Urban Heat Island Effect of Addis Ababa

City: Implications of Urban Green Spaces

for Climate Change Adaptation

Ermias Teferi and Hiwot Abraha

1 Introduction

Change in land use and land cover (LULC), mainly conversion of natural envi-
ronments into impervious surfaces, has become a major environmental concern
worldwide (Weng 2001). In Ethiopia, the rapid population growth together with the
fast economic growth has accelerated unplanned growth of urban or built-up areas.
Urbanization causes drastic changes in vegetation cover, hydrological regimes and
local scale climates. The most obvious climatic impact of urbanization is increase in
land surface temperature (LST) in urban areas relative to the surrounding rural areas
(Zhou and Wang 2011; Pongracz et al. 2006), and this phenomenon is referred to as
the Urban Heat Island (UHI) effect. Elevated temperatures (i.e. heatwaves) can
affect air quality (IPCC 2013) and human health (WHO 2013). Heatwaves can have
undesired effects on urban population. The effect of heatwaves on human health is
exemplified by the deaths of 70,000 Europeans in 2003 (Robine et al. 2008). Thus,
the increasing frequency and intensity of heatwaves is becoming an important
health concern for policymakers (WHO 2013).
The LST is an important factor controlling many physical, chemical and bio-
logical processes of the Earth. Thus, knowledge of the LST is necessary for many
environmental studies and management activities of the Earth’s surface resources
(Li and Becker 1993). It is one of the key parameters in the physics of land surface
processes on regional and global scales, combining the results of all surface–at-
mosphere interactions and energy fluxes between the atmosphere and the ground
(Mannstein 1987).

E. Teferi (&)
Center for Environment and Development Studies,
Addis Ababa University, Addis Ababa, Ethiopia
e-mail: [email protected]
H. Abraha
Addis Ababa Environmental Protection Authority, Addis Ababa, Ethiopia

© Springer International Publishing AG 2017 539

W. Leal Filho et al. (eds.), Climate Change Adaptation in Africa,
Climate Change Management, DOI 10.1007/978-3-319-49520-0_33
540 E. Teferi and H. Abraha

Changes in LST due to urbanization can be assessed by comparing historical

point data from meteorological stations located within urban areas and in the sur-
rounding rural areas where such data exist. In the absence of such records, satellite
remote sensing data can provide useful information to evaluate land surface tem-
perature changes in response to urbanization. Previous studies have demonstrated
that land surface temperature product retrieved from thermal infrared (TIR) sensors
can be used to monitor the UHI effect. For example, Pongracz et al. (2006) used
LST time series data derived from Moderate Resolution Imaging Spectroradiometer
(MODIS) to determine UHI intensities over ten most populated cities of Hungary.
Hung et al. (2006) analyzed the UHI effect in 18 Asian mega cities using
MODIS LST products acquired between 2001 and 2003. They examined spatial
patterns of UHI effects for each city in a diurnal cycle and seasonal variations. They
found that both the magnitude and extent of UHI effect were positively correlated
with population size of the cities, indicating the significant impact of urban
expansion and population growth on local and global climates. Amiri et al. (2009)
examined the relationship between land cover changes and LST using Landsat
images for urban area of Iran. However, in Ethiopia, no study has so far been
conducted, on the Changes in LST due to LUCC.
The impacts of LUCC on thermal environment change can be investigated in
two ways. The first approach involves the comparison of the LST of categorical
land cover data (Chen et al. 2006; Xiao and Weng 2007). The second approach
involves the analysis of LST based on remote sensing indices such as the
Normalized Difference Vegetation Index (NDVI) (Yuan and Bauer 2007) and the
Normalized Difference Built-up Index (NDBI) (Zha et al. 2003). The NDVI is a
proxy for greenness of an area (Chen and Brutsaert 1998). The NDBI is an indicator
of built-up and barren land.
The objective of this study is to appraise the urban heat island effect of Addis
Ababa, which is the capital and largest city of Ethiopia. The specific objectives
were to (i) determine extent and rate of expansion of Addis Ababa city since 1986,
(ii) investigate the urban heat island effect of Addis Ababa city using remote
sensing and GIS technologies, and (iii) suggest some interventions that could
reduce the urban heat island effect in the city. The paper is divided into four
sections. The following section (Sect. 2) presents descriptions of the study area and
data and methods of the study. Section 3 presents results and discussion, and the
final section (Sect. 4) concludes the study.

