Evaluation of Spatial and Temporal Distribution Changes of LST Using Landsat Images (Case Study:Tehran)

In traditional approach, the land surface temperature (LST) is estimated by the permanent or portable ground-based weather stations. Due to the lack of adequate distribution of weather stations, a uniform LST could not be achieved. Todays, With the development of remote sensing from space, satellite data offer the only possibility for measuring LST over the entire globe with sufficiently high temporal resolution and with complete spatially averaged rather than point values. the remote sensing imageries with relatively high spatial and temporal resolution are used as suitable tools to uniformly LST estimation. Time series, generated by remote sensed LST, provide a rich spatial-temporal infrastructure for heat island’s analysis. in this paper, a time series was generated by Landsat8 and Landsat7 satellite images to analysis the changes in the spatial and temporal distribution of the Tehran’s LST. In this process, The Normalized Difference Vegetation Index (NDVI) threshold method was applied to extract the LST; then the changes in spatial and temporal distribution of LST over the period 1999 to 2014 were evaluated by the statistical analysis. Finally, the achieved results show the very low temperature regions and the middle temperature regions were reduced by the rate of 0.54% and 5.67% respectively. On the other hand, the high temperature and the very high temperature regions were increased by 3.68% and 0.38% respectively. These results indicate an incremental procedure on the distribution of the hot regions in Tehran in this period. To quantitatively compare urban heat islands (UHI), an index called Urban Heat Island Ratio Index(URI) was calculated. It can reveal the intensity of the UHI within the urban area. The calculation of the index was based on the ratio of UHI area to urban area. The greater the index, the more intense the UHI was. Eventually, Considering URI between 1999 and 2014, an increasing about 0.03 was shown. The reasons responsible for the changes in spatio-temporal characteristics of the LST were the sharp increase in impervious surfaces, increased use of fossil fuels and greening policies. * Corresponding author


INTRODUCTION
Nowadays, temperatures of the Metropolitan centers are increased due to several reasons (e.g.vegetation suppression, industrial life, population growth and etc) (Rose and Devadas, 2009).In General, urban heat islands (UHI), as an important indicator of the urban development, are defined as the urban regions which have the rather high temperature with respect to the surrounding (Chen et al, 2009).One of the main tools of the heat island's analysis is the estimation of the land surface temperature (LST) (Xu et al, 2009).The LST is one of the most important parameters in the physical processes of surface energy and water balance at local through global scales (Amiri et al, 2009).Knowledge of the LST provides information on the temporal and spatial variations of the surface equilibrium state and is of fundamental importance in many applications.As such, the LST is widely used in a variety of fields including evapotranspiration, climate change, hydrological cycle, vegetation monitoring, urban climate, environmental studies and etc.In urban areas, it helps to learn the local climate and the development of a city (Dai et al, 2010).The total area of districts with higher temperature changes year by year as the city develops.It is therefore significant to analyze the temporal and regional changes of land surface temperature in urban areas.In order to analyze and Evaluation of Spatial and Temporal Distribution Changes of LST ,there are two major approaches: (a) a method based on field observations of meteorological stations and (b) method based on the use of remote sensing technology (Malekpour et al, 2010).
Due to the lack of adequate distribution of weather stations, a uniform LST could not be achieved.Todays, With the development of remote sensing from space, satellite data offer the only possibility for measuring LST over the entire globe with sufficiently high temporal resolution and with complete spatially averaged rather than point values.the remote sensing imageries with relatively high spatial and temporal resolution are used as suitable tools to uniformly LST estimation (Li, 2006).
Study of urban heat islands by remote sensing done the first time by using NOAA satellite data (Balling and Brazell, 1988).Spatial resolution of NOAA satellite proper for providing small scale urban temperature maps.Then the thermal infrared data of Landsat satellite series with spatial resolution of 120 and 60 meters was applied to the surface temperature extraction (Weng, 2001).Xiao and Moddy (2005), investigated the relationship between land-use and land cover changes using Landsat TM and ETM images.Evaluation of urban development effects on the heat islands of the city of Guangzou in China by Weng and Yang (2004), using Landsat images showed that the expansion of construction and urban heat islands enhances the effect of heat islands.In the case of the heat island phenomenon, Pongracz et al. (2006), examined this phenomenon in different areas in the 10 densely populated city of Hungarian by MODIS images.The results of this study showed that with the increase of the population, the temperature difference city with the suburbs (UHI) has also increased.Gaffin et al. (2008), studied the heat island phenomenon and provided the required data from the weather stations inside and around the city.The results represent the temperature increase during the 12-year interval.Percentage of increase of temperature-related to changes of the regional and global climate and other percentage of the increase of the temperature of the city center are related to UHI.Numerous studies and research in related the Urban heat island has been performed (Malekpour et al, 2010).The results of this researches show that the effect of heat island has been more pronounced and consistent with the growth and development of the metropolis of Tehran.Heat island phenomenon can be studied from different aspects that the Earth's surface temperature changes caused by the existence of this phenomenon is one of the most important, therefore, the main objective of the present study is to investigate and evaluate of Spatial and Temporal Distribution Changes of land surface temprature in the metropolis of Tehran using Landsat satellite images.

