
FRAC_MN also measures shape
complexity. FRAC_MN approaches one
for shapes with simple perimeters and
approaches two when shapes are more
complex (unit: none).
Shape and Complexity 1
n×2 Ln 0.25 Pi
Ln √ai
Aggregation Index (AI)
The tendency of impervious surface or
green space patches to be spatially
aggregated (unit: none).
Spatial arrangement AI =gi
max−gi(100)
Note: a
i=
area of patch i; n=number of patches; A=total class area; p
i
=perimeter of patch i;g
i
=number of like adjacencies
(joins) between pixels of patch type (class) i based on the single-count method; max
−
g
i
=maximum number of like
adjacencies (joins) between pixels of patch type (class) i based on the single-count method (details in [
60
]). A patch
is defined as a relatively homogeneous area (i.e., impervious surface or green space in this study) that differs from its
surroundings [61].
2.5.4. Statistical Analysis
Statistical analysis was conducted using the Pearson correlation analysis and scatter plots to
examine the relationship of mean LST and the density of impervious surfaces and green spaces in each
of the 200 m buffer zones created as outlined in Section 2.5.1. We further conducted Pearson correlation
analysis to investigate the relationship between mean LST and spatial metrics based on the 100 grid
polygons created as outlined in Section 2.5.3 for each study area.
3. Results
3.1. LST Relationship with Impervious Surfaces and Green Spaces
The LST and land cover maps for the study areas, Lagos, Nairobi, Addis Ababa and Lusaka, are
shown in Figures 3and 4. Figure 5shows the minimum, maximum and mean LST of impervious
surfaces and green spaces, and the percentage of impervious surfaces and green spaces relative to the
total landscape (40 km x 40 km) considered for each study area. The results revealed that Lagos had
the highest percentage of impervious surface (40%). Compared to Lagos, the other three cities had
very low percentages of impervious surface, i.e., Addis Ababa 12%, Lusaka 11% and the lowest being
Nairobi with 8%. However, Nairobi had the highest percentage of green spaces (32%) followed by
Lagos (25%), Addis Ababa (23%) and the lowest in Lusaka (20%) (Figure 5c).
In terms of the relationship between mean LST and the impervious surfaces and green spaces,
the results revealed that cities with a higher percentage of impervious surface were warmer and vice
versa. The results showed that Lagos was the warmest city and Nairobi was the coolest city, while
Addis Ababa and Lusaka were slightly warmer than Nairobi but cooler than Lagos. Lagos recorded
the highest maximum and minimum LST values of impervious surfaces (i.e., 42.0
◦
C and 25.1
◦
C,
respectively) while Nairobi recorded the lowest maximum and minimum LST values of impervious
surfaces (i.e., 33.5
◦
C and 15.4
◦
C, respectively). The mean LST of impervious surfaces in Lagos was
32.4
◦
C and 27.8
◦
C in Nairobi. Addis Ababa and Lusaka had a mean LST of impervious surfaces
Remote Sens. 2019,11, 1645 9 of 20
of 29.5
◦
C (Figure 5a). With regard to the LST of green spaces, Lagos still had the highest, with a
maximum of 41.2
◦
C and a minimum of 24.6
◦
C, and Nairobi still had the lowest, with maximum and
minimum LST values of 31.5
◦
C and 16.5
◦
C, respectively. For the mean LST of green spaces, Lagos
had 28.4
◦
C, Lusaka had 27.7
◦
C and Addis Ababa had 25.4
◦
C, with the lowest being in Nairobi, 23
◦
C
(Figure 5b).
3.2. LST Relationship with Impervious Surfaces and Green Spaces along the Urban–Rural Gradient
According to the results, the relationships between mean LST and impervious surface and green
space density along the urban–rural gradient of Nairobi, Addis Ababa and Lusaka had similar spatial
patterns (Figure 6). The impervious surface and green space density decreased and increased gradually
along the urban–rural gradient, respectively. However, the mean LST had a similar pattern with
impervious surfaces within the urban footprint, decreasing from the city center to the maximum spatial
extent of the urban area (i.e., around the cross-point of impervious and green space density in Figure 6).
Beyond the spatial extent of the urban area, the mean LST increased gradually, similar to the pattern of
green space density. Unlike the other three cities, the mean LST and impervious surface density in
Lagos decreased while the green space density increased through the urban–rural gradient (Figure 6).
This could be because of urban area spatial extent in Lagos, which dominates the landscape with
almost no rural areas as defined in this study.
Remote Sens. 2019, 11, x FOR PEER REVIEW 9 of 21
versa. The results showed that Lagos was the warmest city and Nairobi was the coolest city, while
Addis Ababa and Lusaka were slightly warmer than Nairobi but cooler than Lagos. Lagos recorded
the highest maximum and minimum LST values of impervious surfaces (i.e., 42.0 °C and 25.1 °C,
respectively) while Nairobi recorded the lowest maximum and minimum LST values of impervious
surfaces (i.e., 33.5 °C and 15.4 °C, respectively). The mean LST of impervious surfaces in Lagos was
32.4 °C and 27.8 °C in Nairobi. Addis Ababa and Lusaka had a mean LST of impervious surfaces of
29.5 °C (Figure 5a). With regard to the LST of green spaces, Lagos still had the highest, with a
maximum of 41.2 °C and a minimum of 24.6 °C, and Nairobi still had the lowest, with maximum and
minimum LST values of 31.5 °C and 16.5 °C, respectively. For the mean LST of green spaces, Lagos
had 28.4 °C, Lusaka had 27.7 °C and Addis Ababa had 25.4 °C, with the lowest being in Nairobi, 23
°C (Figure 5b).
Figure 3. Land surface temperature (LST) distribution in the study areas: Lagos, Nairobi, Addis Ababa
and Lusaka.
Figure 3.
Land surface temperature (LST) distribution in the study areas: Lagos, Nairobi, Addis Ababa
and Lusaka.
Remote Sens. 2019,11, 1645 10 of 20
Remote Sens. 2019, 11, x FOR PEER REVIEW 10 of 21
Figure 4. Land cover maps for the study areas: Lagos, Nairobi, Addis Ababa and Lusaka.
Figure 5. LST distribution and the percentage of impervious surfaces (IS) and green spaces (GS) in
each study area.