2 Materials and Methods

2.1 Study Site: Addis Ababa City

Addis Ababa city, the capital of Ethiopia, is located between 8° 50′N–9° 5′N and
38° 38′E–38° 52′E. The city lies at the foot of an isolated mountain called Entoto,
and elevation varies from 2015 to 3150 m a.s.l (Fig. 1). The high elevation
Urban Heat Island Effect of Addis Ababa City … 541

Fig. 1 Location map of the study area

moderates temperatures year-round, and the city’s position near the equator means
that temperatures are relatively constant from month to month. The city has
increasingly expanded southward, westward and eastward. Expansion to the north
is limited because of topography. Based on the 2007 census results, Addis Ababa
has a total population of 2,738,248, consisting of 1,304,518 men and 1,433,730
women. The city is fully urban, with no rural dwellers within the city’s adminis-
trative boundaries. For the capital city 662,728 households were counted living in
628,984 housing units, which results in an average of 4.2 persons to a household.
Addis Ababa contains 22.9% of all urban dwellers in Ethiopia and for about 3.7%
of the total population of the country.

2.2 Data Used and Image Pre-processing

The study used topographic map of the area at a scale of 1:50,000 dated 1984 and
Landsat images of years 1986 (TM) and 2011 (TM). Scenes were required to be of
the same phenological cycle (dry season) and have little or no cloud cover. All
Landsat images were accessed free of charge from U.S. Geological Survey (USGS)
Center for Earth Resources Observation and Science (EROS) (via http://glovis.usgs.
gov/). All scenes supplied by the EROS Data Center had already been
geo-referenced to the Universal Transverse Mercator (UTM) map projection (Zone
37), WGS 84 datum and ellipsoid. Re-projection to the local level projection system
542 E. Teferi and H. Abraha

was made (UTM, map projection; Clarke 1880, Spheroid, and Adindan Datum).
High resolution SPOT-5 imagery (5 m) was also used for visual feature identifi-
cation and collection of inaccessible ground control points.
In order to eliminate the effects of atmospheric scattering and absorption in the
image and to increase the accuracy of land cover classification, the original Digital
Number (DN) values were converted to reflectance at the surface of the earth by
using FLASH atmospheric correction. This procedure is divided into two stages:
(i) converting DN values to spectral radiance and (ii) transferring the sensor
detected radiance into surface reflectance.

2.3 Image Classification

A hybrid image classification based on the combined use of a thresholding tech-

nique, and supervised/unsupervised classification approach was used to classify the
1986 and 2011 images. The unsupervised classification was carried out using the
Iterative Self-Organizing Data Analysis (ISODATA) algorithm to identify spectral
clusters in the images. The ISODATA method uses minimum spectral distance to
assign a cluster for each candidate pixel in image classification (Ball and Hall
1965). Based on the results of the unsupervised classifications, training sites were
chosen from the images. For each image, spectral signatures for the training sites
were carefully chosen and examined and also using the spectral clusters ground
truth were collected to associate the spectral classes. A maximum likelihood clas-
sifier (MLC) was then employed for the image classification. MLC is one of the
well-known parametric classifiers used for supervised image classification (Foody
et al. 1992). The advantage of the MLC as a parametric classifier is that it takes into
account the variance–covariance within the class distributions. Two remote sensing
indices were computed to assist land cover types identification. Normalized
Difference Vegetation Index (NDVI) was computed using Eq. 1 for the purpose of
estimating emissivity. In Eq. 1, q4 represents reflectance of the near infrared band
of TM and q3 represents the red band of TM image.
q4  q3
NDVI ¼ ð1Þ
q4 þ q3

To improve the accuracy of bare land identification Zhao and Chen (2005)
introduced Normalized Difference Bareness Index (NDBaI). NDBaI was used to
extract the different types of bare lands: primary bare land (no vegetation), bare land
associated with grassland and cultivated soils. Band 5 and Band 6 are the Short
Wave Infrared (SWIR) and Thermal Infrared (TIR) bands Landsat TM, respectively.