Case Study Area
Tehran, the capital of the Islamic Republic of Iran, modern urban, developing and growing urbanization and dating back tens of years, has a high population density and large area compared to the other cities.Tehran city geographically at 51 degrees and 17 minutes to 51 degrees and 33 minutes East and 35 degrees and 36 minutes to 35 degrees and 46 minutes north.This city is located in the South of the Alborz Mountains and the northern edge of the Central deserts of Iran in the relatively smoothly plain.It has an area of about 706 square kilometers and its average elevation is 1600 m above sea level.

Data
In the present study were used two satellite image corresponds to sensors of the Landsat7 (ETM) and Landsat8 (OLI_TIRS) which features of satellite images were described in table (1).It should be noted that the aforementioned satellite images were downloaded from the site http://earthexplorer.usgs.gov.

LST retrieval from ETM data
ETM 6-2 band was used to retrieval the land surface temperature.this band is very suitable for the detection of the temperature difference of the urban environment.In the following present the steps of retrieval the land surface temperature of ETM sensor.To extract the land surface temperature, the following three steps will be performed: a) convert digital number of the band6 to spectral radiance with using equation (1) (Landsat Data User Hanbook, 2006): Where BT= Effective at-sensor brightness temperature [K] K2= Calibration constant 2 [K] is equivalent to the (1282.71)K1= Calibration constant 1 [W/(m2 sr μm)] is equivalent to the (666.09)L= Spectral radiance at the sensor's aperture [W/(m2 sr μm)] Ln = Natural logarithm c) In the third stage, the emissivity and the temperature of the Earth's surface were calculated with using the method based on NDVI.
One of the effective methods to estimate the temperature of the land surface is using of the hybrid model of the percentage of land cover, with the assumption that the soil and vegetation have a known emissivity, so with using the NDVI index, it can be calculated the amount of the compound and the incorporation of soil, vegetation and ...And with using this combination, it can be Obtained the emissivity and the temperature of the land surface (Sobrino et al., 2001).Therefore, one of the operational and functional ways to get emissivity, using the NDVI threshold method, which is based on it emissivity is divided to three level based on NDVI values.1. NDVI < 0.2 : In this case the pixels corresponding to soil dry and the mean value of emissivity is estimated to be 0.978( ). 3. 0.2 ≤ NDVI ≥ 0.5 : In which case the pixels of a combination of soil and vegetation and the emissivity in this case can be calculated on the basis of the equation (3): Where,  is the Land surface emissivity, v  is the emissivity of full vegetation cover area, while s  is emissivity of bare soil and Pv is the vegetation fraction, which was determined by (4) (Sobrino et al., 2004): ).In respect of (3),  d the effect of geometric distribution of natural surfaces, which is an approximate basis using the equation ( 5): Where F is the shape factor, (Sobrino et al., 1990) that the mean value with the assumption of different geometric distribution of levels , is equal 0.55.Getting the values of the emissivity, land surface temperature (LST) through the equation ( 6) calculated based on Kelvin (Artis and Carnahan, 1982): Where  is the emitted radiance wavelength,

LST retrieval from (OLI-TIRS) data
To retrieval the temperature of the Earth's surface, was used the band10 of OLI_TIRS sensor.To extract the LST, such as section (2.3), three steps take place: a) convert digital number of the band10 to spectral radiance with using equation (8): Where cal Q is the digital number of every pixel, Gain is equal to 0.0003342 and bias is 0.1, this information can be obtained from the header file of the images.b) Convert spectral radiance to the Brightness temperatures(BT) based on kelvin with using equation (2).Where BT= Effective at-sensor brightness temperature [K] K2= Calibration constant 2 of the band10 [K] is equivalent to the (1321.08)K1= Calibration constant 1 of the band10 [W/(m2 sr μm)] is equivalent to the (774.89)L= Spectral radiance at the sensor's aperture [W/(m2 sr μm)] c) In the third stage, the emissivity and the LST were calculated such as section (2.3).

