- Email: [email protected]
Land Cover in Tehran
Accepted Manuscript Title: Assessment of Urban Heat Island Based on the Relationship between Land Surface Temperature and Land Use/Land Cover in Tehran Author: Mehdi Bokaie Mirmasoud Kheirkhah Zarkesh Peyman Daneshkar Arasteh Ali Hosseini PII: DOI: Reference:
S2210-6707(16)30040-3 http://dx.doi.org/doi:10.1016/j.scs.2016.03.009 SCS 390
To appear in: Received date: Revised date: Accepted date:
29-12-2015 13-3-2016 15-3-2016
Please cite this article as: Bokaie, Mehdi., Zarkesh, Mirmasoud Kheirkhah., Arasteh, Peyman Daneshkar., & Hosseini, Ali., Assessment of Urban Heat Island Based on the Relationship between Land Surface Temperature and Land Use/Land Cover in Tehran.Sustainable Cities and Society http://dx.doi.org/10.1016/j.scs.2016.03.009 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Assessment of Urban Heat Island Based on the Relationship between Land Surface Temperature and Land Use/Land Cover in Tehran Mehdi Bokaie a, Mirmasoud Kheirkhah Zarkesh a, Peyman Daneshkar Arasteh b, Ali Hosseini c * a
Department of Remote Sensing and GIS, IAU Science and Research Branch, Tehran, Iran
b
Department of Engineering and Technology, IKI University, Ghazvin, Iran
c
Department of Geography and Urban Planning, University of Tehran, 1417854151, Tehran Iran * E-mail: [email protected] (Corresponding author): Ali Hosseini
Abstract In this study, the relationship between land surface temperature (LST) and Land Use/Land Cover (LULC) in Tehran Metropolitan City was studied using Landsat TM satellite image. For this, LST was calculated in accordance with the algorithm (Jiménez‐Muñoz & Sobrino, 2003) and the LULC map was prepared based on supervised classification method. According to the LST map obtained by processing the thermal band of the satellite image, areas affected by Urban Heat Island (UHI) were detected to examine their status in relation to the existing LULC classes and the population density. The results showed that UHI created in Tehran are different in terms of the causative agent. This difference is primarily due to the status of LULC in the region and reflects the close relationship between land cover and land surface temperature. Also, the distribution of vegetation and green spaces in different areas of Tehran City was studied by normalized vegetation index (NDVI) using remotely-sensed data. Correlation study between the land cover and land surface temperature showed a negative correlation between these two parameters. Study of the average surface temperature in six LULC classes indicated full compliance of heat islands with LULC classes. Keywords: Land Surface Temperature (LST), Urban Heat Island (UHI), Normalized Difference Vegetation Index (NDVI), Land Use/ Cover (LULC), Change with Land Use/ Land Cover (LULC).
1.
Introduction
The century 21, during which about half of the global population have been urban dwellers, is called “urban century”. According to forecasts, this amount will reach even more than 70% by the year 2050. Cities are places with sever consumption of various sources and increasing the release of contaminants. They are also a serious threat to ecological resources and can restrict urban facilities and the rapid growth of urbanization (Hosseini et al. 2015; Mirkatouli et al, 2015). The rapid growth of urbanization and the development of areas covered with man-made land cover versus the lower the level of natural cover is one of the main causes of global climate change (Streutker, 2002; Sekizawa et al., 2015). Population growth in urban areas, especially in developing countries, puts a lot of pressure on natural resources and causes gradual loss of these resources (Amiri, 2009; Feizizadeh & Blaschke, 2013; Peron et al. 2015). In recent decades, UHI, which are known as areas with negative temperature gradient compared to the surroundings, is one of the most important factors affecting the quality of human life (Wong et al., 2013; Mirzaei, 2015). UHI phenomenon is a risk arising from uncontrolled growth of urban areas (Mirzaei et al., 2012). Even if the global climate is not getting warmer, cities are now faced with the problem of rising temperature (Hoverter, 2002; Zinzi & Agnoli , 2012; Giannopoulou et al., 2011; Dimoudi et al., 2014; Taha, 2015). Cities with different physical surfaces than the surrounding environments (rural areas) influence on their microclimate (Cai et al., 2011). The behavior of urban surfaces is far different from the behavioral pattern of natural surfaces in terms of the absorption of shortwave and long wave radiation, evaporation, release of man-made heat, and the block of prevailing wind (Mirzaei & Haghighat, 2010). The difference in physical material of urban surfaces is very high. Man-made cover and surfaces such as asphalt, flooring, and concrete, instead of natural pervious surfaces such as soil and vegetation, reduce evapotranspiration and increase sensible heat in cities. As a result, cities experience a warmer weather than surrounding areas. Furthermore, accelerated deformation and design of buildings and the
emergence and spread of skyscrapers as well as increased industrial activities have led to the development of UHIs (Liu & Zhang, 2011; Busato et al., 2014; Radhi et al, 2015). Nowadays, raw materials are designed and built in such a way that have a high potential in absorbing solar radiations, high impermeability, and the desired thermal specifications for energy storage and release of heat. Knowing these thermal characteristics can lead to develop appropriate strategies for mitigation of the impacts of UHIs (Gagliano et al, 2015). Megacities with compact constructions and a variety of activities, especially in the CBD, fully affect on their climate and the climate of surroundings (Boehme et al., 2015; Morris et al., 2015). Natural cooling process in urban environments undergoes problems due to the extensive changes in the pattern of urban design and structure. A complex network of high-rise buildings and narrow streets, that makes trapping the heat absorbed during the day, impairs air circulation and consequently the process of reducing the temperature in urban areas. Calculating the amount of LST is considered a starting point for the analysis of UHI (Mather, 1986; Malekpour et al., 2011). One of the strategies adopted to reduce the heating and cooling load of buildings in urban areas is the increased use of thermal insulation (such as roof coatings). Despite the positive role of thermal insulation in buildings to prevent the loss of heat energy in the winter, they increase indoor temperature of buildings in summer, which results in increased energy use for cooling purposes (Gagliano et al, 2015). In studies of urban climate change, the LST has been recognized as one of the most important parameters affecting the UHI phenomenon (Liu & Zhang, 2011). Further, in many of the studies on the analysis of UHI, a very close relationship between LST and LULC as well as considerable impressibility of UHI by this relationship have been emphasized (Weng et al., 2004). Expansion of synthetic surfaces is a marker for centralized human activities that lead to an increase in Earth's surface temperature (Su et al., 2010).
The exact calculation of the LST and the study of relationship between current LULC and LST can be an important factor in order to solve many problems in the field of climate change in urban areas and interaction between humans and the environment (Ramachandra & Uttam, 2009). Developing remote sensing technology, launching various satellites, and obtaining images with high spatial resolution power at large spectral range as well as providing imaging capabilities at reasonable intervals with repetitive coverage of the Earth's land surface have made possible monitoring and investigating of the land surface (Mallick et al., 2008). So far, a large number of satellites and sensors equipped with thermal infrared bands have been used for the study and monitoring of UHI. In the absence of regular and dense network of ground weather stations, temporal and spatial distribution of LST based on thermal imaging can be used to supply the required input data for the study of UHI (Feizizadeh et al., 2013). Images captured by Landsat TM sensor are one of the most widely used satellite images in the field of environmental researches that uses the data of the band 6 within the range between 10.4 and 12.5 microns to study and analyze the ground surface phenomena such as UHI (Ramachandra et al., 2012). According to the foregoing, the present study was carried out to follow three objectives: 1. The study of LST in Tehran City using Landsat TM satellite images, 2. Analysis of the spatial distribution of LST and its relationship with LULC and Normalized Difference Vegetation Index (NDVI), and 3. Analysis of relationship between LST and the population density as well as its impact on energy consumption and the health of residents in the study area. To calculate LST, one of the widely used remote sensing methods that is based on thermal infrared wavelength band was used (Jiménez‐Muñoz & Sobrino, 2003). The LULC map of the study area was prepared using Landsat TM image classified by the maximum likelihood algorithm. The population distribution was also studied in the form of a map prepared from the latest census in 2011.
2.
Materials and methods
2.1 Study area The present study was conducted in Tehran, the capital city of Iran, which is located in the northern part of the country at the southern hillside of Alborz Mountains within the longitudes of 51° to 51° 40′ E and the latitudes of 35° 30′ to 35° 51′ N (Hosseini et al. 2015). According to the latest Iranian Population and Housing Census in 2011 by the Statistical Centre of Iran, Tehran, with a population of 8,154,051 people, is still ranked first most populous city in Iran with a very distinct demographic difference than other cities (Statistical Center of Iran, 2011) (Figure 1). The study area leads to mountainous areas from the north and to desert areas for the south, so the southern and the northern parts have a different climate. The northern areas have a cold and dry climate, while the southern parts suffer from warm and dry climatic conditions. The altitude of the city varies from 900 m to 1800 m. The great variation in the altitude is due to the large extent of the city. In the Tehran metropolitan, the annual mean temperature changes between 15 to 18 Celsius degrees and given the height differences in the city, the various parts have an average of 3 Celsius degrees difference in temperature. The aviation of monthly mean relative humidity include minimum and maximum relative humidity registered at Mehrabad Station shows that in mornings, humidity changes from a minimum of 38% to a maximum of 79% in July to January. The changes in the relative humidity in midday vary between 15% in June to 47% in February. The annual precipitation in Tehran is mainly affected by the height differences, varying between a maximum of 422 millimeters in the north and a minimum of 145 millimeters in the southeast. The number of rainy days follows the same pattern and varies between 89 days in the north and 33 days in the south. Also in the city area, there are about 205 to 213 days of clear sky with some clouds per year (Habibi & Hourcade, 2005).
Figure 1. Situation of the study area, Tehran City
2.2. Satellite images and meteorological data In order to estimate the LST and to investigate the areas affected by UHI, Landsat TM satellite image dated back to 7 August 2010 was used. To prepare the LULC map, discrete bands of the image in the visible, near-infrared, and mid- infrared spectral range were stacked by layer stack command. Further, the data of air temperature measured in five synoptic weather stations in Tehran, including Chitgar, Dowshantape, Mehrabad, Aqdasieh, and Geophysics, were used as one of the model inputs to estimate LST. These data have been measured within the 3 hour-intervals from 12 a.m. to 9 p.m. GMT (Greenwich Mean Time). So, there are 8 measured temperature data available for each day. The position of the synoptic weather stations is depicted in Figure 2.
Figure 2. Position of the synoptic weather stations in Tehran City
2.3. Pre-processing of Landsat Image Many raw satellite images, due to atmospheric conditions and possible mechanical problems in the sensors, have some distortions and errors that must be corrected using different methods prior to processing them (Yan et al., 2006). Atmospheric and radiometric corrections are considered among the most important pre-process steps in the processing of satellite images for the purposes of image classification and preparation of LULC map, as well as estimation of LST (Song et al., 2001; Li et al., 2004). For radiometric corrections, the Digital Numbers (DNs) of all pixels in the image were converted to spectral radiance using the data in the image header file and based on the Equation 1. Then, to estimate the LST, the brightness temperature was calculated using Equation 2 (Chander et al., 2009). In addition, in order to minimize the effects of atmospheric conditions on the image, ENVI software was used to do atmospheric corrections. Dark Object Subtraction method was used to correct satellite
images atmospherically. This method is available as a pre-processing step for several of the in-
scene based atmospheric correction methods. Dark object subtraction method searches each bands of multi spectral image for the darkest pixel value. Assuming that dark objects reflect no light, any value greater than zero must result from atmospheric scattering. ENVI Classic removes the scattering by subtracting this value from every pixel in the band. This simple technique is effective for haze correction and normally produces improved results for in-scene atmospheric correction in the shorter wavelengths where scattering dominates (Song et al., 2001). 𝐿(𝜆) = (
𝐿𝑀𝑎𝑥 −𝐿𝑀𝑖𝑛 255
) ∗ 𝐷𝑁 + 𝐿𝑀𝑖𝑛 (1)
Where;
DN is the degree of greyness of the pixels and L (λ) (radiance) is in (W/m2/sr/mm). LMax and LMin are the calibration constants of the sensor, equal to the maximum and minimum values of the spectral radiance (in W/m2/sr/ mm) detectable for each band, by the sensor Landsat TM. LMax values for band 3 and band 4 Landsat TM are 264.00 and 221.00, and LMin values for band 3 and band 4 are -1.170 and -1.510 (Chander et al., 2009) T=
𝐾2 𝐾1 𝑙𝑛( +1) 𝐿ℷ
(2)
In which; T= brightness temperature, 𝐾1 and 𝐾2 are correction constants of 607.76 and 1260.56, and Lλ = spectral radiance of sensor in Wm-2sr-1μm-1. Despite the primary geometric corrections on the Landsat products, in order to eliminate possible errors in the image of the study area, it was geometrically rectified using national topographical maps at a scale of 1: 25000. The rectification process was performed by first-order Polynomial model using nearest neighbor re-sampling method. RMSE was calculated using the equation (3) to check the accuracy of
geometric correction. This parameter calculates the difference between observed values and predicted values (Morad et al., 1996) 1
𝑅𝑀𝑆𝐸 = √ ∑𝑁 ̂𝑖 )2 (3) 𝑖=1( 𝑥𝑖 − 𝑥 𝑁
The RMS error of the image was estimated at 0.03.
