This research examines the changes in land use/land cover in the city of Sulaymaniyah north of Iraq and identifies land surface temperature variations among the land cover types. The primary aim of this study is to use Landsat-5 TM imagery with GIS techniques to study and investigate the impact of urban expansion on land surface temperature (LST) for three year periods. Three Landsat 5 TM images were obtained in July 1984, August 2000 and October 2010. Land use categories were derived through the use of supervised classification techniques and the land surface temperature was obtained by computing the brightness temperature from the satellite sensor.
The result showed that between 1984 and 2010, there was a mild decrease in open and barren lands from 69.3% in 1984 to 57.2% in 2010 while the built-up areas increased from 11.5% in 1984 to 15.5% in 2000 and reached 25.5% by 2010.The political and economic changes in the study area are the main factors behind the recent urban expansion. The lowest LST readings were taken from the vegetation lands with values of 28oC in 1984, 29oC in 2000 and 34oC in 2010. The barren lands recorded the highest temperature of 38oC, 38oC and 34oC for the years 1984, 2000 and 2010 respectively. An interesting observation in this study is the fact that the urban areas where found to be cooler than its surroundings zones. This is revealed by the LST analysis conducted, with the recent increase in green spaces in the city playing a major role in cooling the temperature there. In relating NDVI to LST, the study found a strong negative correlation between them having derived correlation values of the values of (-0.70), (-0.69) and (-0.73) for 1984, 2000 and 2010 respectively. Conclusively, remote sensing and GIS proved to be very effective in studying and monitoring the relationship between urban growth and surface temperature. Recommendations were made to encourage the expansion of urban surfaces into the surrounding areas, especially barren lands, in order to cool those areas.
Contents
Abstract
Acknowledgement
List of Figure
List of table
Chapter One: Introduction
1.1: Introduction
1.2: Aim and Objective of the study
1.3: Research Questions
1.4: Problem Definition
1.5 Structure of the thesis
Chapter Two: Literature Review
2.1:The Role of Remote Sensing in Land surface Tempreture Retrieval and Application
2.2: Algorithms Developed for Retrieving LST
2.3: Related Literature on Landsat- based Land Surface Temperature Retrieval
Chapter Three: Data and Methodology
3.1: Study area
3.2: Data Collection
3.3:Research method
1.2: Aim and Objective of the Study
1.3: Research Questions
1.4: Problem Definition
1.5: Structure of the Thesis
3.1: Study area
3.2: Data Collection
3.3: Research method
3.5: Classification scheme
3.6: Image Classification Techniques
3.7: Accuracy Assessment
3.8: Normalized Difference Vegetation Index (NDVI)
Chapter four: Results and Discussion
4.1: Result of Classification Images
4.3: Land Surface Temperature Retrieval
4.4: Spatial Distribution of LST in Sulaymaniyah city
4.5: Spatial Distribution of LST in
4.6: LST Spatial Distribution in
5.1: Conclusion
Abstract
This research examines the changes in land use/land cover in the city of Sulaymaniyah north of Iraq and identifies land surface temperature variations among the land cover types. The primary aim of this study is to use Landsat-5 TM imagery with GIS techniques to study and investigate the impact of urban expansion on land surface temperature (LST) for three year periods. Three Landsat 5 TM images were obtained in July 1984, August 2000 and October 2010. Land use categories were derived through the use of supervised classification techniques and the land surface temperature was obtained by computing the brightness temperature from the satellite sensor.
The result showed that between 1984 and 2010, there was a mild decrease in open and barren lands from 69.3% in 1984 to 57.2% in 2010 while the built-up areas increased from 11.5% in 1984 to 15.5% in 2000 and reached 25.5% by 2010.The political and economic changes in the study area are the main factors behind the recent urban expansion. The lowest LST readings were taken from the vegetation lands with values of 28oC in 1984, 29oC in 2000 and 34oC in 2010. The barren lands recorded the highest temperature of 38oC, 38oC and 34oC for the years 1984, 2000 and 2010 respectively. An interesting observation in this study is the fact that the urban areas where found to be cooler than its surroundings zones. This is revealed by the LST analysis conducted, with the recent increase in green spaces in the city playing a major role in cooling the temperature there. In relating NDVI to LST, the study found a strong negative correlation between them having derived correlation values of the values of (-0.70), (-0.69) and (-0.73) for 1984, 2000 and 2010 respectively. Conclusively, remote sensing and GIS proved to be very effective in studying and monitoring the relationship between urban growth and surface temperature. Recommendations were made to encourage the expansion of urban surfaces into the surrounding areas, especially barren lands, in order to cool those areas.