Band 5  Band 6
NDBaI ¼ ð2Þ
Band 5 þ Band 6
Urban Heat Island Effect of Addis Ababa City … 543

2.4 Land Surface Temperature Retrieval

To obtain a reasonably good quality of LST, four steps may be required (Fig. 2):
(1) conversion to at-sensor spectral radiance; (2) land surface emissivity estimation;
(3) atmospheric correction; and (4) LST retrieval.

Fig. 2 Methodological flowchart

544 E. Teferi and H. Abraha

Conversion to at-sensor spectral radiance: At-sensor spectral radiance (Lsen)

was computed for the Landsat 5-TM quantized calibrated pixel values in DNs
(Qcal ). The conversion from Qcal to at-sensor spectral radiance (Lsen) is performed
using sensor calibration parameters published by Chander et al. (2009) and in image
header file by following equation:

Lsen ¼ Grescale  Qcal þ Brescale ð3Þ

where
Lsen = Spectral radiance at the sensor’s aperture ½W=ðm2 srlmÞ
Qcal = Quantized calibrated pixel value ½DN 
Grescale = Band 6 rescaling gain factor ½ðW=ðm2 srlmÞÞ=DN  (0.05518).
Brescale = Band 6 rescaling bias factor ½W=ðm2 srlmÞ (1.2378). Land surface
emissivity estimation: Emissivity for ground objects from passive sensor data has
been estimated using different techniques. Among other techniques, NDVI methods
are easy to apply (Van de Griend and Owe 1993). A slight modification of the first
one was used in this paper to make the equation fit to the local condition of the
study area (Eq. 4). To derive emissivity image the above equation was written in
the spatial modeler of ERDAS.
(
0:99; NDVI  0:01
e¼ ð4Þ
1:0094 þ 0:047  lnðNDVIÞ; otherwise

Atmospheric correction: The impacts of the atmosphere and the emitted ground
are unavoidably involved in the sensor-observed radiance. Thus, correction is
necessary for retrieving true LST from Landsat TM6 data. An atmospheric cor-
rection was applied to thermal band of Landsat TM using local values for several
meteorological parameters. An online atmospheric correction parameter calculator
(http://atmcorr.gsfc.nasa.gov/) was used to calculate the atmospheric-correction
parameters (τ, L↑, and L↓) required in the atmospheric radiative transfer equation.
However the calculator will not work to correct scenes earlier than 2000. To
overcome this problem similar parameters with that of the 2011 imagery were used
to correct for the earlier scene (e.g. 1986). Since the scenes are from the same time
of year, it is possible to assume a similar atmospheric condition.
The radiance acquired by a thermal band of the remote sensor can be described
quantitatively with the Radiative Transfer Equation (RTE) as follows:

Lsen  L " 1  e
BðLSTÞ ¼  L# ð5Þ
es e

From Eq. 5, LST can be obtained using the inverted Planck function (Eq. 6).
Therefore, the LST obtained is corrected for atmospheric and emissivity effects.
Urban Heat Island Effect of Addis Ababa City … 545

k2
LST ¼   ð6Þ
ln k1
BðLSTÞ þ1

where B(LST) is the blackbody radiance given by the Planck’s law in

W/m^2/sr/um; Latm↑ is the upwelling atmospheric radiance in W/m^2/sr/um (0.79);
Latm↓ is the downwelling atmospheric radiance in (1.32); τ is the total atmospheric
transmissivity between the surface and the sensor (0.89); ε is the land surface
emissivity; LST is the land surface temperature in Kelvin (K); k1 and k2 are the
calibration constants for Landsat TM with k1 = 607.76 W/m^2/sr/um and
k2 = 1260.56 W/m^2/sr/um

3 Results and Discussion

3.1 Land Use and Land Cover Change

Figure 3 depicts the classified maps for 1986 and 2011. The overall accuracy of the
land use/cover maps of 1986 and 2011 were 89.2 and 91.5% respectively. The
Kappa coefficients for the 1986 and 2011 maps were 0.81 and 0.84, respectively.
Applying the methods of Congalton and Green (2009) the above results indicate
strong agreement between the ground truth and the classified classes. Furthermore,
the maps met the minimum accuracy requirements to be used for the subsequent
post-classification operations such as change detection (Anderson et al. 1976).
Table 1 summarizes the producer’s and User’s accuracy figure for the 1986 and
2011 image classification.