2.4. Satellite Image Processing 2.4.1. NDVI NDVI is considered as one of the most widely- and commonly- used indices in environmental studies (Sun & Kafatos, 2007; Sekizawa et al., 2015). If the initial processing and corrections are done well on the image, the values of this index will be assured confidently (LI et al., 2009). This index is the difference between the maximum absorption in the red region of chlorophyll pigments and a maximum reflectivity in the near infrared region of the cell structure leaves (Baret & Guyot, 1991). And is calculated with Red and Near Infrared band in multi spectral satellites using equation”4”
𝑁𝐷𝑉𝐼 =
( 𝑁𝐼𝑅−𝑅𝐸𝐷 ) ( 𝑁𝐼𝑅+𝑅𝐸𝐷 )
(4)
In theory the value of this index is variable in range of (+1 and -1). The value of NDVI for dense vegetation usually differs from 0.3 to 0.6 and negative values indicate the clouds, snow covers and water (Serrano et al., 2000) There have been many studies on the relationship between LST and vegetation cover by analyzing the values of NDVI and surface temperature in which the key role of NDVI and its variations on LST map at different LULC classes are emphasized (Weng et al., 2004; Yue et al., 2007; Raynolds et al., 2008). The present study also examined correlation between changes in LST and vegetation cover measured by NDVI.
2.4.2. Emissivity Total radiation measured from a surface can be reduced or increased as by changes in the emissivity (Prata, 1993). Therefore, when the temperature of a land surface is calculated using satellite images, it would be quite important to estimate accurately the extent of emissivity from the given surface whereas incorrect estimate of emissivity or its ignorance will raise the error in the calculated temperature up to 1 or 2 degrees Kelvin (Caselles et al., 1995). Emissivity of various surfaces is a function of different parameters such as properties, water content, chemical composition, physical structure, and texture of surfaces (Snyder et al., 1998). Further, emissivity of different surfaces varies in different wavelengths. The rate of variations is less in wavelengths between 10.5 μm and 12.5 μm, which is the spectral range of the thermal band in TM sensor (Weng, 2009; Stathopoulou, 2007). In areas with dense land cover, assuming a constant coefficient for the emissivity would be a realistic assumption. However, in poorly vegetated areas or bare lands with various types of soil and minerals, it is necessary to consider the emission factor of each of the pixels in measurement of temperature. In the absence of required facilities for laboratory tests, satellite images are used to calculate emissivity of different materials and ground features at different wavelengths. The NDVI is considered as one of the most common methods of this type (Valor & Caselles, 1996). In the present study, in order to estimate the emissivity, the NDVI was used based on the method presented by Van de Griend (1993).The average emissivity coefficients for barren land, green space, forest parks, water bodies, asphalt surfaces, and residential areas were calculated as 0.932, 0.937, 0.937, 0.931, 0.933, and 0.932, respectively.
2.4.3. Land Surface Temperature retrieval from Landsat TM image After calculation of the image brightness temperature, using the method developed for the sensors with only one thermal bond, the LST was computed based on the Equations 3 to 8 (Jiménez‐Muñoz & Sobrino, 2003).
Ts = γ [ε−1 (φ1 Ls + φ2 ) + φ3 ] + δ γ= {
C2 Ls λ4 T2b
[
C1
Ls + λ−1 ]}
−1
(5)
(6)
δ = −γ Ls + Ts (7) In which, Ts= land surface temperature, λ=effective wavelength, which was regarded to be 11.5 microns in this research, ε= surface emissivity, C1 and C2 =constant values of 1.19104×108 μm 4 m -2 sr -1 and 14387.7 μm K, respectively, and φ1, φ2, φ3 = atmospheric functions that are calculated using the Equations 8 to 10 and the amount of water vapor in the atmosphere at the time of imaging. The average water vapor in the study area was estimated at 0.748. φ1=0.14714 W2 – 0.15583 W+1.1234
(8)
φ2= −1.1836 w2 −0.37607w−0.52894
(9)
φ3= −0.04554 w2 + 1.8719 w−0.39071
(10)
2.4.4. Land cover classification Satellite images are widely used to monitor LULC changes and to prepare new LULC maps (Lobo et al., 2004). LULC refers to the data obtained from satellite images that are classified based upon the values of different pixels in an image. These data are generally in the format of raster or grid. Each pixel has a value that links it to a particular class in the classification process. In this study, Landsat TM image was used to investigate relationship between LST and LULC changes in the study area as well as their influence on the spatial distribution of UHI. For this, various types of LULC were classified in to six classes including bare lands, constructed areas, green spaces, forest park, water bodies, and asphalt-paved surfaces using supervised classification method and maximum likelihood algorithm. Six classes of LULC were defined based on our knowledge of case study (the Tehran city) which it had been provide from
high resolution satellite imagery and field visits. Ensuring the accuracy of the classification, SPOT 2.5 meter satellite image dated of 03 March 2010 was used. Validation is considered one of the most important parts of image classification process because it examines the degree of success of an algorithm in classification of different objects in an image. The accuracy of a classification process is usually assessed by comparing the results of classification with reference data from field visits, high spatial resolution images, or aerial photographs (Rosenfield & Fitzpatrick-Lins, 1986). Results of accuracy assessment are usually presented in the form of a matrix error
(confusion matrix). Overall classification accuracy is computed by dividing the number of pixels categorized in the same class (based on either field visits or satellite images) on the number of all pixels (Bogoliubova,& Tymków, 2014).
𝑇=
∑ 𝐷𝑖𝑖 𝑁
(11)
Where; Dii∑ is the number of pixels that are classified correctly and N represents all the pixels in matrix error Unfortunately, the overall accuracy parameter cannot provide classification accuracy of each land use class, separately. Therefore, user accuracy and producer accuracy parameters were used to check the classification accuracy of each of the classes. Producer accuracy examines the probability that a particular class of land cover on the ground is classified under the same category in the satellite image while user accuracy investigates the probability of correct labeling of a particular land use class assigned to a pixel. For example, in this study, the producer accuracy for the land use class of green lands was calculated as (322⁄330) = 97.57% and the user accuracy was computed as(322⁄323) = 99.69%. As one of the
indices of classification accuracy, Kappa coefficient was applied to verify the accuracy of the classified pixels compared to ground truth (Congalton, 1991) by the use of equation 12.
𝐾=
𝑁 ∑𝑟𝑖=1 𝑥𝑖𝑖 − ∑𝑟𝑖=1(𝑥𝑖+ ∗𝑥+𝑖 ) 𝑁2 − ∑𝑟𝑖=1(𝑥𝑖+ ∗𝑥+𝑖 )
(12)
In which; r =number of rows in the error matrix, xii= number of observations in the ith row and column, x+i= total number of observations in the ith column, and N= total number of observations. The overall accuracy and Kappa coefficient of the classification were 93.41% and 0.9166, respectively. The results of supervised classification error matrix are given in Table 1. Besides, Figure 4 presents the LULC map prepared from the satellite image are illustrated in.
Table 1. Error Matrix of the Landsat TM Image in 2010
3.
Results
3.1. Relationship between LST and LULC Figure 3 demonstrates the LST map prepared from Landsat TM satellite image. The temperature in this map were classified based on standard deviation of values and subsequently, areas affected by UHI were identified. According to the LST map, surface temperature varied between 21.5 ◦C and 57.9 ◦C. At the time of imaging, the lowest average temperature was 35 ◦C related to the water bodies and the maximum average temperature was 48 ◦C recorded from bare lands in the study area. To clarify the relationship between the LST and land cover, particularly vegetation cover whose distribution status in the study was examined with NDVI, variations in the temperature of different LULC classes were studied. Comparing the estimated temperature values and the LULC map revealed the thermal variation in
different LULC classes. According to which, the lowest temperature corresponds to the water bodies and areas covered by natural coatings such as green space, forest parks, and urban parks, while maximum temperature was recorded in bare lands and impervious surfaces such as asphalt- paved areas and other man-made coatings, as well as industrial, commercial, residential, and transport land uses. Generally speaking, areas largely influenced by UHI are mainly located in 4 different places in the study area including various land uses. One of the most important affected areas is District 21 of the Tehran City housed the vast majority of factories, industrial workshops, warehouses, and other related land uses, as a chain of industrial activities. In other words, majority of the district is occupied by industrial land use and man-made coatings. The second area that is very largely influenced by the UHI is District 9. This is mainly due to Mehrabad International Airport, passenger transport terminals, and the gate of Tehran from the west as one of the oldest areas affected by this phenomenon. Heavy road and air traffic, and large volume of public and private vehicles, as well as a variety of pollutants caused by them are the most important negative aspects of this district. The third important UHI-affected area is located in the north and northwest of the District 22 and the north of the District 19 due to the bare lands spread in the region. The causative agent of this phenomenon in these districts is different from the rest areas.
Figure 3. Classified map of land surface temperature, July 2010 Figure 4. LULC map prepared from Landsat TM satellite image
In addition to these areas, it can be noted to the District 12 (Central Market and the old core of Tehran City), District 13 (bare lands around the former Doshan-Tapeh Airport and its surrounding asphalt-paved areas) and the southern areas of Tehran where there are workshops and industrial activities. However, the spatial development of HUI is less severe in these areas. In some parts, such as District 12,
enforcement of new laws and regulations such as amendments to the vehicle and traffic laws and development of pedestrian-only streets is among the steps taken in recent years to reduce the impact and progress of this phenomenon. In contrast to the above-mentioned areas largely affected by UHI phenomenon experienced changes in the temperature and energy consumption, there are areas with natural covering such as forest parks and green spaces marked in green in Figure 3. Moreover, water bodies and residential areas with low to relatively medium population density are less-influenced from the UHI compared to the largely-affected areas. These areas are displayed in the figure among areas with low to medium temperature.