Figure 4. Land cover maps for the study areas: Lagos, Nairobi, Addis Ababa and Lusaka.
Remote Sens. 2019, 11, x FOR PEER REVIEW 10 of 21
Figure 4. Land cover maps for the study areas: Lagos, Nairobi, Addis Ababa and Lusaka.
Figure 5. LST distribution and the percentage of impervious surfaces (IS) and green spaces (GS) in
each study area.
Figure 5.
LST distribution and the percentage of impervious surfaces (IS) and green spaces (GS) in
each study area.
Remote Sens. 2019,11, 1645 11 of 20
Remote Sens. 2019, 11, x FOR PEER REVIEW 11 of 21
3.2. LST Relationship with Impervious Surfaces and Green Spaces along the Urban–Rural Gradient
According to the results, the relationships between mean LST and impervious surface and green
space density along the urban–rural gradient of Nairobi, Addis Ababa and Lusaka had similar spatial
patterns (Figure 6). The impervious surface and green space density decreased and increased
gradually along the urban–rural gradient, respectively. However, the mean LST had a similar pattern
with impervious surfaces within the urban footprint, decreasing from the city center to the maximum
spatial extent of the urban area (i.e., around the cross-point of impervious and green space density in
Figure 6). Beyond the spatial extent of the urban area, the mean LST increased gradually, similar to
the pattern of green space density. Unlike the other three cities, the mean LST and impervious surface
density in Lagos decreased while the green space density increased through the urban–rural gradient
(Figure 6). This could be because of urban area spatial extent in Lagos, which dominates the landscape
with almost no rural areas as defined in this study.
The Pearson’s correlation results showed significant relationships (p < 0.001) between the mean
LST and the density of impervious surfaces (positive) and green spaces (negative) in all the study
areas along the urban–rural gradient (Figure 7). The correlation of impervious surfaces with mean
LST in Lagos (r
2
= 0.9483; slope = 0.0641) and Lusaka (r
2
= 0.5766; slope = 0.0438) was relatively high
compared to Nairobi (r
2
= 0.2783; slope = 0.0258) and Addis Ababa (r
2
= 0.1776; slope = 0.0186). In
contrast, the correlation of green spaces with mean LST was relatively very low in Nairobi (r
2
= 0.3085;
slope = –0.0355) compared to Lagos (r
2
= 0.9482; slope = –0.0676), Addis Ababa (r
2
= 0.7881; slope = –
0.0659) and Lusaka (r
2
= 0.801; slope = −0.0199).
Figure 6. Relationships between the LST and impervious surfaces and green spaces along the urban–
rural gradients of Lagos (a), Nairobi (b), Addis Ababa (c) and Lusaka (d). Note: Urban and rural were
discerned based on the physical extent of the built-up footprint for each city.
Figure 6.
Relationships between the LST and impervious surfaces and green spaces along the
urban–rural gradients of Lagos (
a
), Nairobi (
b
), Addis Ababa (
c
) and Lusaka (
d
). Note: Urban and
rural were discerned based on the physical extent of the built-up footprint for each city.
The Pearson’s correlation results showed significant relationships (p <0.001) between the mean
LST and the density of impervious surfaces (positive) and green spaces (negative) in all the study
areas along the urban–rural gradient (Figure 7). The correlation of impervious surfaces with mean
LST in Lagos (r
2
=0.9483; slope =0.0641) and Lusaka (r
2
=0.5766; slope =0.0438) was relatively
high compared to Nairobi (r
2
=0.2783; slope =0.0258) and Addis Ababa (r
2
=0.1776; slope =0.0186).
In contrast, the correlation of green spaces with mean LST was relatively very low in Nairobi
(r
2
=0.3085; slope =–0.0355) compared to Lagos (r
2
=0.9482; slope =–0.0676), Addis Ababa (r
2
=0.7881;
slope =–0.0659) and Lusaka (r2=0.801; slope =−0.0199).
3.3. SUHI Intensity Patterns along the Urban–Rural Gradient
The SUHI intensity results also showed a different pattern in Lagos compared to Nairobi, Addis
Ababa and Lusaka (Figure 8). In Lagos, the SUHI intensity increased from 0.5
◦
C to 4.0
◦
C, which
indicated a high mean LST around the city center compared to other zones along the urban–rural
gradient. For Nairobi and Addis Ababa, the SUHI intensity results showed a similar pattern for Lagos
within the urban area but opposite in the rural areas. The SUHI intensity values for Nairobi ranged
from 0.5
◦
C to 3.0
◦
C along the urban area and reduced from 3.0
◦
C to 1.0
◦
C along the rural area.
The SUHI intensity values for Addis Ababa ranged from 0.5
◦
C to 2.5
◦
C along the urban area and
also reduced from 3.0
◦
C to 1.0
◦
C along the rural area. While Lusaka showed a somewhat similar
pattern to Nairobi and Addis Ababa, the SUHI intensity results generally showed an irregular pattern
of decreasing mean LST along the urban–rural gradient. The SUHI intensity values for Lusaka varied
from about −1.6 ◦C to 0.2 ◦C across the urban–rural gradient.
Remote Sens. 2019,11, 1645 12 of 20
Remote Sens. 2019, 11, x FOR PEER REVIEW 12 of 21
Figure 7. Correlation between mean LST and density of impervious surfaces (a)–(d) and green
spaces (e)–(h).
Figure 7.
Correlation between mean LST and density of impervious surfaces (
a
–
d
) and green spaces (
e
–
h
).
3.4. LST Relationship with Urban Landscape Metrics
The correlations between mean LST and urban landscape metrics varied across the study areas,
with some variables having stronger positive and negative relationships for impervious surfaces and
green spaces, respectively, and others having no relationship at all. The composition variables (PD and
AREA_MN) had significant positive correlations with impervious surface mean LST in all the cities,
except for the PD in Lagos (p=0.807) and Lusaka (p=0.076), and the AREA_MN in Lusaka (p=0.827).
For the complexity variables, SHAPE_MN showed no relationship with impervious surface mean LST
in Lagos (p=0.180) and Lusaka (p=0.758), while FRAC_MN had no relationship in all four cities.