Acknowledgments
I sincerely thank God for the completion of this research. My heartfelt thanksgose to my dear husband Sartip for his support and my daughters Naivan and Adan for their patience during my study.
I am greatly thankful to my supervisor (Andrew Johns) for his guidance and for taking the time to read the earlier draft of this thesis. I would like to thank all friends in Salahaddinuniversity for providing advice and support to me. I extend my sincere thanks to my professors and loyal friends in Sulaymaniyah University-College of Human Sciences, who have encouraged and supported me in getting the information and data that I needed. I would like to thank my colleagues in Sheffield Hallam university for providing help during my work. I would like to thank all family members for their continued encouragement and love through my life.
This study could not have been conducted and written without help from many people. I would like to thank all others who have contributed directly and indirectly to the success of this research.
List of Figures:
Figure 1 : Study area..
Figure 2 : Work flow for analysing Satellite Images.
Figure 3 : Layer Stacking of Landsat Images
Figure 4 : Subset Study area from Landsat Images
Figure 5 :PrincipalComponent Analysis
Figure 6 : Workflow of Converting Raster file formats into a Vector file in GIS.
Figure 7 : Land Cover Classes in 1984 .
Figure 8 : Land Cover Classes in 2000..
Figure 9 : Land Cover Classes in 2010..
Figure 10 : Percentages of changes in Land Cover classes in 1984, 2000 and 2010.
Figure 11 : Land Surface Temperature for 1984
Figure 12 : Land Surface Temperature for 2000
Figure 13 : Land Surface Temperature for 2010...
Figure 14 : LST Spatial Distribution in the Main Districts of the city of Sulaimanyah
on10-07-1984.Time:07:01:24 AM.
Figure 15 : LST Spatial Distribution in the Main Districts of the city of
Sulaimanyah on 02-10-2010. Time: 07:23:22 AM
Figure 16 : Change Detection Map based on LST from 1984 to 2011.
Figure 17 : Mean LST for Land Cover types in 1984,2000 and 2010.
Figure 18: NDVI Maps for 1984,2000 and 2010 .
Figure 19 : Outcome of Liner Regression analysis of NDVI and LST in 1984
Figure 20 : Outcome of Liner Regression analysis of NDVI and LST in 2000
Figure 21 : Outcome of Liner Regression analysis of NDVI and LST in 2010
Figure 22 : Differences in Mean LST and NDVI value for each Land Cover Class.
List of tables:
Table 1 : Landsat-5 TM Data.
Table 2 : Description of Land Cover Types..
Table 3 :Changes in Land Cover Classes of 1984 and 2010
Table 4 : Result of Classification Accuracy..
Table 5 : Mean LST of Main Districts in Sulaimanyah City on 10-07-1984
Table 6 : Mean LST of Main Districts in Sulaimanyah City on 02-10-2010
Table 7 : Values Collected from LST and NDVI Maps for 1984
Table 8 : Values Collected from LST and NDVI Maps for 2000
Table 9 : Values Collected from LST and NDVI Maps for 2010
Chapter one:Introduction:
1.1:Introduction
Rates of urbanization across the globe are rapid and it has put quality of physical environmental elements such as (temperature, rainfall, soil and water) at high risk. Thus, now every day, all over the world the subject of urban thermal environment attracts wide interests (Guo et al, 2012). Muhammadi et al, (2012) defined urban expansion as the extent of urbanization which mainly occurs by population growth and migration, and has an effect on the environment and natural resource.