3.2. Land Surface Temperature and NDVI Increase the area of vegetation cover and green spaces in urban environments has a significant influence on the thermal condition of cities and development of UHI. Vegetation cover through evapotranspiration process can reduce and adjust the temperature of land surface (Yuan & Bauer, 2007). That is why in many of the studies conducted in order to find solutions to deal with the phenomenon and mitigate its effects; the role of vegetation cover is mentioned as one of the most important ways to reduce the effects of UHI (Rizwan et al., 2008; Corburn, 2009). Although the use of vegetation cover, especially trees, is a perfect solution to reduce emissions and mitigate the impacts of UHI phenomenon, but planting trees and development of green spaces is not easy in dense urban areas (Gaffin et al., 2008). The use of green roofs and development of green surfaces on the upper level of buildings can be considered as one of the ways to reduce the temperature of cities and improve air quality (Yang et al., 2008). Evapotranspiration process can decrease the air temperature at rates between 1 and 5 ° C (Farina, 2001). Generally, areas with suitable vegetation cover and the high NDVI shall experience much lower surface temperature than the areas with poor vegetation cover. On the contrary, the LST is higher in
areas with less vegetation cover. However, the negative correlation between NDVI and LST may be affected by factors such as soil moisture and evapotranspiration status, however, the role of vegetation cover and its impact on the reduction of air and surface temperature is often undeniable (Su et al., 2010). LST maps obtained using remote sensing technology and satellite imagery, in addition to displaying temperature at different LULC classes; reveal spatial distribution of UHI in different areas (Figure 3). The spatial pattern of LST and NDVI in the study area is displayed in Figure 5. At the time of the study, the NDVI value in Tehran varied between 0.15 and 0.34. Areas with the lowest NDVI value are mainly spread in northwest, south, and southwest of the city. According to the LULC map, these areas are mainly covered by bare lands and industrial land uses. Areas with high NDVI are located in the north, northeast, and parts of southeast and west of the study area, which in most cases correspond to the urban parks, forest parks, and parts of agricultural lands (especially in the southeast), respectively. The surface temperature is always influenced by water content and vegetation cover. In other words, there is a direct linear relationship between temperature, and material and coating of the surface (Yue et al., 2007; Raynolds et al., 2008). In this study, a total number of 500 points were selected randomly to study the relationship between LST and NDVI in the form of a two-dimensional scatter plot (Figure 6).
As the figure suggests, LST is negatively correlated with the NDVI so that LST increases by reducing the amount of vegetation cover. It makes more evident the role of vegetation in decreasing temperature. As shown in the graph, the lowest recorded LST (302-304°K) is observed in the areas with the highest levels of vegetation cover. As the graph goes towards the areas with less vegetation cover, the surface temperature is increased. The largest proportion of data is concentrated between the temperature range of 307-310 °K and the NDVI values of 0.17 to 0.2. The downtrend continues up to the end. The fitted line is also an affirmation of the same negative relationship between the data. The coefficient of determination (R2=0.6) of the regression model also showed a good fit to the data. In addition, in order to verify the accuracy of the regression model, the RMS error was computed for the LST and NDVI data.
The error value of 0.0196 can be due to the precision of the developed model and the accuracy of the relationship between LST and NDVI parameters in the study area. According to the figure 6, there is a negative correlation between the LST and NDVI so that the surface temperature of land will increase by reducing the amount of vegetation cover.
Figure 5. NDVI map of Tehran City Figure 6. Relationship between land surface temperature (LST) and vegetation index (NDVI)
3.3. Land Surface Temperature and Population Spatial pattern of heat islands in the study area reflects the different nature of their constituent and aggravating factors. The formation of this phenomenon in proximity of dense residential and commercial areas, together with traffic-induced air pollution, and consequently release of dangerous environmental pollutants are the most important factors threatening the health of urban residents and increasing the risk of many diseases, particularly respiratory and cardiovascular diseases. With the expansion of impervious and man-made surfaces due to the loss of natural land cover, particularly vegetation cover, fundamental problems have emerged in the process of self-purification of pollutants in urban areas (Lo et al., 2003). Heat capacity has increased recently in urban areas due to various constructed structures and the expanded use of unnatural materials such as asphalt, cement, concrete, etc. on the body of these structures. This increase has led to a greater absorption of solar energy, and thus increased temperature of urban areas (Priyadarsini, 2009). High density of population in an area and consequently the expansion of settlements and high-rise residential buildings, on one hand, can cause the energy to be trapped in walls due to high heat capacity, and on the other hand, lead to excessive use of artificial cooling systems by residents and anthropogenic heat in the city.
Based on the population of different districts of Tehran in 2010, the 21, 22, 9, and 12 districts that host the most important UHI cores in the city are among the most sparsely populated areas of Tehran (Figure 7 and Figure 8). Due to the status of different LULC classes, the low population density in these districts seems normal whereas the UHIs in these districts are mainly induced by industrial, airport and transport, and commercial land uses, as well as the barren lands. The LST map and population distribution in the different districts of Tehran reveal areas with a high temperature in the 14, 5, 2, 15, 4 and 20 districts where face a high density of population. Although these areas are not considered among the main cores of UHI in Tehran, but high population density, high traffic volume, and pollution caused by transport, put these areas at the risk of emergence and spread of UHIs.
Figure 7. Spatial distribution of the population in Tehran, 2011 Figure 8. Population of districts of Tehran, 2011
4.
Discussion
Although there is usually observed an island core in single-core cities, however, different urban planning patterns can cause the formation of different forms of heat islands. Considering that areas with dense buildings are generally warmer than the surroundings, multi-core population centers in cities can lead to creation of several heat islands or can exhibit the heat island as a multi-core UHI. As Figure 3 suggests, the study area is faced with different types of heat islands in terms of the causative agent. The only cause common to all them was very high temperature of these lands compared to their surroundings. Examination and comparison of LST and LULC maps as well as field visits showed a significant difference in the nature of areas affected by UHI. The most important areas that are known as heat cores and UHIaffected area in Tehran are adjacent areas to industrial estates and factories (District 21), dense residential and commercial structures with intricate texture of old and new (Districts 12, 16, and 20) or
the areas around the airport and major transport terminals (Districts 9, 18, and 19). The LST and LULC maps represent establishment of some UHI cores in areas with the land use class of barren land, particularly in the north-west of the study area. According to the 2011 Census by the Statistical Center of Iran, Tehran, including an overall population of 8,154,051 people, is ranked first most populous city in Iran with a serious demographic difference compared to other cities. As similar to the previous decades, the city is still the first choice of many immigrants due to the expansion and centralization of facilities, impressively high income rates, as well as centralization of job opportunities and other municipal services. District 22 in the North West of Tehran hosts one of the most important botanical parks and forest parks in the study area. The district has always attracted the attention of city managers as one of the bets weather areas of Tehran, so that following the rapid population growth of the city due to irregular migration, the district was proposed as one of the possible options for new constructions and city expansion in order to provide housing for the new residents. Accordingly, many residential townships have been established or are on the verge of completion. According to the LST and LULC maps as well as analysis the status of vegetation cover by NDVI map indicate the point that UHI-affected areas in this district are mainly barren lands. Unlike the discussed thermal cores in the vicinity of industrial areas, transportation terminals, and dense residential and commercial areas, the emergence and development of heat islands in the District 22 and the North West of Tehran happens under the influence of different changing trend of soil temperature and barren lands compared to the surroundings. This could reflect the fact that the absence of particular natural or even particular coatings in an area can help the emergence and development of UHIs. Increasing development of construction processes in the District 22 in northwest of Tehran, and the use of a variety of construction materials such as cement, concrete, and clay bricks, the severity and extent of UHI-affected areas will be added.
High levels of NDVI in Tehran fully match the coatings of green spaces (parks, gardens, and grass coverings) and forest parks. The negative relationship between the status of vegetation cover and LST changes in the study area also emphasizes the expansion of vegetation cover as a way to deal with the emergence and escalation of UHIs. Comparing maps of vegetation cover, population density, and LST, it can be argued that some areas, such as 13, 10, 11, 7, and 8 Districts, due to high density of residential and commercial areas and the lack of sufficient vegetation cover, are in danger of becoming the new HUI cores in the future. The effect of changes in LULC on LST changes is also evident in the District 21, which has been converted to a long chain of industrial, workshop, and warehousing land uses. Considering the status of LULC and also the poor condition of vegetation cover in this district, the existence of UHIs in this area does not seem so strange. However, factors affecting the emergence and development of UHIs in this industrial district are different from other thermal cores of the city. Unlike the Districts 19 and 21 where the emergence of UHIs is induced by the existence of barren lands or some central parts of Tehran where the emergence of UHIs is attributed to the high density construction and abnormally high levels of artificial construction materials that mismatches with the natural coatings, UHIs in this industrial district have different origins. In other words, construction of industrial estates, factories, and small workshops, as well as conversion of a wide proportion of natural vegetation cover provides a ground for the emergence and spread of UHIs in this district. Nowadays, emissions from these industrial land uses are a contributing factor affecting the temperature difference between the industrial chain and the surrounding environment as well as durability of UHIs.
5.
Conclusion
Emergence and spread of UHIs occur under the influence of various factors. Depending on the circumstances and factors under which the UHIs have been initiated and expanded, there can be offered some strategies to mitigate their dimensions and eventually eliminate them. Analysis of the relationship
between the current LULC and LST as well as the factors influencing the growth and development of heat islands in Tehran are the main goals of this study. According to which, the study investigated the relationship between the LST and spatial pattern of land cover types, as well as the impact of population density on the development of heat islands in Tehran, the capital of Iran. Landsat TM image was used to prepare the surface temperature map of the study area and to analyze UHIs in the city. Before performing the desired processing to produce LST, NDVI, and LULC maps, the satellite images were geometrically corrected using the topographic maps at the scale of 1:25000. The radiometric and atmospheric corrections were also done by the use of ENVI software. Based on the results, several factors contribute to the development of UHI in the region. Industrial factories, airports, bare lands, and dense residential areas are the land uses more affected by UHI. Moreover, population density in different parts of the study area reflects the impact of human activities on the energy consumption and intensification of heat islands. Proximity of dense residential areas to the UHI-affected zones increases the risk of respiratory, and heart diseases among the residents and makes harder the conditions of everyday life, especially for the elderly and children. The LST in the six land use classes approved the full compliance of the two LST and LULC maps. So that areas with natural cover, particularly with vegetation cover and green spaces, with an average temperature of 40 ° C are the coolest places in the city, while bare lands with an average temperature of 48 ◦C, and asphalt-paved surfaces, with an average temperature of 46 ◦C are the hottest places. This shows the influence of natural cover such as water bodies and green spaces on mitigation of the intensity and spread of UHI. The obtained results also confirmed an adverse correlation between the spread of land cover and green spaces, measured by NDVI, and LST estimated by satellite images. As expected, with an increase in the NDVI, the LST was declined and in areas covered with green spaces, a lower average LST was observed.
References 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. Baret, F., & Guyot, G. (1991). Potentials and limits of vegetation indices for LAI and APAR assessment. Remote sensing of environment, 35(2), 161-173. Boehme, P., Berger, M., & Massier, T. (2015). Estimating the building based energy consumption as an anthropogenic contribution to Urban Heat Islands.Sustainable Cities and Society, 373- 384. Bogoliubova, A., & Tymków, P. (2014). Accuracy assessment of automatic image processing for land cover classification of St. Petersburg protected area. Acta Scientiarum Polonorum. Geodesia et Descriptio Terrarum, 13(1-2). Busato, F., Lazzarin, R. M., & Noro, M. (2014). Three years of study of the Urban Heat Island in Padua: Experimental results. Sustainable Cities and Society, 10, 251-258. Cai, G., Du, M., & Xue, Y. (2011). Monitoring of urban heat island effect in Beijing combining ASTER and TM data. International journal of remote sensing, 32(5), 1213-1232. Caselles, V., Coll, C., Valor, E., & Rubio, E. (1995). Mapping land surface emissivity using AVHRR data application to La Mancha, Spain. Remote Sensing Reviews, 12(3-4), 311-333. 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(5), 893-903. Congalton, R. G. (1991). A review of assessing the accuracy of classifications of remotely sensed data. Remote sensing of environment, 37(1), 35-46. Corburn, J. (2009). Cities, climate change and urban heat island mitigation: Localising global environmental science. Urban Studies, 46(2), 413-427.