Fig. 3 Land use and land cover maps of a 1986, and b 2011
546 E. Teferi and H. Abraha

Table 1 Producer’s and user’s accuracy for individual LULC

Land cover classes 1986 accuracy (%) 2011 accuracy (%)
Producer User Producer User
Urban/built up 83.33 71.43 75.09 76.09
Forest 84.62 73.3 83.33 71.43
Agriculture 94.59 92.11 97.87 92.00
Grassland 67.67 88.89 60.09 85.71
Wetland 78.00 79.00 77.78 87.50
Bare land 75.00 85.71 80.00 80.00

Table 2 Summary of land use/cover transitions (%)

Total Total Gain Loss Total Swap Absolute value
1986 2011 change of net change
Urban 19.20 34.45 18.71 3.46 22.17 6.92 15.25
Forest 11.48 9.63 4.19 6.04 10.23 8.39 1.84
Cropland 40.62 37.51 12.85 15.96 28.81 25.69 3.12
Grassland 23.30 14.95 8.13 16.49 24.61 16.25 8.36
Barren land 5.40 3.46 2.51 4.44 6.96 5.02 1.93
Total 100 100 46.39 46.39 46.39 31.14 15.25

The proportion of gain, loss, swap and net change of each land use/cover for the
1986 and 2011 are presented in Table 2. There has been a considerable change
(*46% of the landscape) in land use/cover in the study area during the 24-years
period. Urban class has shown a growth of about *79%. Urban class experienced
the highest gain in about 19% of the landscape, whereas grassland experienced the
highest loss in about 17% of the landscape, followed by cropland in about 16% of
the landscape. Losses in grassland and cropland are most likely due to urban
expansion. The change attributable to quantity (net change) is highest for urban
(about 15% of total change for urban); whereas the change attributable to location
(swap) is highest for cropland (25% of total change for cropland). Swap land
change dynamics accounted for 31% of total landscape change.
Most of the new developments during this study period, took place in the
suburbs as organized clusters for accommodating especially residential expansions,
industrial, commercial, condominium, emerging settlements, warehouses, or
external transportation facilities, in addition to rapid developments on the outskirts
of the old city core. Table 3 shows the spatial occurrence of urban expansion within
sub-cities. The highest urban expansion occurred in Bole (25.12 km2), Akaki Kality
(21.63 km2), Nifassilk Lafto (16.56 km2) and Kolfe Keranyo (15.64 km2)
sub-cities. Massive urban sprawl in eastern and southern part of Addis Ababa can
be attributable to rural urbanization, which is a common phenomenon in the post-
derg regime. Dramatic urban development was observed in the early 2000s. The
reconstruction, expansion and upgrading of Bole airport to international status;
Urban Heat Island Effect of Addis Ababa City … 547

Table 3 Areal extent of urban expansion by sub-cities

Sub-cities Areal extent of urban Contributions of
expansion (km2) Cropland Grassland Barren land Forest
Addis Ketema 0.39 – – – –
Akaki Kality 21.63 16.54 4.16 0.72 0.2
Arada 0.78 – – – –
Bole 25.12 12.5 9.87 2.13 0.6
Gulele 2.34 – – – –
Kirkos 0.96 – – – –
Kolfe Keranyo 15.64 6.48 5.93 0.82 2.4
Lideta 0.99 – – – –
Nifassilk Lafto 16.56 6.84 7.42 1.3 0.98
Yeka 12.81 4.12 5.63 1.4 1.6

buildings of different factory in industrial area; establishment of private College of

Education, Office buildings, hotels, and recreational facilities played greater role in
shaping and added attribute to a modern Addis Ababa city. Ring road was con-
structed to provide the foundation for future expansion.
In the late 2000, Addis Ababa witnessed yet another significant expansion in
terms of both land area and population. This is evident in areas like Ayat, Lebu,
Summit, Asko, Jemo, and other areas where various stages of land development
like condominiums, single and multistory residential and other types of construc-
tions have been taking place. Furthermore, a number of renewals, upgrading or
slum improvement strategies undertaken such as redevelopment Lideta area, Arat
Killo area, Merkato area, and etc. Moreover, the land use and land cover maps
derived from the Landsat images show that the urban and built-up area increased by
15.25% in the 25-year period (1986–2011). Generally, the city has shown rapid
Vertical and horizontal expansion both planned and unplanned.