Remote Sens. 2019,11, 1645 13 of 20
The spatial arrangement variable (AI) had significant positive correlations with impervious surface
density mean LST in all the cities excluding Lusaka (p=0.280). For green space mean LST, PD showed
a significant negative correlation only in Lagos, while AREA_MN had significant negative correlations
in all the cities. SHAPE_MN and FRAC_MN had significant positive correlations with green space
mean LST in all the cities, except for FRAC_MN in Nairobi (p=0.272). The correlation between green
space mean LST and AI was insignificant only in Lagos (p=0.085).
Remote Sens. 2019, 11, x FOR PEER REVIEW 13 of 21
3.3. SUHI Intensity Patterns along the Urban–Rural Gradient
The SUHI intensity results also showed a different pattern in Lagos compared to Nairobi, Addis
Ababa and Lusaka (Figure 8). In Lagos, the SUHI intensity increased from 0.5 °C to 4.0 °C, which
indicated a high mean LST around the city center compared to other zones along the urban–rural
gradient. For Nairobi and Addis Ababa, the SUHI intensity results showed a similar pattern for Lagos
within the urban area but opposite in the rural areas. The SUHI intensity values for Nairobi ranged
from 0.5 °C to 3.0 °C along the urban area and reduced from 3.0 °C to 1.0 °C along the rural area. The
SUHI intensity values for Addis Ababa ranged from 0.5 °C to 2.5 °C along the urban area and also
reduced from 3.0 °C to 1.0 °C along the rural area. While Lusaka showed a somewhat similar pattern
to Nairobi and Addis Ababa, the SUHI intensity results generally showed an irregular pattern of
decreasing mean LST along the urban–rural gradient. The SUHI intensity values for Lusaka varied
from about −1.6 °C to 0.2 °C across the urban–rural gradient.
Figure 8. Urban heat island intensity (∆ mean LST) patterns along the urban–rural gradients of Lagos
(a), Nairobi (b), Addis Ababa (c) and Lusaka (d). Note: Urban and rural were discerned based on the
physical extent of the built-up footprint for each city.
3.3. LST Relationship with Urban Landscape Metrics
The correlations between mean LST and urban landscape metrics varied across the study areas,
with some variables having stronger positive and negative relationships for impervious surfaces and
green spaces, respectively, and others having no relationship at all. The composition variables (PD
and AREA_MN) had significant positive correlations with impervious surface mean LST in all the
cities, except for the PD in Lagos (p = 0.807) and Lusaka (p = 0.076), and the AREA_MN in Lusaka (p
= 0.827). For the complexity variables, SHAPE_MN showed no relationship with impervious surface
mean LST in Lagos (p = 0.180) and Lusaka (p = 0.758), while FRAC_MN had no relationship in all
four cities. The spatial arrangement variable (AI) had significant positive correlations with
impervious surface density mean LST in all the cities excluding Lusaka (p = 0.280). For green space
Figure 8.
Urban heat island intensity (
∆
mean LST) patterns along the urban–rural gradients of Lagos
(
a
), Nairobi (
b
), Addis Ababa (
c
) and Lusaka (
d
). Note: Urban and rural were discerned based on the
physical extent of the built-up footprint for each city.
4. Discussion
4.1. Influence of Impervious Surfaces and Green Spaces on LST
In this study, we conducted a comparative study of SUHIs in African cities by examining the
relationship of the spatial patterns, composition and configuration of impervious surfaces and green
spaces with LST using Landsat-8 OLI/TIRS. The results show that Lagos, with the highest percentage
(40%) of impervious surfaces relative to the study area, was the warmest city, i.e., at least 3
◦
C warmer
than Addis Ababa and Lusaka and 4
◦
C warmer than Nairobi. These results could be attributed
to Lagos being a megacity with a population of over 10 million people while the other three cities
still have less than 5 million people [
35
]. These results are dissimilar to the findings of [
8
] in Asian
megacities, where they observed a city with the highest percentage of impervious surfaces to be the
coolest and attributed it to geographical location and background climate.
On the other hand, Nairobi, with the highest and lowest percentage of green spaces and impervious
surfaces, respectively, was the coolest city, i.e., at least 5
◦
C cooler than Lagos and Lusaka and 2
◦
C
cooler than Addis Ababa. The ratio of green spaces to impervious surfaces was also highest in Nairobi
(4.0) and Lowest in Lagos (0.63), while Lusaka and Addis Ababa had ratios of 1.92 and 1.83, respectively.
Despite this, we observed that, although most African cities have a relatively larger green space to
impervious surface ratio (e.g., Addis Ababa, Nairobi and Lusaka) compared to cities in other regions,
the SUHI effect is still evident. This could be because impervious surfaces have a greater impact on
Remote Sens. 2019,11, 1645 14 of 20
surface temperature than green spaces [
8
,
28
,
29
,
34
]. Still, this means that, without the mitigating effect
of green spaces that provide the cool island effect, surface temperatures are expected to escalate. For
example, Lusaka, with the lowest percentage of green spaces, recorded the second highest overall
mean LST of 28.6
◦
C, while Lagos, with the highest percentage of impervious surfaces, had 30.4
◦
C.
Accordingly, Nairobi had the lowest overall mean LST of 25.4
◦
C, while Addis Ababa recorded 27.4
◦
C.
Interestingly, Lusaka had the least overall difference of 1.8
◦
C between the mean LST of impervious
surfaces and green spaces compared to Lagos (4.0 ◦C), Addis Ababa (4.0 ◦C) and Nairobi (4.9 ◦C).
Our results are analogous to other SUHI studies in other regions based on Landsat data as shown
in Figure 9. For example, in Japan, the authors of [
19
] found overall mean LST values of 23.7
◦
C and
24.0
◦
C, with differences between the mean LST of impervious surfaces and green spaces of 1.7
◦
C and
1.8
◦
C in Tsukuba and Tsuchiura, respectively. The authors of [
8
] found overall mean LST values of
27.6
◦
C, 27.9
◦
C and 27.4
◦
C with differences between the mean LST of impervious surfaces and green
spaces of 2.9
◦
C, 3.7
◦
C and 2.2
◦
C in Manila (Philippines), Jakarta (Indonesia) and Bangkok (Thailand),
respectively. In the city of Tehran, Iran, the authors of [
62
] found both a much higher overall mean
LST (43.0
◦
C) and difference (6
◦
C) between the mean LST of impervious surfaces and green spaces .