Dimoudi, A., Zoras, S., Kantzioura, A., Stogiannou, X., Kosmopoulos, P., & Pallas, C. (2014). Use of cool materials and other bioclimatic interventions in outdoor places in order to mitigate the urban heat island in a medium size city in Greece. Sustainable Cities and Society, 13, 89-96. Farina, A. (2011). Exploring the relationship between land surface temperature and vegetation abundance for urban heat island mitigation in Seville, Spain. LUMA-GIS Thesis. Feizizadeh, B., & Blaschke, T. (2013). Land suitability analysis for Tabriz County, Iran: a multi-criteria evaluation approach using GIS. Journal of Environmental Planning and Management, 56(1), 1-23. Feizizadeh, B., Blaschke, T., Nazmfar, H., Akbari, E., & Kohbanani, H. R. (2013). Monitoring land surface temperature relationship to land use/land cover from satellite imagery in Maraqeh County, Iran. Journal of Environmental Planning and Management, 56(9), 1290-1315. Gaffin, S. R., Rosenzweig, C., Khanbilvardi, R., Parshall, L., Mahani, S., Glickman, H., Goldberg. R., Blake, R., Slosberg, R. B. & Hillel, D. (2008). Variations in New York City’s urban heat island strength over time and space. Theoretical and applied climatology, 94(1-2), 1-11. Gagliano, A., Detommaso, M., Nocera, F., & Evola, G. (2015). A multi-criteria methodology for comparing the energy and environmental behavior of cool, green and traditional roofs. Building and Environment, 90, 71-81. Giannopoulou, K., Livada, I., Santamouris, M., Saliari, M., Assimakopoulos, M., & Caouris, Y. G. (2011). On the characteristics of the summer urban heat island in Athens, Greece. Sustainable Cities and Society, 1(1), 16-28. Habibi, M., Hourcade, B. (2005). Atlas of Tehran Metropolis. Tehran Geographic Information Center and CNRS, Vol. 1. - Land and People Hosseini, A., Pourahmad, A., & Pajoohan, M. (2015). Assessment of Institutions in Sustainable UrbanManagement Effects on Sustainable Development of Tehran: Learning from a Developing Country. Journal of Urban Planning and Development, 05015009.
Hoverter, S. P. (2012). "Adapting to urban heat: a tool kit for local governments. “Georgetown Climate Center. Jiménez‐Muñoz, J. C., & Sobrino, J. A. (2003). A generalized single‐channel method for retrieving land surface temperature from remote sensing data.Journal of Geophysical Research: Atmospheres (1984–2012), 108(D22). Li, F., Jackson, T. J., Kustas, W. P., Schmugge, T. J., French, A. N., Cosh, M. H., & Bindlish, R. (2004). Deriving land surface temperature from Landsat 5 and 7 during SMEX02/SMACEX. Remote sensing of environment, 92(4), 521-534. Li, H., Liu, Q. H., & Zou, J. (2009). Relationships of LST to NDBI and NDVI in Changsha-Zhuzhou-Xiangtan area based on MODIS data. Scientia Geographica Sinica, 2, 018. Liu, L., & Zhang, Y. (2011). Urban heat island analysis using the Landsat TM data and ASTER data: A case study in Hong Kong. Remote Sensing, 3(7), 1535-1552. Lo, C. P., & Lo, C. P., & Quattrochi, D. A. (2003). Land-Use and Land-Cover Change, Urban Heat Island Phenomenon, and Health Implications. Photogrammetric Engineering & Remote Sensing, 69(9), 1053-1063. Lobo, A., Legendre, P., Rebollar, J. L. G., Carreras, J., & Ninot, J. M. (2004). Land cover classification at a regional scale in Iberia: separability in a multi-temporal and multi-spectral data set of satellite images. International journal of remote sensing, 25(1), 205-213. Malekpour, P., Taleai, M., & ASSIST, P. (2011). Modeling of Relationship between Land use/Cover and land Surface Temperature Using ASTER datasets. Journal of Environmental Studies, 37(58). Mallick, J., Kant, Y., & Bharath, B. D. (2008). Estimation of land surface temperature over Delhi using Landsat-7 ETM+. J Indian Geophys Union, 12(3), 131-140. Mather, A. S. (1986). Land use (p. 286). London; New York: Longman.
Meng, F., & Liu, M. (2013). Remote-sensing image-based analysis of the patterns of urban heat islands in rapidly urbanizing Jinan, China. International Journal of Remote Sensing, 34(24), 8838-8853. Mirkatouli, J., Hosseini, A., & Neshat, A. (2015). Analysis of land use and land cover spatial pattern based on Markov chains modelling. City, Territory and Architecture, 2(1), 1-9. Mirzaei, P. A. (2015). Recent Challenges in Modelling of Urban Heat Island. Sustainable Cities and Society, 19, 200-206. Mirzaei, P. A., & Haghighat, F. (2010). Approaches to study urban heat island–abilities and limitations. Building and Environment, 45(10), 2192-2201. Mirzaei, P. A., Haghighat, F., Nakhaie, A. A., Yagouti, A., Giguère, M., Keusseyan, R., & Coman, A. (2012). Indoor thermal condition in urban heat Island–Development of a predictive tool. Building and environment, 57, 7-17. Morad, M., Chalmers, A. I., & O’regan, P. R. (1996). The role of root-mean-square error in the geotransformation of images in GIS. International Journal of Geographical Information Science, 10(3), 347-353. Morris, K. I., Salleh, S. A., Chan, A., Ooi, M. C. G., Abakr, Y. A., Oozeer, M. Y., & Duda, M. (2015). Computational study of urban heat island of Putrajaya, Malaysia. Sustainable Cities and Society, 359-372. Peron, F., De Maria, M. M., Spinazzè, F., & Mazzali, U. (2015). An analysis of the urban heat island of Venice mainland. Sustainable Cities and Society,19, 300-309. Prata, A. J. (1993). Land surface temperatures derived from the advanced very high resolution radiometer and the along‐track scanning radiometer: 1. Theory.Journal of Geophysical Research: Atmospheres (1984–2012), 98(D9), 16689-16702. Priyadarsini, R. (2009). Urban Heat Island and its impact on building energy consumption. Advances in building energy research, 3(1), 261-270.
Radhi, H., Sharples, S., & Assem, E. (2015). Impact of urban heat islands on the thermal comfort and cooling energy demand of artificial islands—A case study of AMWAJ Islands in Bahrain. Sustainable Cities and Society,19, 310-318. Ramachandra, T. V., & Uttam, K. (2009). Land surface temperature with land cover dynamics: multiresolution, spatio-temporal data analysis of Greater Bangalore. International Journal of Geoinformatics, 5(3). Ramachandra, T. V., Aithal, B. H., & Sanna, D. (2012). Land Surface Temperature Analysis in an Urbanising Landscape through Multi-Resolution Data. Research & Reviews: Journal of Space Science & Technology, 1(1), 1-10 Raynolds, M. K., Comiso, J. C., Walker, D. A., & Verbyla, D. (2008). Relationship between satellite-derived land surface temperatures, arctic vegetation types, and NDVI. Remote Sensing of Environment, 112(4), 1884-1894. Rizwan, A. M., Dennis, L. Y., & Chunho, L. I. U. (2008). A review on the generation, determination and mitigation of Urban Heat Island. Journal of Environmental Sciences, 20(1), 120-128. Rosenfield, G. H., & Fitzpatrick-Lins, K. (1986). A coefficient of agreement as a measure of thematic classification accuracy. Photogrammetric engineering and remote sensing, 52(2), 223-227. Serrano, L., Filella, I., & Penuelas, J. (2000). Remote sensing of biomass and yield of winter wheat under different nitrogen supplies. Crop Science, 40(3), 723-731. Snyder, W. C., Wan, Z., Zhang, Y., & Feng, Y. Z. (1998). Classification-based emissivity for land surface temperature measurement from space. International Journal of Remote Sensing, 19(14), 27532774. Song, C., Woodcock, C. E., Seto, K. C., Lenney, M. P., & Macomber, S. A. (2001). Classification and change detection using Landsat TM data: when and how to correct atmospheric effects? Remote sensing of Environment, 75(2), 230-244.
Stathopoulou, M., Cartalis, C., & Petrakis, M. (2007). Integrating Corine Land Cover data and Landsat TM for surface emissivity definition: application to the urban area of Athens, Greece. International Journal of Remote Sensing, 28(15), 3291-3304. Statistical Center of Iran. (2011). Yearbook of Iran, Tehran. Streutker, D. R. (2002). A remote sensing study of the urban heat island of Houston, Texas. International Journal of Remote Sensing, 23(13), 2595-2608. Su, W., Gu, C., & Yang, G. (2010). Assessing the impact of land use/land cover on urban heat island pattern in Nanjing City, China. Journal of Urban Planning and Development. Sun, D., & Kafatos, M. (2007). Note on the NDVI‐LST relationship and the use of temperature‐related drought indices over North America. Geophysical Research Letters, 34(24). Taha, H. (2015). Meteorological, air-quality, and emission-equivalence impacts of urban heat island control in California. Sustainable Cities and Society, 19, 207-221. Touchaei, A. G., & Wang, Y. (2015). Characterizing urban heat island in Montreal (Canada)—Effect of urban morphology. Sustainable Cities and Society, 19, 395–402. Valor, E., & Caselles, V. (1996). Mapping land surface emissivity from NDVI: Application to European, African, and South American areas. Remote sensing of Environment, 57(3), 167-184. Van de Griend, A. 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(6), 1119-1131. Weng, Q. (2009). Thermal infrared remote sensing for urban climate and environmental studies: Methods, applications, and trends. ISPRS Journal of Photogrammetry and Remote Sensing, 64(4), 335-344. Weng, Q., Lu, D., & Schubring, J. (2004). Estimation of land surface temperature–vegetation abundance relationship for urban heat island studies. Remote sensing of Environment, 89(4), 467-483.
Wong, M. S., & Nichol, J. E. (2013). Spatial variability of frontal area index and its relationship with urban heat island intensity. International journal of remote sensing, 34(3), 885-896. Yan, G., Mas, J. F., Maathuis, B. H. P., Xiangmin, Z., & Van Dijk, P. M. (2006). Comparison of pixel‐based and object‐oriented image classification approaches—a case study in a coal fire area, Wuda, Inner Mongolia, China. International Journal of Remote Sensing, 27(18), 4039-4055. Yang, J., Yu, Q., & Gong, P. (2008). Quantifying air pollution removal by green roofs in Chicago. Atmospheric environment, 42(31), 7266-7273. 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. Yue, W., Xu, J., Tan, W., & Xu, L. (2007). The relationship between land surface temperature and NDVI with remote sensing: application to Shanghai Landsat 7 ETM+ data. International Journal of Remote Sensing, 28(15), 3205-3226. Zinzi, M., & Agnoli, S. (2012). Cool and green roofs: An energy and comfort com-parison between passive cooling and mitigation urban heat island techniquesfor residential buildings in the Mediterranean region. Energy and Buildings, 55, 66–76.