3.2 Changes in Land Surface Temperature

in Response to LULC Change

Figure 4 depicts the LST maps for 1986 and 2011. It is evident from the map that
there is a thermal gradient as progressed from the central part to the countryside.
The lowest temperature in the built-up areas (290.16 to 300.95 K) appeared in:
(i) Northern part foot of Entoto hill and areas along Gojam and Ambo road.
(ii) Eastern part around Kotebe (iii) Southeastern parts around Adwa Park and in the
outskirts of city except southern parts along Deberezeit road. The standard devia-
tion of the 2011 LST is greater than that in 1986 (Table 4), indicating that the
surfaces experienced relatively considerable variation in land surface temperature
during these periods. The central parts showed a high temperature of over
548 E. Teferi and H. Abraha

1986 2011

Fig. 4 Derived land surface temperature (LST)

Table 4 Land Surface Temperature (LST) averaged over each land cover classes
Land use and land cover LST 1986 (K) LST 2011 (K) dT (K)
Urban/built-up 303.4 ± 2.88 304.42 ± 3.01 1.02
Forest 296.51 ± 3.3 295 ± 2.99 −1.51
Agriculture 302.93 ± 2.69 302.98 ± 2.92 0.05
Grass land 302.79 ± 2.28 303.81 ± 2.46 1.02
Wet land 301.53 ± 1.45 302.48 ± 1.68 0.95
Bare land 303.3 ± 2.27 304.25 ± 2.64 0.95
dT is the Mean temperature difference between 1986 and 2011

323.23 K, while in the outskirts, a lower temperature of 288.28 K exist. As areas in

cities develop, more vegetation is lost, and more surfaces are paved or covered with
buildings. The change in ground cover results in less shade and moisture to keep
urban areas cool. Thus, built up areas contribute to elevated surface and air tem-
peratures, while green spaces moderate temperatures by providing shade, thus
helping reduce the risk of heat-related illnesses for city dwellers (IPCC 2013; WHO
2013; Wolch et al. 2014).
The distinctive land surface temperature patterns are associated with the thermal
characteristics of land cover classes. To better understand the impact of urban
development on land surface temperatures, the thermal signature of each land cover
type was obtained by overlaying a land surface temperature image with a land use
and land cover map of the same year. The average value of land surface temperature
by land cover types is summarized in Table 4. Of all the LULC categories, urban
and built-up had the highest LST for both years. This implied that urban devel-
opment, due to the replacing of natural vegetation with non-evaporating,
non-transpiring surfaces, such as stone, metal, and concrete, did have an effect on
raising LST. The lowest LST in 1986 was observed in forest, followed by wet land,
grass land, agriculture, and bare land. The average LSTs for urban/built-up area are
303.4 and 304.42 K in 1986 and 2011, respectively. Forest land exhibited a
Urban Heat Island Effect of Addis Ababa City … 549

Table 5 Land surface Land use and land cover dT (s.d.) (K) Adjusted dT (K)
temperature based on land change categories
cover change categories
Unchanged urban 1.54 (2.91) 0.00
Forest to built-up 1.21 (2.54) 3.32
Agriculture to built-up 1.97 (2.99) 1.65
Grass land to built-up 2.42 (2.84) 1.27
Bare land to built-up 1.24 (2.53) 0.29

decreasing trend in LST from 296.51 K in 1986 to 295 K 2011. Bare land pos-
sessed the highest LST next to built-up areas among all the LULC types. In general,
the LST of the city shows an increasing trend in the years under study due to
changes in land cover and land use.
To investigate the effect of landscape transformation on LST change, LST dif-
ference was calculated by subtracting LST in 1986 from LST in 2011 for each land
cover change category (Table 5). LST increased about 3.32, 1.65, 1.27 and 0.29 K
for areas where forests, agricultural lands, grass lands and bare lands converted into
built-up areas, respectively.