Urbanization is an important topic for decision makers, planners and environmentalists (Jat et al, 2008). Urban expansion causes dramatic conversion in the areas around it. It leads to the loss of agricultural lands and their subsequent transformation to build up areas (Rimal 2012). In other words, the land surface is mostly covered by man-made features while the other natural land covers, including vegetation and water bodies, decreases. Urban growth leads to change in land surface characteristics including thermal capacity, surface albedo and the moisture of soil. This causes the differences in temperature between urban areas and the areas around it, as the surface materials are different between urban and suburban areas, such as asphalt, road pavements in the urban areas which highly absorb heat in days and emit it at nights; this is different in the suburban areas as it covered mostly by vegetation (Gallo et al, 1993) (Matthews, 2012). According to (Vernberg et al, 1996) urbanization affects the natural environment negatively; therefore having an effect on human activities and health directly and indirectly.
Urban Heat Islands (UHIs) are one of the main problems that result from urban growth. According to Stabler et al, (2005) currently, Urban Heat Island is one of the important issues on the environment in terms of its effect on land use and vegetation cover. Frazer (2005) reported that impervious surfaces have a negative relationship with the environment, as they have a greater thermal capacity than vegetated areas.
Myint and Okin, (2009), stated that urban features are dark and they lead to increase in temperature unlike soil or vegetation which create and enhance humidity. Occurring vegetation assist in cooling the air in its area as it converts solar radiation into the humidity in the evaporation process, which can reduce the temperature (Reducing Urban Heat Islands, 2009).
Land surface temperature (LST) is an important parameter which is involved in energy balance and the procedure of evapotranspiration. It entails energy fluctuations and interactions between the ground and the atmosphere (Alsultan et al, 2005). LST differs from the land surface air temperature (LSAT) because it entails heating and cooling processes of earth's surface and it is faster in LST similar approach in the air surface temperature. Valiente et al (2010) defines land surface air temperature as the air temperature near the earth's surface usually measured by meteorological stations. LST means the skin temperature of the land surface. Consequently, the process of land surface heating is very intricate and is affected by many factors including surface emission, the rate of moisture in the soil, type of the materials in the earth's surface and solar radiation (Sun, 2003 ) (Rinner and Hussain, 2011).
LST has adverse impacts on the environment and atmosphere. For instance, it involves in the process of rising land radiation and heat flux exchanges in the atmosphere (Alsultan, 2005). Thus, LST is a key parameter for estimating surface and atmosphere energy exchanges and for studying local, regional and global environment change (Lo and Quattrochi, 2003; Wan and Dozier, 1996; Mallicket al, 2008). LST is a source for acquiring information about different kinds of land surface as it has an important role in the dynamics of land surface processes. Consequently, LSTestimationis essential for diverse field of studies such as climate, ecology, hydrology, vegetation monitoring, estimating soil moisture and geology (Wan and Dozier, 1996; Tang et al, 2008).Satellite remote sensing is crucial for providing data at different spatial and temporal resolutions, which are necessary for many studies including climatology, geography, ecology and hydrology. Thermal infrared remote sensing (TIR) is an excellent tool in modelling of environmental elements. It enables the collection of data on physical parameters of land surface based on the amount energy reflected, radiated or emitted (Kiage et al, 2007; Dousset and Gourmelon, 2003; Albright et al, 2011). As a result, remote sensing has emerged as a valuable and widely accepted monitoring tool for meteorological stations.
LST measurements are increasingly becoming popular for modelling climate change (e.g. Global warming) and greenhouse effects (Dousset and Gourmelon, 2003). In addition, this has an implication for urban planning as unplanned urban areas are more likely to have a higher air surface temperature owing to the impact of LST, while well planned urban areas with wide range of green spaces are more likely to have a lower air surface temperature. (Kumar et al, 2012). Planners and decision makers are using the LST information for urban planning as it is the most important factor of controlling the urban climate.
In order to estimate thermal situations of land surface by satellite image it is necessary to explore the connection between different types of land cover and surface temperature. The most popular approach for this is the Normalized Difference Vegetation Index (NDVI) (Takemata et al, 2004). LST is important in visualizing the rate of vegetation in the area, as this has an effect on evapotranspiration and soil moisture of the surface (Ahmed et al, 2005). NDVI and LST are both used to identify the effect of vegetation and built-up lands on the urban heat island (UHI). This research was aimed at examining the effects of urban sprawl on the land surface temperature of the city of Sulaymaniyah, Iraq Recently, Sulaimanyah city has experienced a rapid urban expansion which has been documented in several literature works (Rasul, 1996). In addition, there has been no research conducted on the spatial effects of urban development on LST by utilizing remote sensing and GIS. Furthermore, in the last two decades, many environmental problems have occurred in the area, like sand storms in the summer period and droughts during winter (Sulaimanyah weather station’s data)..