Figure Captions
Figure 1. Situation of the study area, Tehran
Figure 2. Position of the synoptic weather stations in Tehran
Figure 3. Classified map of land surface temperature, July 2010
Figure 4. LULC map prepared from Landsat TM satellite image]
Figure 5. NDVI map of Tehran
Figure 6. Relationship between land surface temperature (LST) and vegetation index (NDVI)
Figure 7. Spatial distribution of the population in Tehran, 2011
Figure 8. Population of districts of Tehran, 2011
Figure 9. Landsat TM imagery of 07 August 2010 in 164 pass and 35 row
Tables Table 1. Error Matrix of the Landsat TM Image in 2010 Classes
Frost Park
Bare land
Green lands
Water bodies
Constructed areas
Asphalt
Total
User accuracy
Frost Park
254
0
8
0
0
7
269
94,42%
Bare land
0
217
0
0
4
0
221
98,19%
Green lands Water bodies Constructed areas
1
0
322
0
0
0
323
99,69%
0
0
0
111
0
0
111
100%
0
0
0
0
164
0
164
100%
Asphalt
0
0
0
0
0
73
73
100%
Total Producer accuracy
255
217
330
111
168
80
1161
-
99,60%
100%
97,57%
100%
97,61%
91,25%
-
-
S2210-6707(16)30040-3 http://dx.doi.org/doi:10.1016/j.scs.2016.03.009 SCS 390
To appear in: Received date: Revised date: Accepted date:
29-12-2015 13-3-2016 15-3-2016
Please cite this article as: Bokaie, Mehdi., Zarkesh, Mirmasoud Kheirkhah., Arasteh, Peyman Daneshkar., & Hosseini, Ali., Assessment of Urban Heat Island Based on the Relationship between Land Surface Temperature and Land Use/Land Cover in Tehran.Sustainable Cities and Society http://dx.doi.org/10.1016/j.scs.2016.03.009 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Assessment of Urban Heat Island Based on the Relationship between Land Surface Temperature and Land Use/Land Cover in Tehran Mehdi Bokaie a, Mirmasoud Kheirkhah Zarkesh a, Peyman Daneshkar Arasteh b, Ali Hosseini c * a
Department of Remote Sensing and GIS, IAU Science and Research Branch, Tehran, Iran
b
Department of Engineering and Technology, IKI University, Ghazvin, Iran
c
Department of Geography and Urban Planning, University of Tehran, 1417854151, Tehran Iran * E-mail: [email protected] (Corresponding author): Ali Hosseini
Abstract In this study, the relationship between land surface temperature (LST) and Land Use/Land Cover (LULC) in Tehran Metropolitan City was studied using Landsat TM satellite image. For this, LST was calculated in accordance with the algorithm (Jiménez‐Muñoz & Sobrino, 2003) and the LULC map was prepared based on supervised classification method. According to the LST map obtained by processing the thermal band of the satellite image, areas affected by Urban Heat Island (UHI) were detected to examine their status in relation to the existing LULC classes and the population density. The results showed that UHI created in Tehran are different in terms of the causative agent. This difference is primarily due to the status of LULC in the region and reflects the close relationship between land cover and land surface temperature. Also, the distribution of vegetation and green spaces in different areas of Tehran City was studied by normalized vegetation index (NDVI) using remotely-sensed data. Correlation study between the land cover and land surface temperature showed a negative correlation between these two parameters. Study of the average surface temperature in six LULC classes indicated full compliance of heat islands with LULC classes. Keywords: Land Surface Temperature (LST), Urban Heat Island (UHI), Normalized Difference Vegetation Index (NDVI), Land Use/ Cover (LULC), Change with Land Use/ Land Cover (LULC).
1.
Introduction
The century 21, during which about half of the global population have been urban dwellers, is called “urban century”. According to forecasts, this amount will reach even more than 70% by the year 2050. Cities are places with sever consumption of various sources and increasing the release of contaminants. They are also a serious threat to ecological resources and can restrict urban facilities and the rapid growth of urbanization (Hosseini et al. 2015; Mirkatouli et al, 2015). The rapid growth of urbanization and the development of areas covered with man-made land cover versus the lower the level of natural cover is one of the main causes of global climate change (Streutker, 2002; Sekizawa et al., 2015). Population growth in urban areas, especially in developing countries, puts a lot of pressure on natural resources and causes gradual loss of these resources (Amiri, 2009; Feizizadeh & Blaschke, 2013; Peron et al. 2015). In recent decades, UHI, which are known as areas with negative temperature gradient compared to the surroundings, is one of the most important factors affecting the quality of human life (Wong et al., 2013; Mirzaei, 2015). UHI phenomenon is a risk arising from uncontrolled growth of urban areas (Mirzaei et al., 2012). Even if the global climate is not getting warmer, cities are now faced with the problem of rising temperature (Hoverter, 2002; Zinzi & Agnoli , 2012; Giannopoulou et al., 2011; Dimoudi et al., 2014; Taha, 2015). Cities with different physical surfaces than the surrounding environments (rural areas) influence on their microclimate (Cai et al., 2011). The behavior of urban surfaces is far different from the behavioral pattern of natural surfaces in terms of the absorption of shortwave and long wave radiation, evaporation, release of man-made heat, and the block of prevailing wind (Mirzaei & Haghighat, 2010). The difference in physical material of urban surfaces is very high. Man-made cover and surfaces such as asphalt, flooring, and concrete, instead of natural pervious surfaces such as soil and vegetation, reduce evapotranspiration and increase sensible heat in cities. As a result, cities experience a warmer weather than surrounding areas. Furthermore, accelerated deformation and design of buildings and the
emergence and spread of skyscrapers as well as increased industrial activities have led to the development of UHIs (Liu & Zhang, 2011; Busato et al., 2014; Radhi et al, 2015). Nowadays, raw materials are designed and built in such a way that have a high potential in absorbing solar radiations, high impermeability, and the desired thermal specifications for energy storage and release of heat. Knowing these thermal characteristics can lead to develop appropriate strategies for mitigation of the impacts of UHIs (Gagliano et al, 2015). Megacities with compact constructions and a variety of activities, especially in the CBD, fully affect on their climate and the climate of surroundings (Boehme et al., 2015; Morris et al., 2015). Natural cooling process in urban environments undergoes problems due to the extensive changes in the pattern of urban design and structure. A complex network of high-rise buildings and narrow streets, that makes trapping the heat absorbed during the day, impairs air circulation and consequently the process of reducing the temperature in urban areas. Calculating the amount of LST is considered a starting point for the analysis of UHI (Mather, 1986; Malekpour et al., 2011). One of the strategies adopted to reduce the heating and cooling load of buildings in urban areas is the increased use of thermal insulation (such as roof coatings). Despite the positive role of thermal insulation in buildings to prevent the loss of heat energy in the winter, they increase indoor temperature of buildings in summer, which results in increased energy use for cooling purposes (Gagliano et al, 2015). In studies of urban climate change, the LST has been recognized as one of the most important parameters affecting the UHI phenomenon (Liu & Zhang, 2011). Further, in many of the studies on the analysis of UHI, a very close relationship between LST and LULC as well as considerable impressibility of UHI by this relationship have been emphasized (Weng et al., 2004). Expansion of synthetic surfaces is a marker for centralized human activities that lead to an increase in Earth's surface temperature (Su et al., 2010).
The exact calculation of the LST and the study of relationship between current LULC and LST can be an important factor in order to solve many problems in the field of climate change in urban areas and interaction between humans and the environment (Ramachandra & Uttam, 2009). Developing remote sensing technology, launching various satellites, and obtaining images with high spatial resolution power at large spectral range as well as providing imaging capabilities at reasonable intervals with repetitive coverage of the Earth's land surface have made possible monitoring and investigating of the land surface (Mallick et al., 2008). So far, a large number of satellites and sensors equipped with thermal infrared bands have been used for the study and monitoring of UHI. In the absence of regular and dense network of ground weather stations, temporal and spatial distribution of LST based on thermal imaging can be used to supply the required input data for the study of UHI (Feizizadeh et al., 2013). Images captured by Landsat TM sensor are one of the most widely used satellite images in the field of environmental researches that uses the data of the band 6 within the range between 10.4 and 12.5 microns to study and analyze the ground surface phenomena such as UHI (Ramachandra et al., 2012). According to the foregoing, the present study was carried out to follow three objectives: 1. The study of LST in Tehran City using Landsat TM satellite images, 2. Analysis of the spatial distribution of LST and its relationship with LULC and Normalized Difference Vegetation Index (NDVI), and 3. Analysis of relationship between LST and the population density as well as its impact on energy consumption and the health of residents in the study area. To calculate LST, one of the widely used remote sensing methods that is based on thermal infrared wavelength band was used (Jiménez‐Muñoz & Sobrino, 2003). The LULC map of the study area was prepared using Landsat TM image classified by the maximum likelihood algorithm. The population distribution was also studied in the form of a map prepared from the latest census in 2011.
2.
Materials and methods
2.1 Study area The present study was conducted in Tehran, the capital city of Iran, which is located in the northern part of the country at the southern hillside of Alborz Mountains within the longitudes of 51° to 51° 40′ E and the latitudes of 35° 30′ to 35° 51′ N (Hosseini et al. 2015). According to the latest Iranian Population and Housing Census in 2011 by the Statistical Centre of Iran, Tehran, with a population of 8,154,051 people, is still ranked first most populous city in Iran with a very distinct demographic difference than other cities (Statistical Center of Iran, 2011) (Figure 1). The study area leads to mountainous areas from the north and to desert areas for the south, so the southern and the northern parts have a different climate. The northern areas have a cold and dry climate, while the southern parts suffer from warm and dry climatic conditions. The altitude of the city varies from 900 m to 1800 m. The great variation in the altitude is due to the large extent of the city. In the Tehran metropolitan, the annual mean temperature changes between 15 to 18 Celsius degrees and given the height differences in the city, the various parts have an average of 3 Celsius degrees difference in temperature. The aviation of monthly mean relative humidity include minimum and maximum relative humidity registered at Mehrabad Station shows that in mornings, humidity changes from a minimum of 38% to a maximum of 79% in July to January. The changes in the relative humidity in midday vary between 15% in June to 47% in February. The annual precipitation in Tehran is mainly affected by the height differences, varying between a maximum of 422 millimeters in the north and a minimum of 145 millimeters in the southeast. The number of rainy days follows the same pattern and varies between 89 days in the north and 33 days in the south. Also in the city area, there are about 205 to 213 days of clear sky with some clouds per year (Habibi & Hourcade, 2005).
Figure 1. Situation of the study area, Tehran City
2.2. Satellite images and meteorological data In order to estimate the LST and to investigate the areas affected by UHI, Landsat TM satellite image dated back to 7 August 2010 was used. To prepare the LULC map, discrete bands of the image in the visible, near-infrared, and mid- infrared spectral range were stacked by layer stack command. Further, the data of air temperature measured in five synoptic weather stations in Tehran, including Chitgar, Dowshantape, Mehrabad, Aqdasieh, and Geophysics, were used as one of the model inputs to estimate LST. These data have been measured within the 3 hour-intervals from 12 a.m. to 9 p.m. GMT (Greenwich Mean Time). So, there are 8 measured temperature data available for each day. The position of the synoptic weather stations is depicted in Figure 2.
Figure 2. Position of the synoptic weather stations in Tehran City
2.3. Pre-processing of Landsat Image Many raw satellite images, due to atmospheric conditions and possible mechanical problems in the sensors, have some distortions and errors that must be corrected using different methods prior to processing them (Yan et al., 2006). Atmospheric and radiometric corrections are considered among the most important pre-process steps in the processing of satellite images for the purposes of image classification and preparation of LULC map, as well as estimation of LST (Song et al., 2001; Li et al., 2004). For radiometric corrections, the Digital Numbers (DNs) of all pixels in the image were converted to spectral radiance using the data in the image header file and based on the Equation 1. Then, to estimate the LST, the brightness temperature was calculated using Equation 2 (Chander et al., 2009). In addition, in order to minimize the effects of atmospheric conditions on the image, ENVI software was used to do atmospheric corrections. Dark Object Subtraction method was used to correct satellite
images atmospherically. This method is available as a pre-processing step for several of the in-
scene based atmospheric correction methods. Dark object subtraction method searches each bands of multi spectral image for the darkest pixel value. Assuming that dark objects reflect no light, any value greater than zero must result from atmospheric scattering. ENVI Classic removes the scattering by subtracting this value from every pixel in the band. This simple technique is effective for haze correction and normally produces improved results for in-scene atmospheric correction in the shorter wavelengths where scattering dominates (Song et al., 2001). 𝐿(𝜆) = (
𝐿𝑀𝑎𝑥 −𝐿𝑀𝑖𝑛 255
) ∗ 𝐷𝑁 + 𝐿𝑀𝑖𝑛 (1)
Where;
DN is the degree of greyness of the pixels and L (λ) (radiance) is in (W/m2/sr/mm). LMax and LMin are the calibration constants of the sensor, equal to the maximum and minimum values of the spectral radiance (in W/m2/sr/ mm) detectable for each band, by the sensor Landsat TM. LMax values for band 3 and band 4 Landsat TM are 264.00 and 221.00, and LMin values for band 3 and band 4 are -1.170 and -1.510 (Chander et al., 2009) T=
𝐾2 𝐾1 𝑙𝑛( +1) 𝐿ℷ
(2)
In which; T= brightness temperature, 𝐾1 and 𝐾2 are correction constants of 607.76 and 1260.56, and Lλ = spectral radiance of sensor in Wm-2sr-1μm-1. Despite the primary geometric corrections on the Landsat products, in order to eliminate possible errors in the image of the study area, it was geometrically rectified using national topographical maps at a scale of 1: 25000. The rectification process was performed by first-order Polynomial model using nearest neighbor re-sampling method. RMSE was calculated using the equation (3) to check the accuracy of
geometric correction. This parameter calculates the difference between observed values and predicted values (Morad et al., 1996) 1
𝑅𝑀𝑆𝐸 = √ ∑𝑁 ̂𝑖 )2 (3) 𝑖=1( 𝑥𝑖 − 𝑥 𝑁
The RMS error of the image was estimated at 0.03.