3.3 The Urban Heat Island Effect at Addis Ababa

The growth and expansion of Addis Ababa has resulted in change in land surface
temperature. Since 1986, the urban built up area has expanded dramatically.
Figure 5 shows the urban-rural temperature difference between the urban center and
its surrounding areas. Transect #1 (Red line in Fig. 4) shows the maximum LST of
the urban center is 315 K (41.85 °C) and the LST of the surrounding rural areas is
300 K (26.85 °C). Thus, the urban-rural temperature difference between the urban
center and its surrounding areas can reach a maximum of 15 °C. The land surface
temperature in the urban center was a mean of 5 K warmer than that in surrounding
rural areas. The results show that UHI effect existed significantly in the city of
Addis Ababa. The increased temperature is a reflection of urban expansion during
the 25-year period considered. In addition to urban expansion, there have been
significant changes in the old parts of the city. A large area of land in these parts has
been re-developed by the government or private investors for residential, com-
mercial and industrial developments. The conversion of the surrounding forest and
agricultural land into built-up land has contributed to the increased land surface
temperature. The government has relocated many residents to the outskirts of the
city in order to rebuild the city. The new houses and infrastructure were frequently
located in high-quality agricultural land or forestland. This relocation has affected
agricultural area and increased local temperatures.
550 E. Teferi and H. Abraha

Fig. 5 Land surface temperature difference between Addis Ababa urban center and its rural
surrounding

3.4 Urban Greening to Adapt Addis Ababa City

to Climate Change

The relationship between LST and NDVI clearly shows that vegetation has great
impact on reducing UHI effect (Fig. 6). The strong, negative correlation between
LST and NDVI implies that the higher biomass a land cover has, the lower the land
surface temperature. Because of this relationship between LST and NDVI, changes
in land use/cover have an indirect impact on surface temperatures through NDVI.
Thus, it is very crucial to promote and support urban greening, such as new planting
in the public areas and green infrastructure. In many existing urban areas, it is not
feasible to create large new greenspaces. In that case, the creative use of the green
infrastructure such as street tree planting is one of the most promising opportunities
for adaptation.
Urban Heat Island Effect of Addis Ababa City … 551

Fig. 6 The relationship between Land Surface Temperature (LST) and Normalized Vegetation
Index (NDVI)

4 Conclusion

This study has examined LULC changes in the capital of Ethiopia, Addis Ababa
from 1986 to 2011. The results indicate that urban/built-up areas expanded dra-
matically, while agricultural land and forest declined. Barren land increased, mainly
in the boundary areas between forest and dry croplands, especially in steeply
sloping areas. The observed changes in LULC were largely attributed to population
pressure on the land, a rapidly growing infrastructure and poor land use planning.
Changes in LULC were accompanied by changes in LST. Moreover, temperature
differences between the urban/built-up and the surrounding rural areas significantly
widened. The study assessed the UHI spatial patterns and temporal variations in the
Addis Ababa city. The urban-rural temperature differences between the urban core
and its surrounding areas of Addis Ababa show a maximum difference of 15 K.
This indicates the existence of intensified urban heat island effect in Addis Ababa
city. The results of this study suggest that an increase in urban green space could
significantly ameliorate the rising temperatures associated with climate change and
the UHI effect. Thus, development of UHI effect reduction strategies such as the
creation of greenways, community gardens, and street gardens is very important.