Evaluation of Spatial and Temporal Distribution of LST
To study and Evaluation of Spatial and Temporal Distribution changes of LST in the metropolis of Tehran, we divided the LST into 5 categories using Mean-Standard deviation Method (Xu et al., 2011) Table 2. Usage of mean-standard deviation method to devided land surface temperature In the table 2, LST represent the land surface temperature,  and std, respectively represent the average and standard deviation of surface temperature of any images.
To quantitatively compare urban heat islands (UHI), in Tehran metropolis, between the years 1999 to 2014, an index called Urban Heat Island Ratio Index (URI) was calculated (equation ( 9)) (Xu and Chen, 2004).It can reveal the intensity of the UHI within the urban area.The calculation of the index was based on the ratio of UHI area to urban area.
In which, m is the number of temperature Classes , n is the number of temperaturer classes higher than the mean temperature.i w is the amount of weight of temperature classes higher than the mean temperature that means the amount of weight the of classes 4 and 5, which the amount of weight can be considered respectively as 4 and 5. i P is the Percent of occupied area by the temperature classes of higher than mean temperature.

RESULTS AND DISCUSSION
Using the equations expressed in sections (2.3) and (2.4), the maps of LST of Tehran were produced during the 1999 and 2014 (Figure 2 & 3).  the middle temperature regions, covered the extent area of the study area and particularly is centralized in city center.By Visual comparison of figures 4 and 5, obviously can be understand that some changes take place within a relatively cool, cool and middle regions during the passing of time, especially in the northern and Western zones, that seems to be the main cause of this type of transformation, changes of land use/land cover.Table 3. the percentage of the area of each of the classes of surface temperature in Tehran metropolis.
In order to Evaluation of temporal distribution changes of the LST of tehran, was investigated the trend of temporal changes for every classes, separately.
According to the table 3, the achieved results show (between 1999 to 2014) the verey low temperature regions and the middle temperature regions were reduced by the rate of 0.54% and 5.67% respectively.On the other hand, the high temperature and the very high temperature regions were increased by 3.68% and 0.38% respectively.These results indicate an incremental procedure on the distribution of the hot regions in Tehran in this period.In the following in order to investigate the temporal change of heat island intensity of Tehran metropolis, between the years 1999 to 2014, the Urban heat island ratio index (URI) was applied.The index (URI) based on the equation ( 9) for 1999 is equal 0.11 and for 2014 is estimated 0.14.The greater the index, the more intense the UHI was.Eventually, Considering URI between 1999 and 2014, an increasing about 0.03 was shown.

CONCLUSIONS
in this paper, a time series was generated by Landsat8 and Landsat7 satellite images to analysis the changes in the spatial and temporal distribution of the Tehran's LST.In this process, The Normalized Difference Vegetation Index (NDVI) threshold method was applied to extract the LST; then the changes in spatial and temporal distribution of LST over the period 1999 to 2014 were evaluated by the statistical analysis.Finally, the achieved results show the very low temperature regions and the middle temperature regions were reduced by the rate of 0.54% and 5.67% respectively.On the other hand, the high temperature and the very high temperature regions were increased by 3.68% and 0.38% respectively.These results indicate an incremental procedure on the distribution of the hot regions in Tehran in this period.To quantitatively compare urban heat islands (UHI), an index called Urban Heat Island Ratio Index(URI) was calculated.It can reveal the intensity of the UHI within the urban area.The calculation of the index was based on the ratio of UHI area to urban area.The greater the index, the more intense the UHI was.Eventually, Considering URI between 1999 and 2014, an increasing about 0.03 was shown.The reasons responsible for the changes in spatiotemporal characteristics of the LST were the sharp increase in impervious surfaces, increased use of fossil fuels and greening policies.
Figure (1), shows position of the study area.

Figure 1 .
Figure 1.Location of case study area (22 zones of Tehran) radiance at the sensor's aperture [W/(m2 sr μm)] cal Q = Quantized calibrated pixel value [DN] at-sensor radiance that is scaled to min cal Q [W/(m2 sr μm)]  LMAX = Spectral at-sensor radiance that is scaled to max cal Q [W/(m2 srμm)] b) Convert spectral radiance to the Brightness temperatures(BT) based on kelvin with using equation (2):

5 :
That represents the domain with abundant vegetation and the mean value of emissivity is estimated to be 0 emissivity.In the final stage, the equation (7) was used to convert kelvin to centigrade.

Figure 4 .Figure 5 .
Figure 4.The map of Spatial Distribution Changes LST of the Tehran metropolis (1 August 1999)

Table 1 .
characteristics of satellite images used in the analysis The class of high temperature is expanded, in the West and South-west regions and East and South-East regions.By comparing the figures 4 and 5, is clearly evident the development of the high temperature class and create very high temperature class in the range to the West, South West, East and South East.

Table 3 ,
shows the percentage of the area of each of the classes of surface temperature in Tehran metropolis.