2.4. Satellite Image Processing 2.4.1. NDVI NDVI is considered as one of the most widely- and commonly- used indices in environmental studies (Sun & Kafatos, 2007; Sekizawa et al., 2015). If the initial processing and corrections are done well on the image, the values of this index will be assured confidently (LI et al., 2009). This index is the difference between the maximum absorption in the red region of chlorophyll pigments and a maximum reflectivity in the near infrared region of the cell structure leaves (Baret & Guyot, 1991). And is calculated with Red and Near Infrared band in multi spectral satellites using equation”4”
𝑁𝐷𝑉𝐼 =
( 𝑁𝐼𝑅−𝑅𝐸𝐷 ) ( 𝑁𝐼𝑅+𝑅𝐸𝐷 )
(4)
In theory the value of this index is variable in range of (+1 and -1). The value of NDVI for dense vegetation usually differs from 0.3 to 0.6 and negative values indicate the clouds, snow covers and water (Serrano et al., 2000) There have been many studies on the relationship between LST and vegetation cover by analyzing the values of NDVI and surface temperature in which the key role of NDVI and its variations on LST map at different LULC classes are emphasized (Weng et al., 2004; Yue et al., 2007; Raynolds et al., 2008). The present study also examined correlation between changes in LST and vegetation cover measured by NDVI.
2.4.2. Emissivity Total radiation measured from a surface can be reduced or increased as by changes in the emissivity (Prata, 1993). Therefore, when the temperature of a land surface is calculated using satellite images, it would be quite important to estimate accurately the extent of emissivity from the given surface whereas incorrect estimate of emissivity or its ignorance will raise the error in the calculated temperature up to 1 or 2 degrees Kelvin (Caselles et al., 1995). Emissivity of various surfaces is a function of different parameters such as properties, water content, chemical composition, physical structure, and texture of surfaces (Snyder et al., 1998). Further, emissivity of different surfaces varies in different wavelengths. The rate of variations is less in wavelengths between 10.5 μm and 12.5 μm, which is the spectral range of the thermal band in TM sensor (Weng, 2009; Stathopoulou, 2007). In areas with dense land cover, assuming a constant coefficient for the emissivity would be a realistic assumption. However, in poorly vegetated areas or bare lands with various types of soil and minerals, it is necessary to consider the emission factor of each of the pixels in measurement of temperature. In the absence of required facilities for laboratory tests, satellite images are used to calculate emissivity of different materials and ground features at different wavelengths. The NDVI is considered as one of the most common methods of this type (Valor & Caselles, 1996). In the present study, in order to estimate the emissivity, the NDVI was used based on the method presented by Van de Griend (1993).The average emissivity coefficients for barren land, green space, forest parks, water bodies, asphalt surfaces, and residential areas were calculated as 0.932, 0.937, 0.937, 0.931, 0.933, and 0.932, respectively.
2.4.3. Land Surface Temperature retrieval from Landsat TM image After calculation of the image brightness temperature, using the method developed for the sensors with only one thermal bond, the LST was computed based on the Equations 3 to 8 (Jiménez‐Muñoz & Sobrino, 2003).
Ts = γ [ε−1 (φ1 Ls + φ2 ) + φ3 ] + δ γ= {
C2 Ls λ4 T2b
[
C1
Ls + λ−1 ]}
−1
(5)
(6)
δ = −γ Ls + Ts (7) In which, Ts= land surface temperature, λ=effective wavelength, which was regarded to be 11.5 microns in this research, ε= surface emissivity, C1 and C2 =constant values of 1.19104×108 μm 4 m -2 sr -1 and 14387.7 μm K, respectively, and φ1, φ2, φ3 = atmospheric functions that are calculated using the Equations 8 to 10 and the amount of water vapor in the atmosphere at the time of imaging. The average water vapor in the study area was estimated at 0.748. φ1=0.14714 W2 – 0.15583 W+1.1234
(8)
φ2= −1.1836 w2 −0.37607w−0.52894
(9)
φ3= −0.04554 w2 + 1.8719 w−0.39071
(10)
2.4.4. Land cover classification Satellite images are widely used to monitor LULC changes and to prepare new LULC maps (Lobo et al., 2004). LULC refers to the data obtained from satellite images that are classified based upon the values of different pixels in an image. These data are generally in the format of raster or grid. Each pixel has a value that links it to a particular class in the classification process. In this study, Landsat TM image was used to investigate relationship between LST and LULC changes in the study area as well as their influence on the spatial distribution of UHI. For this, various types of LULC were classified in to six classes including bare lands, constructed areas, green spaces, forest park, water bodies, and asphalt-paved surfaces using supervised classification method and maximum likelihood algorithm. Six classes of LULC were defined based on our knowledge of case study (the Tehran city) which it had been provide from
high resolution satellite imagery and field visits. Ensuring the accuracy of the classification, SPOT 2.5 meter satellite image dated of 03 March 2010 was used. Validation is considered one of the most important parts of image classification process because it examines the degree of success of an algorithm in classification of different objects in an image. The accuracy of a classification process is usually assessed by comparing the results of classification with reference data from field visits, high spatial resolution images, or aerial photographs (Rosenfield & Fitzpatrick-Lins, 1986). Results of accuracy assessment are usually presented in the form of a matrix error
(confusion matrix). Overall classification accuracy is computed by dividing the number of pixels categorized in the same class (based on either field visits or satellite images) on the number of all pixels (Bogoliubova,& Tymków, 2014).
𝑇=
∑ 𝐷𝑖𝑖 𝑁
(11)
Where; Dii∑ is the number of pixels that are classified correctly and N represents all the pixels in matrix error Unfortunately, the overall accuracy parameter cannot provide classification accuracy of each land use class, separately. Therefore, user accuracy and producer accuracy parameters were used to check the classification accuracy of each of the classes. Producer accuracy examines the probability that a particular class of land cover on the ground is classified under the same category in the satellite image while user accuracy investigates the probability of correct labeling of a particular land use class assigned to a pixel. For example, in this study, the producer accuracy for the land use class of green lands was calculated as (322⁄330) = 97.57% and the user accuracy was computed as(322⁄323) = 99.69%. As one of the
indices of classification accuracy, Kappa coefficient was applied to verify the accuracy of the classified pixels compared to ground truth (Congalton, 1991) by the use of equation 12.
𝐾=
𝑁 ∑𝑟𝑖=1 𝑥𝑖𝑖 − ∑𝑟𝑖=1(𝑥𝑖+ ∗𝑥+𝑖 ) 𝑁2 − ∑𝑟𝑖=1(𝑥𝑖+ ∗𝑥+𝑖 )
(12)
In which; r =number of rows in the error matrix, xii= number of observations in the ith row and column, x+i= total number of observations in the ith column, and N= total number of observations. The overall accuracy and Kappa coefficient of the classification were 93.41% and 0.9166, respectively. The results of supervised classification error matrix are given in Table 1. Besides, Figure 4 presents the LULC map prepared from the satellite image are illustrated in.
Table 1. Error Matrix of the Landsat TM Image in 2010
3.
Results
3.1. Relationship between LST and LULC Figure 3 demonstrates the LST map prepared from Landsat TM satellite image. The temperature in this map were classified based on standard deviation of values and subsequently, areas affected by UHI were identified. According to the LST map, surface temperature varied between 21.5 ◦C and 57.9 ◦C. At the time of imaging, the lowest average temperature was 35 ◦C related to the water bodies and the maximum average temperature was 48 ◦C recorded from bare lands in the study area. To clarify the relationship between the LST and land cover, particularly vegetation cover whose distribution status in the study was examined with NDVI, variations in the temperature of different LULC classes were studied. Comparing the estimated temperature values and the LULC map revealed the thermal variation in
different LULC classes. According to which, the lowest temperature corresponds to the water bodies and areas covered by natural coatings such as green space, forest parks, and urban parks, while maximum temperature was recorded in bare lands and impervious surfaces such as asphalt- paved areas and other man-made coatings, as well as industrial, commercial, residential, and transport land uses. Generally speaking, areas largely influenced by UHI are mainly located in 4 different places in the study area including various land uses. One of the most important affected areas is District 21 of the Tehran City housed the vast majority of factories, industrial workshops, warehouses, and other related land uses, as a chain of industrial activities. In other words, majority of the district is occupied by industrial land use and man-made coatings. The second area that is very largely influenced by the UHI is District 9. This is mainly due to Mehrabad International Airport, passenger transport terminals, and the gate of Tehran from the west as one of the oldest areas affected by this phenomenon. Heavy road and air traffic, and large volume of public and private vehicles, as well as a variety of pollutants caused by them are the most important negative aspects of this district. The third important UHI-affected area is located in the north and northwest of the District 22 and the north of the District 19 due to the bare lands spread in the region. The causative agent of this phenomenon in these districts is different from the rest areas.
Figure 3. Classified map of land surface temperature, July 2010 Figure 4. LULC map prepared from Landsat TM satellite image
In addition to these areas, it can be noted to the District 12 (Central Market and the old core of Tehran City), District 13 (bare lands around the former Doshan-Tapeh Airport and its surrounding asphalt-paved areas) and the southern areas of Tehran where there are workshops and industrial activities. However, the spatial development of HUI is less severe in these areas. In some parts, such as District 12,
enforcement of new laws and regulations such as amendments to the vehicle and traffic laws and development of pedestrian-only streets is among the steps taken in recent years to reduce the impact and progress of this phenomenon. In contrast to the above-mentioned areas largely affected by UHI phenomenon experienced changes in the temperature and energy consumption, there are areas with natural covering such as forest parks and green spaces marked in green in Figure 3. Moreover, water bodies and residential areas with low to relatively medium population density are less-influenced from the UHI compared to the largely-affected areas. These areas are displayed in the figure among areas with low to medium temperature.