In another study in Nanjing, China, the authors of [
26
] found an overall mean LST of 30.0
◦
C and a
3.1
◦
C difference between the mean LST of impervious surfaces and green spaces. In a much earlier
study, the authors of [
17
] found an overall mean LST of 29.0
◦
C and a 5.4
◦
C difference between the
mean LST of impervious surfaces and green spaces in Indianapolis City, IN, USA. The variations
in the overall mean LST in this study and the other studies cited above could also be attributed to
geographical location and the respective local climates.
Remote Sens. 2019, 11, x FOR PEER REVIEW 15 of 21
(Thailand), respectively. In the city of Tehran, Iran, the authors of [62] found both a much higher
overall mean LST (43.0 °C) and difference (6 °C) between the mean LST of impervious surfaces and
green spaces . In another study in Nanjing, China, the authors of [26] found an overall mean LST of
30.0 °C and a 3.1 °C difference between the mean LST of impervious surfaces and green spaces. In a
much earlier study, the authors of [17] found an overall mean LST of 29.0 °C and a 5.4 °C difference
between the mean LST of impervious surfaces and green spaces in Indianapolis City, IN, USA. The
variations in the overall mean LST in this study and the other studies cited above could also be
attributed to geographical location and the respective local climates.
Figure 9. Overall mean LST and differences between the mean LST of impervious surfaces (IS) and
green spaces (GS) in African and other cities [8,17,19,26,62].
4.2. Influence of Impervious Surfaces and Green Spaces on LST and SUHI Intensity Patterns along the
Urban–Rural Gradient.
Considering the relatively small urban/built-up areas of Nairobi (8%), Addis Ababa (12%) and
Lusaka (11%) against the unit area of analysis (40 km × 40 km) in this study, for discussion purposes,
we marked urban and rural ranges of the study areas based on the physical extent of the built-up
footprint as defined in remote sensing urban studies (e.g., [58,59]) (see Figures 6 and 8 and Section
2.4.2). While all the African cities present evidence of the SUHI phenomenon, the results show an
interesting variation. Expectedly, the megacity Lagos, which is almost all urban, had the highest
mean LST and UHI intensity in the zones close to the city center, which decreased gradually towards
the rural zones. The pattern of mean LST, SUHI intensity and impervious surface density within the
urban area of Nairobi (0–7 km) and Addis Ababa (0–10 km) along the urban–rural gradient was
somewhat similar to Lagos, with the highest values at the 0 km zone and gradually decreasing to the
cross-point of the urban and rural ranges. The density of green spaces in the two cities gradually
increased from the 0 km zone to the cross-point of the urban and rural ranges. Lusaka, on the other
hand, showed an irregular pattern of mean LST, with its peak at about 7 km within the urban area
(0–10 km) range. Likewise, the SUHI intensity results generally showed an irregular pattern similar
Figure 9.
Overall mean LST and differences between the mean LST of impervious surfaces (IS) and
green spaces (GS) in African and other cities [8,17,19,26,62].
4.2. Influence of Impervious Surfaces and Green Spaces on LST and SUHI Intensity Patterns along the
Urban–Rural Gradient
Considering the relatively small urban/built-up areas of Nairobi (8%), Addis Ababa (12%) and
Lusaka (11%) against the unit area of analysis (40 km
×
40 km) in this study, for discussion purposes,
Remote Sens. 2019,11, 1645 15 of 20
we marked urban and rural ranges of the study areas based on the physical extent of the built-up
footprint as defined in remote sensing urban studies (e.g., [
58
,
59
]) (see Figures 6and 8and Section 2.5.2).
While all the African cities present evidence of the SUHI phenomenon, the results show an interesting
variation. Expectedly, the megacity Lagos, which is almost all urban, had the highest mean LST and
UHI intensity in the zones close to the city center, which decreased gradually towards the rural zones.
The pattern of mean LST, SUHI intensity and impervious surface density within the urban area of
Nairobi (0–7 km) and Addis Ababa (0–10 km) along the urban–rural gradient was somewhat similar to
Lagos, with the highest values at the 0 km zone and gradually decreasing to the cross-point of the
urban and rural ranges. The density of green spaces in the two cities gradually increased from the
0 km zone to the cross-point of the urban and rural ranges. Lusaka, on the other hand, showed an
irregular pattern of mean LST, with its peak at about 7 km within the urban area (0–10 km) range.
Likewise, the SUHI intensity results generally showed an irregular pattern similar to mean LST along
the urban–rural gradient. The pattern of the impervious surface and green space density in Lusaka
could help explain the irregular pattern of mean LST and SUHI intensity, which can be likened to
the findings of [
8
] in Bangkok and Manila. This is because, although the African cities in this study
generally have their green spaces located outside the urban zones, Lusaka appears to have some green
spaces within the urban area, especially in the eastern part of the city.
Another key observation in this study was the gradual increase in the mean LST (low SUHI
intensity) in Nairobi, Addis Ababa and Lusaka within the defined rural area from the urban–rural
cross-point. This could be explained based on the land cover in the study areas. Unlike Lagos,
where the remaining land cover beyond the urban footprint was dominated by water, the other three
cities’ remaining land cover was mainly characterized by bare lands and abandoned crop fields. This
could have contributed to the observed higher LST values as bare lands can also elevate surface
temperatures [18,63].
4.3. Influence of Spatial Landscape Configuration on LST
In this study, we used five spatial metrics (Table 3) to assess the influence of the spatial landscape
configuration (i.e., composition, shape, complexity and spatial arrangement) on mean LST. Generally,
the results show that the correlation between mean LST and the selected spatial metrics was statistically
significant, i.e., positive for impervious surfaces and negative for green spaces. These results are
consistent with several previous studies. For example, the authors of [
6
] found significant relationships
between mean LST and the PD of patches of residential impervious surfaces (positive) and urban
green spaces (negative) in Shanghai, China. In Baltimore, MD, USA, the authors of [
7
] correlated mean
LST with the AREA_MN and SHAPE_MN and found significant positive and negative relationships
from the patches of impervious surfaces and green spaces, respectively. The authors of [
8
] also found
significant relationships between mean LST and the AI of the patches of impervious surface (positive)
and green space (negative) in megacities in Asia.