It has become necessary to thoroughly analyze and observe environmental alterations in the urban areas so that suitable decisions can be made to improve declining environmental conditions in those areas. Conducting such research is necessary for understanding the causes and consequences of changes in the urban areas, especially using remote sensing and GIS techniques. Understanding the characteristics of LST distribution can also assist planners in discovering new ways of solving urban problems and in developing informed regional plans for the city of Sulimanyah.
1.2: Aim and Objective of the Study
The main purpose of this research was to examine the effects of urbanization on land surface temperature using both Geographic Information Systems (GIS) and remote sensing based techniques.
Research Objectives:
- To examine the spatial pattern of urban land use/cover change in the study area.
- To identify land surface temperature changes on each land use/cover type during the period of the study.
- To offer recommendations for further studies.
1.3: Research Questions
- During the period under study, what changes have occurred in the land cover classes?
- Is there a spatial variation in the Land surface temperature in Sulaimanyah city?
- What is the relationship between land cover (e.g. Vegetation and built-up area) and land surface temperature?
- To what extent has urban expansion affected land surface temperature over the period of the study?
1.4: Problem Definition
The urbanization trends occurring in the study area of Sulaymaniyah city show a rapid urban expansion. This phenomenon has been documented in literatures and non-formal resources, which have expounded on its adverse impacts on the ecology, consequently causing pollution, global warming and greenhouse gas effects.
Therefore planners and decision makers need reliable, current and accurate information on LST, in order to make suitable decisions to improve declining environmental conditions in the area. Conducting such research is imperative for understanding the causes and consequences of the changes in the urban areas, as no research on LST monitoring has been carried out in the study area, especially using remote sensing and GIS techniques. There is a need for this study and this research is a beginning for filling this knowledge gap. As a result, this study is important and unique. Particularly, because it is using remote sensing and GIS techniques.
1.5: Structure of the Thesis
This study organized into five main chapters. The first chapter outlines an overview of the study that covers research aims, research questions, significance of the study and problem definition. The second chapter reviews key issues in the literatures of urban temperature monitoring in both remote sensing and GIS based methods. In the third chapter, the study area is presented, data collection, study methods and the methodology used. The fourth chapter presents the results of the study and discussion of the research findings and the findings of other studies. The fourth chapter focuses on the study limitations; research contributes to the body of knowledge; recommendations and the conclusions.
Chapter Two: Literature Review:
2.1 The Role of Remote Sensing in Land Surface Temperature Retrieval and Application
Remote sensing is the means of acquiring information about a phenomenon or object without physically making contact with the object. (Lillesand and Kiefer 2004) agrees with this in their definition of remote sensing as the art of obtaining information about an object or phenomenon remotely. In other words, remote sensing involves obtaining information on any target object from a remote distance, and geographers find this information to be of important use. This explains why for issues relating to environmental changes, one of the most important tools used is remote sensing as it provides a great amount of data on the environment.
Remote sensing has been, and is still in great use in contemporary GIS analysis. Going from trends and research works found, remote sensing has proven to be more effective than traditional methods like field based surveys. It is curious to note that several literatures have highlighted the limitations of remote sensing. An argument against remote sensing was given by Carver et al, (1995) who advocated for the use of field based surveys in GIS analysis as an alternative to remote sensing, in the area of environmental classification, modelling and evaluation support. Osborne et al, (2001) and Raup et al, (2007) complained about the high cost of obtaining remotely sensed data Bastiaanssen et al, (2000) also pointed out that the low availability of experts is a bottleneck to the implementation of remote sensing technology. However, remote sensing has time and time again offered much more benefit than field surveys. In the area of environmental classification alone, Foody and Curran (1994) highlighted a number of uses including terrestrial global environment research, land cover variations monitoring, tropical forests regenerative states, snow monitoring, urban land cover monitoring and a host of other uses. Benefits found in its use in other areas like agriculture was given by Clevers (1997) where the time and quantitative manner in which information was provided was well acknowledged. Furthermore, in contemporary times, we have witnessed the continuous reduction in the cost of obtaining satellite data. In fact, Landsat ETM+, TM, and MSS data, made available from USGS, is now free (NASA 2013), this means that more researches today can be carried out in a cost effective manner. The governments of the world are making heavy investments in procuring satellite services, so that they can obtain remotely sensed data for use in national development. India (Indian Space Research Organization 2008), US, China, Russia to name a few, are among these heavy investors, and it is expected that remote sensing professionals will greatly increase in number in the near future to meet growing demands for them.