3.2. Land Surface Temperature and NDVI Increase the area of vegetation cover and green spaces in urban environments has a significant influence on the thermal condition of cities and development of UHI. Vegetation cover through evapotranspiration process can reduce and adjust the temperature of land surface (Yuan & Bauer, 2007). That is why in many of the studies conducted in order to find solutions to deal with the phenomenon and mitigate its effects; the role of vegetation cover is mentioned as one of the most important ways to reduce the effects of UHI (Rizwan et al., 2008; Corburn, 2009). Although the use of vegetation cover, especially trees, is a perfect solution to reduce emissions and mitigate the impacts of UHI phenomenon, but planting trees and development of green spaces is not easy in dense urban areas (Gaffin et al., 2008). The use of green roofs and development of green surfaces on the upper level of buildings can be considered as one of the ways to reduce the temperature of cities and improve air quality (Yang et al., 2008). Evapotranspiration process can decrease the air temperature at rates between 1 and 5 ° C (Farina, 2001). Generally, areas with suitable vegetation cover and the high NDVI shall experience much lower surface temperature than the areas with poor vegetation cover. On the contrary, the LST is higher in
areas with less vegetation cover. However, the negative correlation between NDVI and LST may be affected by factors such as soil moisture and evapotranspiration status, however, the role of vegetation cover and its impact on the reduction of air and surface temperature is often undeniable (Su et al., 2010). LST maps obtained using remote sensing technology and satellite imagery, in addition to displaying temperature at different LULC classes; reveal spatial distribution of UHI in different areas (Figure 3). The spatial pattern of LST and NDVI in the study area is displayed in Figure 5. At the time of the study, the NDVI value in Tehran varied between 0.15 and 0.34. Areas with the lowest NDVI value are mainly spread in northwest, south, and southwest of the city. According to the LULC map, these areas are mainly covered by bare lands and industrial land uses. Areas with high NDVI are located in the north, northeast, and parts of southeast and west of the study area, which in most cases correspond to the urban parks, forest parks, and parts of agricultural lands (especially in the southeast), respectively. The surface temperature is always influenced by water content and vegetation cover. In other words, there is a direct linear relationship between temperature, and material and coating of the surface (Yue et al., 2007; Raynolds et al., 2008). In this study, a total number of 500 points were selected randomly to study the relationship between LST and NDVI in the form of a two-dimensional scatter plot (Figure 6).
As the figure suggests, LST is negatively correlated with the NDVI so that LST increases by reducing the amount of vegetation cover. It makes more evident the role of vegetation in decreasing temperature. As shown in the graph, the lowest recorded LST (302-304°K) is observed in the areas with the highest levels of vegetation cover. As the graph goes towards the areas with less vegetation cover, the surface temperature is increased. The largest proportion of data is concentrated between the temperature range of 307-310 °K and the NDVI values of 0.17 to 0.2. The downtrend continues up to the end. The fitted line is also an affirmation of the same negative relationship between the data. The coefficient of determination (R2=0.6) of the regression model also showed a good fit to the data. In addition, in order to verify the accuracy of the regression model, the RMS error was computed for the LST and NDVI data.
The error value of 0.0196 can be due to the precision of the developed model and the accuracy of the relationship between LST and NDVI parameters in the study area. According to the figure 6, there is a negative correlation between the LST and NDVI so that the surface temperature of land will increase by reducing the amount of vegetation cover.
Figure 5. NDVI map of Tehran City Figure 6. Relationship between land surface temperature (LST) and vegetation index (NDVI)
3.3. Land Surface Temperature and Population Spatial pattern of heat islands in the study area reflects the different nature of their constituent and aggravating factors. The formation of this phenomenon in proximity of dense residential and commercial areas, together with traffic-induced air pollution, and consequently release of dangerous environmental pollutants are the most important factors threatening the health of urban residents and increasing the risk of many diseases, particularly respiratory and cardiovascular diseases. With the expansion of impervious and man-made surfaces due to the loss of natural land cover, particularly vegetation cover, fundamental problems have emerged in the process of self-purification of pollutants in urban areas (Lo et al., 2003). Heat capacity has increased recently in urban areas due to various constructed structures and the expanded use of unnatural materials such as asphalt, cement, concrete, etc. on the body of these structures. This increase has led to a greater absorption of solar energy, and thus increased temperature of urban areas (Priyadarsini, 2009). High density of population in an area and consequently the expansion of settlements and high-rise residential buildings, on one hand, can cause the energy to be trapped in walls due to high heat capacity, and on the other hand, lead to excessive use of artificial cooling systems by residents and anthropogenic heat in the city.
Based on the population of different districts of Tehran in 2010, the 21, 22, 9, and 12 districts that host the most important UHI cores in the city are among the most sparsely populated areas of Tehran (Figure 7 and Figure 8). Due to the status of different LULC classes, the low population density in these districts seems normal whereas the UHIs in these districts are mainly induced by industrial, airport and transport, and commercial land uses, as well as the barren lands. The LST map and population distribution in the different districts of Tehran reveal areas with a high temperature in the 14, 5, 2, 15, 4 and 20 districts where face a high density of population. Although these areas are not considered among the main cores of UHI in Tehran, but high population density, high traffic volume, and pollution caused by transport, put these areas at the risk of emergence and spread of UHIs.
Figure 7. Spatial distribution of the population in Tehran, 2011 Figure 8. Population of districts of Tehran, 2011
4.
Discussion
Although there is usually observed an island core in single-core cities, however, different urban planning patterns can cause the formation of different forms of heat islands. Considering that areas with dense buildings are generally warmer than the surroundings, multi-core population centers in cities can lead to creation of several heat islands or can exhibit the heat island as a multi-core UHI. As Figure 3 suggests, the study area is faced with different types of heat islands in terms of the causative agent. The only cause common to all them was very high temperature of these lands compared to their surroundings. Examination and comparison of LST and LULC maps as well as field visits showed a significant difference in the nature of areas affected by UHI. The most important areas that are known as heat cores and UHIaffected area in Tehran are adjacent areas to industrial estates and factories (District 21), dense residential and commercial structures with intricate texture of old and new (Districts 12, 16, and 20) or
the areas around the airport and major transport terminals (Districts 9, 18, and 19). The LST and LULC maps represent establishment of some UHI cores in areas with the land use class of barren land, particularly in the north-west of the study area. According to the 2011 Census by the Statistical Center of Iran, Tehran, including an overall population of 8,154,051 people, is ranked first most populous city in Iran with a serious demographic difference compared to other cities. As similar to the previous decades, the city is still the first choice of many immigrants due to the expansion and centralization of facilities, impressively high income rates, as well as centralization of job opportunities and other municipal services. District 22 in the North West of Tehran hosts one of the most important botanical parks and forest parks in the study area. The district has always attracted the attention of city managers as one of the bets weather areas of Tehran, so that following the rapid population growth of the city due to irregular migration, the district was proposed as one of the possible options for new constructions and city expansion in order to provide housing for the new residents. Accordingly, many residential townships have been established or are on the verge of completion. According to the LST and LULC maps as well as analysis the status of vegetation cover by NDVI map indicate the point that UHI-affected areas in this district are mainly barren lands. Unlike the discussed thermal cores in the vicinity of industrial areas, transportation terminals, and dense residential and commercial areas, the emergence and development of heat islands in the District 22 and the North West of Tehran happens under the influence of different changing trend of soil temperature and barren lands compared to the surroundings. This could reflect the fact that the absence of particular natural or even particular coatings in an area can help the emergence and development of UHIs. Increasing development of construction processes in the District 22 in northwest of Tehran, and the use of a variety of construction materials such as cement, concrete, and clay bricks, the severity and extent of UHI-affected areas will be added.
High levels of NDVI in Tehran fully match the coatings of green spaces (parks, gardens, and grass coverings) and forest parks. The negative relationship between the status of vegetation cover and LST changes in the study area also emphasizes the expansion of vegetation cover as a way to deal with the emergence and escalation of UHIs. Comparing maps of vegetation cover, population density, and LST, it can be argued that some areas, such as 13, 10, 11, 7, and 8 Districts, due to high density of residential and commercial areas and the lack of sufficient vegetation cover, are in danger of becoming the new HUI cores in the future. The effect of changes in LULC on LST changes is also evident in the District 21, which has been converted to a long chain of industrial, workshop, and warehousing land uses. Considering the status of LULC and also the poor condition of vegetation cover in this district, the existence of UHIs in this area does not seem so strange. However, factors affecting the emergence and development of UHIs in this industrial district are different from other thermal cores of the city. Unlike the Districts 19 and 21 where the emergence of UHIs is induced by the existence of barren lands or some central parts of Tehran where the emergence of UHIs is attributed to the high density construction and abnormally high levels of artificial construction materials that mismatches with the natural coatings, UHIs in this industrial district have different origins. In other words, construction of industrial estates, factories, and small workshops, as well as conversion of a wide proportion of natural vegetation cover provides a ground for the emergence and spread of UHIs in this district. Nowadays, emissions from these industrial land uses are a contributing factor affecting the temperature difference between the industrial chain and the surrounding environment as well as durability of UHIs.
5.
Conclusion
Emergence and spread of UHIs occur under the influence of various factors. Depending on the circumstances and factors under which the UHIs have been initiated and expanded, there can be offered some strategies to mitigate their dimensions and eventually eliminate them. Analysis of the relationship
between the current LULC and LST as well as the factors influencing the growth and development of heat islands in Tehran are the main goals of this study. According to which, the study investigated the relationship between the LST and spatial pattern of land cover types, as well as the impact of population density on the development of heat islands in Tehran, the capital of Iran. Landsat TM image was used to prepare the surface temperature map of the study area and to analyze UHIs in the city. Before performing the desired processing to produce LST, NDVI, and LULC maps, the satellite images were geometrically corrected using the topographic maps at the scale of 1:25000. The radiometric and atmospheric corrections were also done by the use of ENVI software. Based on the results, several factors contribute to the development of UHI in the region. Industrial factories, airports, bare lands, and dense residential areas are the land uses more affected by UHI. Moreover, population density in different parts of the study area reflects the impact of human activities on the energy consumption and intensification of heat islands. Proximity of dense residential areas to the UHI-affected zones increases the risk of respiratory, and heart diseases among the residents and makes harder the conditions of everyday life, especially for the elderly and children. The LST in the six land use classes approved the full compliance of the two LST and LULC maps. So that areas with natural cover, particularly with vegetation cover and green spaces, with an average temperature of 40 ° C are the coolest places in the city, while bare lands with an average temperature of 48 ◦C, and asphalt-paved surfaces, with an average temperature of 46 ◦C are the hottest places. This shows the influence of natural cover such as water bodies and green spaces on mitigation of the intensity and spread of UHI. The obtained results also confirmed an adverse correlation between the spread of land cover and green spaces, measured by NDVI, and LST estimated by satellite images. As expected, with an increase in the NDVI, the LST was declined and in areas covered with green spaces, a lower average LST was observed.
References 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. Baret, F., & Guyot, G. (1991). Potentials and limits of vegetation indices for LAI and APAR assessment. Remote sensing of environment, 35(2), 161-173. Boehme, P., Berger, M., & Massier, T. (2015). Estimating the building based energy consumption as an anthropogenic contribution to Urban Heat Islands.Sustainable Cities and Society, 373- 384. Bogoliubova, A., & Tymków, P. (2014). Accuracy assessment of automatic image processing for land cover classification of St. Petersburg protected area. Acta Scientiarum Polonorum. Geodesia et Descriptio Terrarum, 13(1-2). Busato, F., Lazzarin, R. M., & Noro, M. (2014). Three years of study of the Urban Heat Island in Padua: Experimental results. Sustainable Cities and Society, 10, 251-258. Cai, G., Du, M., & Xue, Y. (2011). Monitoring of urban heat island effect in Beijing combining ASTER and TM data. International journal of remote sensing, 32(5), 1213-1232. Caselles, V., Coll, C., Valor, E., & Rubio, E. (1995). Mapping land surface emissivity using AVHRR data application to La Mancha, Spain. Remote Sensing Reviews, 12(3-4), 311-333. 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(5), 893-903. Congalton, R. G. (1991). A review of assessing the accuracy of classifications of remotely sensed data. Remote sensing of environment, 37(1), 35-46. Corburn, J. (2009). Cities, climate change and urban heat island mitigation: Localising global environmental science. Urban Studies, 46(2), 413-427.