References

Anderson, J. R. (1976). A land use and land cover classification system for use with remote sensor
data (Vol. 964). US Government Printing Office.
Amiri, R., Weng, Q., Alimohammadi, A., & Alavipanah, S. K. (2009). Spatial–temporal dynamics
of land surface temperature in relation to fractional vegetation cover and land use/cover in the
Tabriz urban area. Iran Remote Sensing of Environment, 113(12), 2606–2617.
Ball, G. H., & Hall, D. J. (1965). ISODATA, a novel method of data analysis and pattern
classification. Melno Park, CA: Stanford Research Institute.
552 E. Teferi and H. Abraha

Chander, G., Markham, B. L., & Helder, D. L. (2009). Summary of current radiometric calibration
coefficients for landsat MSS, TM, ETM+, and EO-1 ALI sensors. Remote Sensing of
Environment, 113, 893–903.
Chen, D., & Brutsaert, W. (1998). Satellite-sensed distribution and spatial patterns of vegetation
parameters over a tallgrass prairie. Journal of the Atmospheric Sciences, 55(7), 1225–1238.
Chen, X. L., Zhao, H. M., Li, P. X., & Yin, Z. Y. (2006). Remote sensing image-based analysis of
the relationship between urban heat island and land use/cover changes. Remote Sensing of
Environment, 104(2), 133–146. doi:10.1016/j.rse.2005.11.016.
Congalton, R. G., & Green, K. (2009). Assessing the accuracy of remotely sensed data: Principles
and practices. CRC Press.
Foody, G. M., Campbell, N. A., Trodd, N. M., & Wood, T. F. (1992). Derivation and applications
of probabilistic measures of class membership from the maximum-likelihood classification.
Photogrammetric Engineering and Remote Sensing, 58(9), 1335–1341.
Hung, T., Uchihama, D., Ochi, S., & Yasuoka, Y. (2006). Assessment with satellite data of the
urban heat island effects in Asian mega cities. International Journal of Applied Earth
Observation and Geoinformation, 8(1), 34–48.
IPCC, Climate Change. (2013). The physical science basis—working group I contribution to the
fifth assessment report of the intergovernmental panel on climate change. New York:
Cambridge University Press.
Li, Z. L., & Becker, F. (1993). Feasibility of land surface temperature and emissivity determination
from AVHRR data. Remote Sensing of Environment, 43, 67–85.
Mannstein, H. (1987). Surface energy budget, surface temperature and thermal inertia. In Remote
Sensing Applications in Meteorology and Climatology, (pp. 391–410). Springer Netherlands.
Pongrácz, R., & Bartholy, J. (2006). Dezső, Zs. Remotely sensed thermal information applied to
urban climate analysis. Advances in Space Research, 37, 2191–2196.
Robine, J.-M., Cheung, S. L. K., Le Roy, S., Van Oyen, H., Griffiths, C., Michel, J.-P., et al.
(2008). Death toll exceeded 70,000 in Europe during the summer of 2003. Comptes Rendus
Biologies, 331(2), 171–178.
Van de Griend, A., & Owe, M. (1993). On the relationship between thermal emissivity and the
normalized difference vegetation index for natural surfaces. International Journal of Remote
Sensing, 14, 1119–1131.
Weng, Q., & Lo, C. P. (2001). Spatial analysis of urban growth impacts on vegetative greenness
with Landsat TM data. Geocarto International, 16(4), 17–25.
WHO. (2013). Protecting health from climate change: vulnerability and adaptation assessment.
Geneva, Switzerland: World Health Organization.
Wolch, J. R., Byrne, J., & Newell, J. P. (2014). Urban green space, public health, and
environmental justice: The challenge of making cities ‘just green enough’. Landscape and
Urban Planning, 125, 234–244.
Xiao, H., & Weng, Q. (2007). The impact of land use and land cover changes on land surface
temperature in a karst area of China. Journal of Environmental Management, 85(1), 245–257.
Yuan, F., & Bauer, M. E. (2007). Comparison of impervious surface area and normalized
difference vegetation index as indicators of surface urban heat island effects in Landsat
imagery. Remote Sensing of Environment, 106(3), 375–386.
Zha, Y., Gao, J., & Ni, S. (2003). Use of normalized difference built-up index in automatically
mapping urban areas from TM imagery. International Journal of Remote Sensing, 24(3),
583–594.
Zhao, H. M., & Chen, X. L. (2005). Use of normalized difference bareness index in quickly
mapping bare areas from TM/ETM+. Geoscience and Remote Sensing Symposium, 3(25–29),
1666−1668.
Zhou, X., & Wang, Y. C. (2011). Dynamics of land surface temperature in response to
land-use/cover change. Geographical Research, 49(1), 23–36.