In the area of climatology, remote sensing offers great advantages over the use of traditional methods. Traditional methods, for instance, have been developed to compute and obtain urban air temperature, especially from mounted temperature sensors or observation stations based on land. In spite of these, limitations to these methods have been found (Mallick et al, 2008). In the first place, it is very expensive to use these traditional methods for large areas. It would require the construction of meteorological stations in large numbers, which is not really feasible. Due to the limited number of climate stations, representative meteorological conditions cannot always be retrieved from them (Southworth 2004). The speed in obtaining meteorological data is another advantage remote sensing has over traditional methods, and this data can be provided for large areas, at high resolution and for a very large number of time periods. Retrieval of urban climate with traditional methods is rather cumbersome, but remote sensing can obtain this data and generate accurate analysis in a short time, making it the preferable choice. Inaccessible regions (e.g. Many of the tropical forests) make it unfeasible to employ traditional micro-meteorological techniques to study forest canopy thermal budgets. Thermal remote sensing, in such locations, can serve as an excellent source of information (Luvall et al, 1990).
Land surface temperature (LST) is a very significant climate component used to detect the amount of sunlight energy reflection, which. It is connected with surface energy balance and the atmospheric integrated thermal state within the boundary layer of the earth (Jin and Dickinson 1996). Being the main parameter in land surface processes, LST acts as an indicator of climate change and is used to monitor sensible and latent heat flux exchanges within the atmosphere (Sun 2003). These surface heat dynamics cause local convection in the atmosphere and create alterations in air temperature, cloudiness, surface winds, and precipitation (Aires et al. 2001). Prata et al, (1995) regarded it as a primary variable to correctly represent the budget of surface energy. He added that it can be used to study the factors behind the variations in the two-way transfer of energy between the atmosphere and the earth's surface. This explains why LST is used globally and in a widespread manner.
Land surface temperature can be interpreted for a number of purposes. Earth's surface temperature daytime cycle estimates give rise to information about the moisture of the soil. A relationship has been found to exist between LST and the variability of rainfall in the North American monsoon (Matsui, Lakshmi and Small 2003). The skin temperature of the earth's surface, derived from LST, can be used to assess surface energy balance, monitor vegetation water stress, detect land surface disturbances and monitor conditions that are suitable for insect-vector disease proliferation, etc. (Pinheiro et al. 2006). The surface temperature varies with different types of land cover, thus making it useful in classifying land use in an area. It has also been used greatly in the study of changes over time in the land use and land cover of an area. Xiao and Weng, (2007) Used it in the study of Guizhou Province, south of China, and compared changes between 1991 and 2001. The result of their research showed a decrease in agricultural area and an increase in urban areas over that period of time. There is increasing interest in the study of urban climate change, due to the increasing urbanization in many parts of the world. Streutker, (2002) emphasized on this in his work where he used land surface temperature to identify urban heat islands, which are also known as high urban concentration areas. He further commented on the helpfulness of remotely sensed temperature data in the study of heat islands of urban areas, in the comparative analyses performance and the ease of obtaining such data, especially for global studies. Tran et al, (2006) also agreed on the LST's usefulness in studying urban climate and added that it aids in providing insight on the environmental conditions residents of an area are living in. They applied LST in the study of urban development of certain Asian cities.
For the purpose of obtaining LST, thermal sensors are predominantly used and the appropriate wave band for analysing LST is the satellite thermal band. It is suitable for LST retrieval for several reasons. Thermal infrared sensors are widely used in retrieving land surface temperature. This preference for thermal infrared data is due to its contribution to the enhanced knowledge of processes on the land surface, as explained by Sobrino et al, (2006); he gave two reasons for this. Firstly, it can measure the surface temperature of particular landscape and biophysical compositions. Secondly, comparisons can be made between surface temperatures and fluctuations in energy for particular landscape occur or processes can be identified. Various other land surface processes can be studied through the use of thermal data. It can also be used to monitor the form of energy exchange between the atmosphere and the land surface (Hurtado and Caselles, 1996) as well as in conducting detailed analysis of the changes in surface temperature of a landscape. Its use in determining land surface temperature has aided a lot of research, especially those focused on urban processes.