Dimoudi, A., Zoras, S., Kantzioura, A., Stogiannou, X., Kosmopoulos, P., & Pallas, C. (2014). Use of cool materials and other bioclimatic interventions in outdoor places in order to mitigate the urban heat island in a medium size city in Greece. Sustainable Cities and Society, 13, 89-96. Farina, A. (2011). Exploring the relationship between land surface temperature and vegetation abundance for urban heat island mitigation in Seville, Spain. LUMA-GIS Thesis. Feizizadeh, B., & Blaschke, T. (2013). Land suitability analysis for Tabriz County, Iran: a multi-criteria evaluation approach using GIS. Journal of Environmental Planning and Management, 56(1), 1-23. Feizizadeh, B., Blaschke, T., Nazmfar, H., Akbari, E., & Kohbanani, H. R. (2013). Monitoring land surface temperature relationship to land use/land cover from satellite imagery in Maraqeh County, Iran. Journal of Environmental Planning and Management, 56(9), 1290-1315. Gaffin, S. R., Rosenzweig, C., Khanbilvardi, R., Parshall, L., Mahani, S., Glickman, H., Goldberg. R., Blake, R., Slosberg, R. B. & Hillel, D. (2008). Variations in New York City’s urban heat island strength over time and space. Theoretical and applied climatology, 94(1-2), 1-11. Gagliano, A., Detommaso, M., Nocera, F., & Evola, G. (2015). A multi-criteria methodology for comparing the energy and environmental behavior of cool, green and traditional roofs. Building and Environment, 90, 71-81. Giannopoulou, K., Livada, I., Santamouris, M., Saliari, M., Assimakopoulos, M., & Caouris, Y. G. (2011). On the characteristics of the summer urban heat island in Athens, Greece. Sustainable Cities and Society, 1(1), 16-28. Habibi, M., Hourcade, B. (2005). Atlas of Tehran Metropolis. Tehran Geographic Information Center and CNRS, Vol. 1. - Land and People Hosseini, A., Pourahmad, A., & Pajoohan, M. (2015). Assessment of Institutions in Sustainable UrbanManagement Effects on Sustainable Development of Tehran: Learning from a Developing Country. Journal of Urban Planning and Development, 05015009.
Hoverter, S. P. (2012). "Adapting to urban heat: a tool kit for local governments. “Georgetown Climate Center. Jiménez‐Muñoz, J. C., & Sobrino, J. A. (2003). A generalized single‐channel method for retrieving land surface temperature from remote sensing data.Journal of Geophysical Research: Atmospheres (1984–2012), 108(D22). Li, F., Jackson, T. J., Kustas, W. P., Schmugge, T. J., French, A. N., Cosh, M. H., & Bindlish, R. (2004). Deriving land surface temperature from Landsat 5 and 7 during SMEX02/SMACEX. Remote sensing of environment, 92(4), 521-534. Li, H., Liu, Q. H., & Zou, J. (2009). Relationships of LST to NDBI and NDVI in Changsha-Zhuzhou-Xiangtan area based on MODIS data. Scientia Geographica Sinica, 2, 018. Liu, L., & Zhang, Y. (2011). Urban heat island analysis using the Landsat TM data and ASTER data: A case study in Hong Kong. Remote Sensing, 3(7), 1535-1552. Lo, C. P., & Lo, C. P., & Quattrochi, D. A. (2003). Land-Use and Land-Cover Change, Urban Heat Island Phenomenon, and Health Implications. Photogrammetric Engineering & Remote Sensing, 69(9), 1053-1063. Lobo, A., Legendre, P., Rebollar, J. L. G., Carreras, J., & Ninot, J. M. (2004). Land cover classification at a regional scale in Iberia: separability in a multi-temporal and multi-spectral data set of satellite images. International journal of remote sensing, 25(1), 205-213. Malekpour, P., Taleai, M., & ASSIST, P. (2011). Modeling of Relationship between Land use/Cover and land Surface Temperature Using ASTER datasets. Journal of Environmental Studies, 37(58). Mallick, J., Kant, Y., & Bharath, B. D. (2008). Estimation of land surface temperature over Delhi using Landsat-7 ETM+. J Indian Geophys Union, 12(3), 131-140. Mather, A. S. (1986). Land use (p. 286). London; New York: Longman.
Meng, F., & Liu, M. (2013). Remote-sensing image-based analysis of the patterns of urban heat islands in rapidly urbanizing Jinan, China. International Journal of Remote Sensing, 34(24), 8838-8853. Mirkatouli, J., Hosseini, A., & Neshat, A. (2015). Analysis of land use and land cover spatial pattern based on Markov chains modelling. City, Territory and Architecture, 2(1), 1-9. Mirzaei, P. A. (2015). Recent Challenges in Modelling of Urban Heat Island. Sustainable Cities and Society, 19, 200-206. Mirzaei, P. A., & Haghighat, F. (2010). Approaches to study urban heat island–abilities and limitations. Building and Environment, 45(10), 2192-2201. Mirzaei, P. A., Haghighat, F., Nakhaie, A. A., Yagouti, A., Giguère, M., Keusseyan, R., & Coman, A. (2012). Indoor thermal condition in urban heat Island–Development of a predictive tool. Building and environment, 57, 7-17. Morad, M., Chalmers, A. I., & O’regan, P. R. (1996). The role of root-mean-square error in the geotransformation of images in GIS. International Journal of Geographical Information Science, 10(3), 347-353. Morris, K. I., Salleh, S. A., Chan, A., Ooi, M. C. G., Abakr, Y. A., Oozeer, M. Y., & Duda, M. (2015). Computational study of urban heat island of Putrajaya, Malaysia. Sustainable Cities and Society, 359-372. Peron, F., De Maria, M. M., Spinazzè, F., & Mazzali, U. (2015). An analysis of the urban heat island of Venice mainland. Sustainable Cities and Society,19, 300-309. Prata, A. J. (1993). Land surface temperatures derived from the advanced very high resolution radiometer and the along‐track scanning radiometer: 1. Theory.Journal of Geophysical Research: Atmospheres (1984–2012), 98(D9), 16689-16702. Priyadarsini, R. (2009). Urban Heat Island and its impact on building energy consumption. Advances in building energy research, 3(1), 261-270.
Radhi, H., Sharples, S., & Assem, E. (2015). Impact of urban heat islands on the thermal comfort and cooling energy demand of artificial islands—A case study of AMWAJ Islands in Bahrain. Sustainable Cities and Society,19, 310-318. Ramachandra, T. V., & Uttam, K. (2009). Land surface temperature with land cover dynamics: multiresolution, spatio-temporal data analysis of Greater Bangalore. International Journal of Geoinformatics, 5(3). Ramachandra, T. V., Aithal, B. H., & Sanna, D. (2012). Land Surface Temperature Analysis in an Urbanising Landscape through Multi-Resolution Data. Research & Reviews: Journal of Space Science & Technology, 1(1), 1-10 Raynolds, M. K., Comiso, J. C., Walker, D. A., & Verbyla, D. (2008). Relationship between satellite-derived land surface temperatures, arctic vegetation types, and NDVI. Remote Sensing of Environment, 112(4), 1884-1894. Rizwan, A. M., Dennis, L. Y., & Chunho, L. I. U. (2008). A review on the generation, determination and mitigation of Urban Heat Island. Journal of Environmental Sciences, 20(1), 120-128. Rosenfield, G. H., & Fitzpatrick-Lins, K. (1986). A coefficient of agreement as a measure of thematic classification accuracy. Photogrammetric engineering and remote sensing, 52(2), 223-227. Serrano, L., Filella, I., & Penuelas, J. (2000). Remote sensing of biomass and yield of winter wheat under different nitrogen supplies. Crop Science, 40(3), 723-731. Snyder, W. C., Wan, Z., Zhang, Y., & Feng, Y. Z. (1998). Classification-based emissivity for land surface temperature measurement from space. International Journal of Remote Sensing, 19(14), 27532774. Song, C., Woodcock, C. E., Seto, K. C., Lenney, M. P., & Macomber, S. A. (2001). Classification and change detection using Landsat TM data: when and how to correct atmospheric effects? Remote sensing of Environment, 75(2), 230-244.
Stathopoulou, M., Cartalis, C., & Petrakis, M. (2007). Integrating Corine Land Cover data and Landsat TM for surface emissivity definition: application to the urban area of Athens, Greece. International Journal of Remote Sensing, 28(15), 3291-3304. Statistical Center of Iran. (2011). Yearbook of Iran, Tehran. Streutker, D. R. (2002). A remote sensing study of the urban heat island of Houston, Texas. International Journal of Remote Sensing, 23(13), 2595-2608. Su, W., Gu, C., & Yang, G. (2010). Assessing the impact of land use/land cover on urban heat island pattern in Nanjing City, China. Journal of Urban Planning and Development. Sun, D., & Kafatos, M. (2007). Note on the NDVI‐LST relationship and the use of temperature‐related drought indices over North America. Geophysical Research Letters, 34(24). Taha, H. (2015). Meteorological, air-quality, and emission-equivalence impacts of urban heat island control in California. Sustainable Cities and Society, 19, 207-221. Touchaei, A. G., & Wang, Y. (2015). Characterizing urban heat island in Montreal (Canada)—Effect of urban morphology. Sustainable Cities and Society, 19, 395–402. Valor, E., & Caselles, V. (1996). Mapping land surface emissivity from NDVI: Application to European, African, and South American areas. Remote sensing of Environment, 57(3), 167-184. Van de Griend, A. 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(6), 1119-1131. Weng, Q. (2009). Thermal infrared remote sensing for urban climate and environmental studies: Methods, applications, and trends. ISPRS Journal of Photogrammetry and Remote Sensing, 64(4), 335-344. Weng, Q., Lu, D., & Schubring, J. (2004). Estimation of land surface temperature–vegetation abundance relationship for urban heat island studies. Remote sensing of Environment, 89(4), 467-483.
Wong, M. S., & Nichol, J. E. (2013). Spatial variability of frontal area index and its relationship with urban heat island intensity. International journal of remote sensing, 34(3), 885-896. Yan, G., Mas, J. F., Maathuis, B. H. P., Xiangmin, Z., & Van Dijk, P. M. (2006). Comparison of pixel‐based and object‐oriented image classification approaches—a case study in a coal fire area, Wuda, Inner Mongolia, China. International Journal of Remote Sensing, 27(18), 4039-4055. Yang, J., Yu, Q., & Gong, P. (2008). Quantifying air pollution removal by green roofs in Chicago. Atmospheric environment, 42(31), 7266-7273. 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. Yue, W., Xu, J., Tan, W., & Xu, L. (2007). The relationship between land surface temperature and NDVI with remote sensing: application to Shanghai Landsat 7 ETM+ data. International Journal of Remote Sensing, 28(15), 3205-3226. Zinzi, M., & Agnoli, S. (2012). Cool and green roofs: An energy and comfort com-parison between passive cooling and mitigation urban heat island techniquesfor residential buildings in the Mediterranean region. Energy and Buildings, 55, 66–76.
Figure Captions
Figure 1. Situation of the study area, Tehran
Figure 2. Position of the synoptic weather stations in Tehran
Figure 3. Classified map of land surface temperature, July 2010
Figure 4. LULC map prepared from Landsat TM satellite image]
Figure 5. NDVI map of Tehran
Figure 6. Relationship between land surface temperature (LST) and vegetation index (NDVI)
Figure 7. Spatial distribution of the population in Tehran, 2011
Figure 8. Population of districts of Tehran, 2011
Figure 9. Landsat TM imagery of 07 August 2010 in 164 pass and 35 row
Tables Table 1. Error Matrix of the Landsat TM Image in 2010 Classes
Frost Park
Bare land
Green lands
Water bodies
Constructed areas
Asphalt
Total
User accuracy
Frost Park
254
0
8
0
0
7
269
94,42%
Bare land
0
217
0
0
4
0
221
98,19%
Green lands Water bodies Constructed areas
1
0
322
0
0
0
323
99,69%
0
0
0
111
0
0
111
100%
0
0
0
0
164
0
164
100%
Asphalt
0
0
0
0
0
73
73
100%
Total Producer accuracy
255
217
330
111
168
80
1161
-
99,60%
100%
97,57%
100%
97,61%
91,25%
-
-