Our results indicate that cities with large and more aggregated patches of impervious surfaces
experience significant increases in LST, exacerbating the SUHI phenomena [
7
], than those with
fragmented smaller patches of the impervious surface [
8
]. Lagos, for example, had the largest patches
and showed significant correlation between mean LST and the AREA_MN and AI of the patches
impervious surfaces, which could explain the high surface temperatures in Lagos. Similarly, Nairobi
recorded significant correlation between mean LST and the AI of the patches of green spaces, despite
having large patches of green spaces. Addis Ababa and Lusaka had more dispersed patches of green
spaces. This explains the higher surface temperatures in Addis Ababa and Lusaka than in Nairobi.
This is in agreement with other studies that have shown that the size of green spaces is an important
factor in mitigating the SUHI effects [
6
,
64
]. The spatial arrangement of green spaces is also important
in providing the cool island effect [
65
]. Larger and contiguous green spaces produce stronger cool
island effects than those of several smaller patches of green space whose total area equals the large,
contiguous patches [14,65].
Remote Sens. 2019,11, 1645 16 of 20
Table 3. Correlations between mean LST and selected urban landscape metrics.
Study Area PD AREA_MN SHAPE_MN FRAC_MN AI
rSig. rSig. rSig. rSig. rSig.
Impervious Surface Mean LST vs. Spatial Metrics
Lagos 0.027 0.807 0.519 0.000 0.147 0.180 0.182 0.096 0.432 0.000
Nairobi 0.339 0.002 0.383 0.001 0.277 0.014 0.148 0.197 0.402 0.000
Addis Ababa 0.540 0.000 0.244 0.029 0.265 0.018 0.197 0.080 0.289 0.009
Lusaka 0.199 0.076 0.025 0.827 0.035 0.758 0.097 0.391 0.122 0.280
Green Space Mean LST vs. Spatial Metrics
Lagos −0.362 0.001 −0.316 0.004 −0.421 0.000 −0.378 0.000 −0.190 0.085
Nairobi −0.151 0.179 −0.281 0.011 −0.418 0.000 −0.123 0.272 −0.700 0.000
Addis Ababa −0.135 0.229 −0.221 0.047 −0.489 0.000 −0.485 0.000 −0.321 0.004
Lusaka −0.189 0.091 −0.313 0.004 −0.336 0.002 −0.298 0.007 −0.244 0.028
More interestingly, the results consistently showed a low complexity of the impervious surface
patches in all the African cities. While other factors may be at play, it is plausible to speculate that this
could be caused by the unplanned urban developed pattern in Afrcian cities that tends to be clustered
around the city center [
5
]. This could also be the reason for the lack of greenspaces within the urban
area. Of note is that this pattern of development is dominated by highly dense informal settlements
that are more susceptible to the SUHI effect.
4.4. Implications for Mitigating SUHIs in African Cities
The results of this study have shown that, although most African cities have a relatively larger
green space to impervious surface ratio compared to cities in other regions, the SUHI effect is still
evident. Altogether, we observed that there is a clear separation between the impervious surfaces and
green spaces in the African cities. Most of the green spaces are found beyond the urban area. This
could have emanated from the unplanned and uncontrolled urbanization of African cities that has
been well documented. Unplanned urban development has likely worsened the SUHI effects. This
means urban areas have continuously lost ecosystem services in the process. The observed high values
of mean LST and impervious surfaces within the zones close to the city center in this study present a
typical case of cities that follow central business district (CBD)-oriented urban development, which
characterizes most African cities. The implication is that urban planners and policy makers in African
cities, while attempting to control the unplanned development, should consider restoring the urban
ecosystems through a diverse set of habitats by increasing the amount of vegetation in the remaining
opens spaces such as parks, cemeteries, vacant lots, gardens and yards [
66
]. Increasing vegetation
cover or surface water could significantly decrease LST, and thus help to mitigate excess heat in urban
areas [7].
This study supports the findings of various other SUHI studies in Africa and other regions that
recommend incorporating strategic combinations of impervious surfaces and green spaces in future
urban and landscape planning to mitigate the SUHI effect. Some of the mitigation strategies in other
studies proposed that African urban planners and policy makers could consider: The use of green walls
that can mitigate indoor temperatures in tropical countries by about 2.4
◦
C [
34
,
67
]; the establishment
of green belts along the main roads and residential areas to promote cool islands that can reduce
heat stress and energy demand for urban dwellers [
33
]; as well as encouraging vertical, rather than
horizontal, urban development to preserve space for urban greening [68].
5. Conclusions
Taking four cities, Lagos (Nigeria), Nairobi (Kenya), Addis Ababa (Ethiopia) and Lusaka (Zambia),
a comparative study of SUHIs in African cities was conducted by examining the relationship of the
spatial patterns, composition and configuration of impervious surfaces and green spaces with LST
using Landsat-8 OLI/TIRS data. The study employed various techniques: Urban–rural gradient, urban
Remote Sens. 2019,11, 1645 17 of 20
heat island intensity, urban landscape metrics and statistical analysis. The results show a significantly
strong correlation between mean LST and the density of impervious surface (positive) and green space
(negative) along the urban–rural gradients of the four African cities. The study also found high urban
heat island intensities in the urban area zones within the 0 to 10 km distance from the city center,
where the density of green space is low. We also found significant correlations between the mean LST
and urban landscape metrics (patch density, size, shape, complexity and aggregation) of impervious
surfaces (positive) and green spaces (negative). The observed high values of mean LST and impervious
surfaces within the zones close to the city center in this study present a typical case of cities that follow
CBD-oriented urban development, which characterizes most African cities. We, therefore, suggest the
urban planners and policy makers in African cities should consider decentralizing through setting
up satellite economic zones in the periphery rural areas. The SUHI effects can then be mitigated by
restoring the urban ecosystems in the remaining open areas such as parks, cemeteries, vacant lots,
gardens, yards and campus areas; and blue spaces, mainly, streams, ponds and dams.
This study has further revealed that, although most African cities have a relatively larger green
space to impervious surface ratio compared to cities in other regions, the SUHI effect is still evident.