Several satellite sensors capture thermal data; among them are AVHRR, MODIS, and ASTER. Other satellite sensors like Quickbird-2, IKONOS-2 and Spot 4 and 5, despite their high spatial and temporal resolutions, do not have thermal bands and thus cannot provide data to map LST. Among all the current satellite sensors, other than Landsat, ASTER data have been said to have the highest thermal band resolution. In addition, ASTER offers a product of high level that provide temperature and reflectance values that can be used as direct inputs in LST studies (Wang et al. 2010). However, this data is not free and ASTER's temporal resolution period is 16 days like Landsat. Its LST value has also been found to be of limited accuracy (Liu, Hiyama and Yamaguchi 2006). Nevertheless, ASTER data has been found to more appropriate for research purposes than for purposes of operation, and it is used because of its extensive use and global applicability (Prata et al, 1995) e.g. in detecting urban heat islands (Peña 2009). Landsat thermal data, on the other hand, has been increasingly used successfully in classifying land cover (Mausel et al. 1993)(Wang et al. 2010). Of all the satellite sensors, Landsat 7 ETM+ has the highest thermal band resolution. Issues with Landsat use include the temporal resolution of 16 days which makes its image acquisition problematic and unsuitable for emergency situations. Problems with the scan line corrector have degraded the quality of the image captured so that only the middle 20km of the imagery is useful (Clark et al. 2003). Unlike ASTER, Landsat does not offer high level products like surface temperature values, which makes its users go through extra work in doing the computation. Despite these limitations, the capacity of Landsat data to sense different land cover types separates it from the other satellite sensors and makes it possible to study how the different land cover types relate to the temperature of the earth's surface. The outcomes derived will aid in assessing each land cover's role in impacting LST (Walawender and Hajto 2009). Landsat thermal data have been said to have significant advantages over other satellite sensors in terms of cost and data quality (Morse and Allen, 2008). In addition, Landsat data has been freely made available by NASA and can be accessed easily. The low cost of acquiring Landsat images makes it easier to start remote sensing research projects. In addition, multi temporal studies, which require data from distant past periods, can be carried out because Landsat has an extensive global archive dating back to 1972. Furthermore, initial processing of images like atmospheric corrections and ortho-rectifications have already been carried out by Landsat, so that its images can be used straight away (Scantherma 2013). Being the longest and oldest satellite sensor, it is popularly used in monitoring and detecting environmental changes. Sensitive within the range of 10.4-12.5um, the Landsat thematic mapper thermal band is assigned as TM band 6. In comparison to the reflective TM bands, its radiometric sensitivity is lower and its spatial resolution is coarser (120m x 120m) compared to the other bands that have a spatial resolution of 30m x 30m (Southworth 2004). Although the TM Band 6 is acquired at a resolution of 120 meters, products processed after the 25th of February 2010 will be re-sampled to 30-meter pixels (United States Geological Survey 2013).
2.2: Algorithms Developed for Retrieving LST
Several factors are considered when calculating LST. The amount of water in the soil considerably influences the reflectance value. A wet soil will record lesser temperature value than dry ones. This is supported in (van Leeuwen et al. 2011) where it observed that deforestation activities, which in turn causes desertification, is a factor behind increasing land surface temperature because aridification and land surface degradation causes the radiative properties of the surface to change. In other words, the dry bare soils record higher surface temperature than well vegetated surfaces. Weng (2002) also observed that heat exchanges were suppressed or lower in more vegetated areas, while it was considerably in areas with scanty vegetation such as urban areas (Oke, 1982). This has therefore established vegetation density as another important factor in LST measurements. The roughness of the land surface is another factor considered in LST measurements (weng, 2002). Surface roughness accounts for features on the land surface such as buildings that affect their reflectance values (Snyder, 2000). Accounting for land surface roughness properties is a major reason why corrections are made to brightness temperatures used for LST computation. Varying viewing angles of the satellite sensor has considerable effects on the radiance observed from space (wang and Dozier, 2001). An experiment in (Atitar et al. 2004) was conducted to measure surface temperature over urban areas from different angles. The results revealed variations of surface brightness temperature ranging between -5 and 7 degrees Celsius between nadir and off-nadir calculations according to the azimuth view angles. The highest emissivity occurs at nadir and decreases as the view zenith angle increases (Rasmussen, 2010). Air humidity is taken into account when atmospheric corrections are applied to satellite data. Atmospheric corrections involve removing the atmospheric effects and humidity is one of them. Cloud cover pose as hindrances in the observation of sea and land surfaces because with the presence of clouds over an area, only the temperatures at the top of the clouds are measured, not the land surface (Ackerman, 2005).