We found that cities with a larger percentage of urban area relative to the study area unit were warmer,
i.e., they had mean LST values at least 3
−
4
◦
C higher than the coolest city, resulting in strong SUHI
effects. Accordingly, the important mitigating effect of green spaces has been highlighted, with the
coolest city having the largest percentage of green space. Another important observation highlighted
in this study is that there is a general separation between the impervious surfaces and green spaces in
the African cities. Most of the green spaces are found beyond the urban area. The results revealed a
distinct variation in the relationship of mean LST with the density of impervious surfaces and green
spaces within and beyond the urban footprint, especially in the cities with relatively small urban
footprints. We attribute this to the unplanned and uncontrolled urbanization of African cities that
have potentially worsened the SUHI effects. It is therefore recommended that urban planners and
policy makers in African cities, while attempting to control the unplanned development, should
consider the dispersion of built-up areas and paved surfaces (e.g., buildings, roads and parking lots)
and maintaining or improving vegetation (e.g., grass, shrubs and trees) cover. The study, therefore,
provides useful information that can help control the effects of the uncontrolled and unplanned
urbanization in Africa to provide better urban environmental conditions for the urban dwellers and
further encourage sustainable urban development in African cities.
In terms of future research, the current study did not evaluate the sensitivity to grid-spacing when
examining the influence of impervious surface and green space on LST in African cities. This is an area
worth investigating in future studies.
Author Contributions:
All the authors (M.S., M.R., R.C.E. and Y.M.) participated in the research concept design
and implementation, data processing and analysis, and writing of the manuscript.
Funding:
This research was supported by the Japan Society for the Promotion of Science (JSPS) through
Grant-in-Aid for Scientific Research (B) 18H00763 (2018-20, representative: Yuji Murayama).
Acknowledgments:
The authors are grateful to the editor and the anonymous reviewers for their helpful comments
and suggestions to improve the quality of this paper.
Conflicts of Interest: The authors declare no conflicts of interest.
References
1.
Guneralp, B.; Shuaib, L.; Hillary, M.; Parnell, S.; Seto, K.C. Urbanization in Africa: Challenges and
opportunities for conservation Urbanization in Africa: Challenges and opportunities for conservation.
Environ. Res. Lett. 2018,13. [CrossRef]
2.
Roxburgh, C.; Dörr, N.; Leke, A.; Tazi-Riffi, A.; Van Wamelen, A.; Lund, S.; Zeino-Mahmalat, T. Lions on the
Move: The Progress and Potential of African Economies; McKinsey Global Institute: New York, NY, USA, 2010.
Remote Sens. 2019,11, 1645 18 of 20
3.
UN-DESA. World Urbanization Prospects: The 2018 Revision, Highlights; United Nations publication; United
Nations,Department ofEconomic and Social Affairs(UN/DESA),PopulationDivision: 2018. Availableonline: https:
//www.un.org/development/desa/publications/2018-revision-of-world-urbanization-prospects.html/(accessed on
23 June 2019).
4.
Parnell, S.; Pieterse, E. Africa’s Urban Revolution; Parnell, S., Pieterse, E., Eds.; Zed Books Ltd.: London, UK;
New York, NY, USA, 2014.
5.
Simwanda, M.; Murayama, Y. Spatiotemporal patterns of urban land use change in the rapidly growing city
of Lusaka, Zambia: Implications for sustainable urban development. Sustain. Cities Soc.
2018
,39, 262–274.
[CrossRef]
6.
Li, J.; Song, C.; Cao, L.; Zhu, F.; Meng, X.; Wu, J. Impacts of landscape structure on surface urban heat islands:
A case study of Shanghai, China. Remote Sens. Environ. 2011,115, 3249–3263. [CrossRef]
7.
Zhou, W.; Huang, G.; Cadenasso, M.L. Does spatial configuration matter? Understanding the effects of
land cover pattern on land surface temperature in urban landscapes. Landsc. Urban Plan.
2011
,102, 54–63.
[CrossRef]
8.
Estoque, R.C.; Murayama, Y.; Myint, S.W. Effects of landscape composition and pattern on land surface temperature:
An urban heat island study in the megacities of Southeast Asia. Sci. Total Environ.
2017
,577, 349–359. [CrossRef]
[PubMed]
9.
Zhang, X.; Estoque, R.C.; Murayama, Y. An urban heat island study in Nanchang City, China based on land
surface temperature and social-ecological variables. Sustain. Cities Soc. 2017,32, 557–568. [CrossRef]
10.
Ranagalage, M.; Estoque, R.; Zhang, X.; Murayama, Y. Spatial changes of urban heat island formation in the
Colombo District, Sri Lanka: Implications for sustainability planning. Sustainability
2018
,10, 1367. [CrossRef]
11.
Buyantuyev, A.; Wu, J. Urban heat islands and landscape heterogeneity: Linking spatiotemporal variations
in surface temperatures to land-cover and socioeconomic patterns. Landsc. Ecol.
2010
,25, 17–33. [CrossRef]
12.
Voogt, J.A. Urban Heat Islands: Hotter Cities. Available online: http://www.actionbioscience.org/
environment/voogt.html (accessed on 29 April 2018).
13.
EPA (US Environmental Protection Agency). Reducing Urban Heat Islands: Compendium of Strategies; US
Environmental Protection Agency: Washington, DC, USA, 2008.
14.
Li, X.; Zhou, W.; Ouyang, Z. Relationship between land surface temperature and spatial pattern ofgreenspace:
What are the effects of spatial resolution? Landsc. Urban Plan. 2013,114, 1–8. [CrossRef]
15.
Weng, Q. Thermal infrared remote sensing for urban climate and environmental studies: Methods,
applications, and trends. ISPRS J. Photogramm. Remote Sens. 2009,64, 335–344. [CrossRef]
16.
Liu, L.; Zhang, Y. Urban heat island analysis using the Landsat TM Data and ASTER Data: A Case Study in
Hong Kong. Remote Sens. 2011,3, 1535–1552. [CrossRef]
17.
Weng, Q.; Lu, D.; Schubring, J. Estimation of land surface temperature-vegetation abundance relationship
for urban heat island studies. Remote Sens. Environ. 2004,89, 467–483. [CrossRef]
18.
Sun, Q.; Wu, Z.; Tan, J. The relationship between land surface temperature and land use/land cover in
Guangzhou, China. Environ. Earth Sci. 2012,65, 1687–1694. [CrossRef]
19.
Ranagalage, M.; Estoque, R.C.; Handayani, H.H.; Zhang, X.; Morimoto, T.; Tadono, T.; Murayama, Y. Relation
between urban volume and land surface temperature: A comparative study of planned and traditional cities
in Japan. Sustainability 2018,10, 2366. [CrossRef]
20.
Liu, H.; Zhan, Q.; Gao, S. Seasonal Variation of the Spatially Non-Stationary Association Between Land
Surface Temperature and Urban Landscape. Remote Sens. 2019,11, 1016. [CrossRef]
21.
Xiao, R.; Ouyang, Z.; Zheng, H.; Li, W.; Schienke, E.W.; Wang, X. Spatial pattern of impervious surfaces
and their impacts on land surface temperature in Beijing, China. J. Environ. Sci. (China)
2007
,19, 250–256.
[CrossRef]
22.
Ogashawara, I.; Brum, S. A Quantitative approach for analyzing the relationship between urban heat islands
and land cover. Remote Sens. 2012,4, 3596–3618. [CrossRef]
23.
Estoque, R.C.; Murayama, Y. Monitoring surface urban heat island formation in a tropical mountain city
using Landsat data (1987–2015). ISPRS J. Photogramm. Remote Sens. 2017,133, 18–29. [CrossRef]
24.
Maimaitiyiming, M.; Ghulam, A.; Tiyip, T.; Pla, F.; Latorre-Carmona, P.; Halik, Ü.; Sawut, M.; Caetano, M.
Effects of green space spatial pattern on land surface temperature: Implications for sustainable urban
planning and climate change adaptation. ISPRS J. Photogramm. Remote Sens. 2014,89, 59–66. [CrossRef]
Remote Sens. 2019,11, 1645 19 of 20
25.
Ranagalage, M.; Dissanayake, D.; Murayama, Y.; Zhang, X.; Estoque, R.C.; Perera, E.; Morimoto, T.
Quantifying surface urban heat island formation in the world heritage tropical mountain city of Sri Lanka.
Int. J. Geo-Inf. 2018,7, 341. [CrossRef]
26.
Kong, F.; Yin, H.; James, P.; Hutyra, L.R.; He, H.S. Effects of spatial pattern of greenspace on urban cooling in
a large metropolitan area of eastern China. Landsc. Urban Plan. 2014,128, 35–47. [CrossRef]
27.
Bowler, D.E.; Buyung-ali, L.; Knight, T.M.; Pullin, A.S. Urban greening to cool towns and cities: A systematic
review of the empirical evidence. Landsc. Urban Plan. 2010,97, 147–155. [CrossRef]
28.
Ranagalage, M.; Estoque, R.C.; Murayama, Y. An urban heat island study of the Colombo Metropolitan Area,
Sri Lanka, based on Landsat Data (1997–2017). ISPRS Int. J. Geo-Inf. 2017,6, 189. [CrossRef]
29.
Rousta, I.; Sarif, O.; Gupta, R.D.; Olafsson, H.; Ranagalage, M.; Murayama, Y.; Zhang, H.; Mushore, T.D.
Spatiotemporal analysis of land use /land cover and its effects on surface urban heat island using Landsat
Data: A case study of Metropolitan City Tehran. Sustainability 2018,10, 4433. [CrossRef]
30.
Memon, R.A.; Leung, D.Y.C.; Liu, C.H. An investigation of urban heat island intensity (UHII) as an indicator
of urban heating. Atmos. Res. 2009,94, 491–500. [CrossRef]
31.
Sheng, L.; Tang, X.; You, H.; Gu, Q.; Hu, H. Comparison of the urban heat island intensity quantified by
using air temperature and Landsat land surface temperature in Hangzhou. Ecol. Indic.
2017
,72, 738–746.
[CrossRef]
32.
Ngie, A.; Abutaleb, K.; Ahmed, F.; Taiwo, O.J.; Darwish, A.A.; Ahmed, M. An estimation of land surface
temperatures from Landsat ETM +images for Durban, South Africa. Rwanda J. 2008,1. [CrossRef]
33.
Dissanayake, D.; Morimoto, T.; Murayama, Y.; Ranagalage, M.; Handayani, H.H. Impact of urban surface
characteristics and socio-economic variables on the spatial variation of land surface temperature in Lagos
City, Nigeria. Sustainability 2018,11, 25. [CrossRef]
34.
Dissanayake, D.; Morimoto, T.; Murayama, Y.; Ranagalage, M. Impact of landscape structure on the variation
of land surface temperature in sub-saharan region: A case study of Addis Ababa using Landsat Data.
Sustainability 2019,11, 2257. [CrossRef]
35.
UN-DESA The World’s Cities in 2016—Data Booklet (ST/ESA/SER.A/392); United Nations publication; United
Nations, Department of Economic and Social Affairs (UN/DESA), Population Division: 2016. Available
online: https://bit.ly/2JBQ2S6 (accessed on 23 June 2019).
36.
ClimaTemps World Climate and Temperature; Lagos, Addis Ababa, Nairobi and Lusaka. Available online:
http://www.climatemps.com/(accessed on 23 January 2019).
37.
Simwanda, M.; Murayama, Y. Integrating geospatial techniques for urban land use classification in the
developing sub-saharan African City of Lusaka, Zambia. ISPRS Int. J. Geo-Inf. 2017,6, 102. [CrossRef]
38.
Gong, H.; Simwanda, M.; Murayama, Y. An internet-based GIS platform providing data for visualization
and spatial analysis of urbanization in major Asian and African Cities. ISPRS Int. J. Geo-Inf.
2017
,6, 257.
[CrossRef]
39.
Tran, H.; Uchihama, D.; Ochi, S.; Yasuoka, Y. Assessment with satellite data of the urban heat island effects in
Asian mega cities. Int. J. Appl. Earth Obs. Geoinf. 2006,8, 34–48. [CrossRef]
40.
Chander, G.; Markham, B.L.; Helder, D.L. Summary of current radiometric calibration coefficients for Landsat
MSS, TM, ETM+, and EO-1 ALI sensors. Remote Sens. Environ. 2009,113, 893–903. [CrossRef]
41.
USGS (United States Geological Survey). Product Guide: Provisional Landsat 8 Surface Reflectance Product;
Department of the Interior, U.S. Geological Survey: Reston, VA, USA, 2016.
42.
Sobrino, J.A.; Jim
é
nez-Muñoz, J.C.; Paolini, L. Land surface temperature retrieval from LANDSAT TM 5.
Remote Sens. Environ. 2004,90, 434–440. [CrossRef]
43.
Artis, D.A.; Carnahan, W.H. Survey of emissivity variability in thennography of urban areas. Remote Sens.
Environ. 1982,12, 313–329. [CrossRef]
44.
Xu, H. Modification of normalised difference water index (NDWI) to enhance open water features in remotely
sensed imagery. Int. J. Remote Sens. 2006,27, 3025–3033. [CrossRef]
45.
Estoque, R.C.; Murayama, Y. Classification and change detection of built-up lands from Landsat-7 ETM +
and Landsat-8 OLI /TIRS imageries: A comparative assessment of various spectral indices. Ecol. Indic.
2015
,
56, 205–217. [CrossRef]
46.
Zha, Y.; Gao, J.; Ni, S. Use of normalized di fference built-up index in automatically mapping urban areas
from TM imagery. Int. J. Remote Sens. 2003,24, 583–594. [CrossRef]
Remote Sens. 2019,11, 1645 20 of 20
47.
Zhang, L.; Wu, J.; Zhen, Y.; Shu, J. A GIS-based gradient analysis of urban landscape pattern of Shanghai
metropolitan area, China. Landsc. Urban Plan. 2004,69, 1–16. [CrossRef]
48.
Zhou, X.; Wang, Y. Spatial—Temporal dynamics of urban green space in response to rapid urbanization and
greening policies. Landsc. Urban Plan. 2011,100, 268–277. [CrossRef]
49.
Kroll, F.; Müller, F.; Haase, D.; Fohrer, N. Rural—Urban gradient analysis of ecosystem services supply and
demand dynamics. Land Use Policy 2012,29, 521–535. [CrossRef]
50. Oke, T.R. City size and the urban heat island. Atmos. Environ. 1973,7, 769–779. [CrossRef]
51.
Levermore, G.; Parkinson, J.; Lee, K.; Laycock, P.; Lindley, S. The increasing trend of the urban heat island
intensity. Urban Clim. 2017,24, 360–368. [CrossRef]
52.
Takebayashi, H.; Senoo, M. Analysis of the relationship between urban size and heat island intensity using
WRF model. Urban Clim. 2016,24, 287–298. [CrossRef]
53.
Martin-vide, J.; Sarricolea, P.; Moreno-garc
í
a, M.C. On the definition of urban heat island intensity: The
“rural” reference. Front. Earth Sci. 2015,3, 24. [CrossRef]
54.
Montgomery, M.; Stren, R.; Cohen, B.; Reed, H. Cities Transformed: Demographic Change and Its Implications in
the Developing World; National Academies Press: Washington, DC, USA, 2003.
55.
Potere, D.; Schneider, A.; Angel, S.; Civco, D.L. Mapping urban areas on a global scale: Which of the eight
maps now available is more accurate? Int. J. Remote Sens. 2009,30, 6531–6558. [CrossRef]
56.
Zlatanova, S.; Ledoux, H.; Fendel, E.M.; Rumor, M. Urban and regional data management: UDMS Annual
2011. In Urban Data Management Society Symposium, Delft, The Netherlands, 28-30 Septemebr 2011; CRC
Press/Balkema: Leiden, The Netherlands, 2012.
57.
Taubenböck, H.; Esch, T.; Felbier, A.; Wiesner, M.; Roth, A.; Dech, S. Monitoring urbanization in mega cities
from space. Remote Sens. Environ. 2012,117, 162–176. [CrossRef]
58.
Herold, M.; Scepan, J.; Clarke, K.C. The use of remote sensing and landscape metrics to describe structures
and changes in urban land uses. Environ. Plan. A 2002,34, 1443–1458. [CrossRef]
59.
Van de Voorde, T.; Jacquet, W.; Canters, F. Mapping form and function in urban areas: An approach based on
urban metrics and continuous impervious surface data. Landsc. Urban Plan. 2011,102, 143–155. [CrossRef]
60.
Mcgarigal, K. FRAGSTATS v4.2: Spatial Pattern Analysis Program for Categorical and Continuous Maps.
Computer Produced by the Authors. 2015, pp. 1–182. Available online: https://www.umass.edu/landeco/
research/fragstats/downloads/fragstats_downloads.html/(accessed on 23 January 2019).
61.
Forman, R.T.T. Land Mosaics, the Ecology of Landscapes and Regions; Cambridge University Press: Cambridge,
UK, 1995.
62.
Bokaie, M.; Zarkesh, M.K.; Arasteh, P.D.; Hosseini, A. Assessment of urban heat island based on the
relationship between land surface temperature and Land Use/Land Cover in Tehran. Sustain. Cities Soc.
2016
,
23, 94–104. [CrossRef]
63. Haylemariyam, M.B. Detection of land surface temperature in relation to land use land cover change: Dire
Dawa City, Ethiopia. J. Remote Sens. GIS 2018,7, 245. [CrossRef]
64.
Chang, C.; Li, M.; Chang, S. A preliminary study on the local cool-island intensity of Taipei city parks.
Landsc. Urban Plan. 2007,80, 386–395. [CrossRef]
65.
Zhang, X.; Zhong, T.; Feng, X.; Wang, K. Estimation of the relationship between vegetation patches and
urban land surface temperature with remote sensing. Int. J. Remote 2009,8, 2105–2118. [CrossRef]
66.
Elmqvist, T.; Setälä, H.; Handel, S.N.; van der Ploeg, S.; Aronson, J.; Blignaut, J.N.; G
ó
mez-Baggethun, E.;
Nowak, D.J.; Kronenberg, J.; de Groot, R. Benefits of restoring ecosystem services in urban areas. Curr. Opin.
Environ. Sustain. 2015,14, 101–108. [CrossRef]
67.
Galagoda, R.U.; Jayasinghe, G.Y.; Halwatura, R.U.; Rupasinghe, H.T. The impact of urban green infrastructure
as a sustainable approach towards tropical micro-climatic changes and human thermal comfort. Urban For.
Urban Green. 2018,34, 1–9. [CrossRef]
68.
Lin, J.; Huang, B.; Chen, M.; Huang, Z. Modeling urban vertical growth using cellular automata d Guangzhou
as a case study. Appl. Geogr. 2014,53, 172–186. [CrossRef]
©
2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).