Increased use, in contemporary times, of Landsat data for the retrieval of land surface temperature has led to the creation of models cantered on the only thermal band (TM6) of Landsat sensors.
The Radiative transfer equation was developed for LST retrieval but the major problem with this model is that it requires in situ radio-sounding to be launched concurrently with satellite movements (Sobrino, 2003).
To resolve this problem of reliance of radio-sounding by radiative transfer equation, a mono-window algorithm was developed in Qin et al., (2001). It acquires the LST from the thermal band (TM6) of the thematic mapper sensor which is boarded on the Landsat 5 satellite. The parameters used in this algorithm include the at-sensor brightness temperature in Kelvin, the total atmospheric transmittance, the land surface emissivity and the effective mean atmospheric temperature in Kelvin (Copertini, 2007). The mono-window algorithm is founded on the brightness temperature obtained from satellite sensor, emissivity of the land surface and solar zenith angle. It is regarded as a more robust algorithm when retrieving LST because it accounts for other factors other than the brightness temperature.
In calculating LST using the mono-window algorithm, the brightness temperature is first calculated. Thereafter, the surface emissivity and incoming solar radiation are provided as required parameters to the algorithm (Lim et al 2012). The values for surface emissivity are taken from NDVI values. The advantage of using mono-window is that after the solar angle values and the surface emissivity values have been added to the model, there is a higher correlation between the retrieved land surface temperature and the brightness temperature (Alsultan, 2005). In a correlation and regression test done in (Lim et al 2012) between the predicted LST and the LST values computed through the mono-window algorithm, the relationship was found to be more positive when compared to other algorithms. Mia and Fujimitsu (2011) used this algorithm in his study which aimed to forge a relationship between land surface temperature, spectral emissivity and land covers, using Landsat ETM+ and ASTER images of Kuju volcano for 2000 and 2006. He observed a negative correlation between the retrieved LST values and NDVI values.
The single-channel algorithm was proposed in Jimenez-Munoz and Sobrino (2003) and it utilizes parameters such as the at-sensor radiance, at-sensor brightness temperature, atmospheric functions and the effective wavelength. In 2007, the RM-NN algorithm was presented in Mao et al, (2007) to retrieve LST from MODIS data using a combination of radiative transfer model and neural network algorithm. It has however been proven to be hard to obtain in situ ground calculation of LST using this method. Moreover, no literature has been found using this algorithm on Landsat data.
A major problem faced when retrieving land surface temperature from Landsat TM 5 thermal channel images is the need for atmospheric profile parameters. This is because Landsat TM does not have a robust set of thermal infrared bands that can be used to calculate atmospheric profile parameters (Zheng, 2006). When comparing mono-window algorithm and split window algorithms on MODIS data, (Zheng, 2006) argued that a major issue with using these retrieval algorithms is their requirement for atmospheric profile parameters and surface emissivity values, which are not directly provided in Landsat TM thermal channel images. Extra computation has to be done to obtain these parameters, thereby making the LST retrieval process difficult and strenuous.
Landsat TM has only one thermal band unlike other satellite sensors like MODIS that have more than one thermal band (Alsultan, 2005). Performing atmospheric correction will involve the use of complex algorithms. NASA designates raw Landsat data received at the ground station as 'Level 0', which are the raw telemetry from the satellite. For the data to be of use, it must be processed and corrected, usually to the standard of Level 1 (Pancroma 2011). However, Landsat data downloaded from the USGS Earth Explorer website has already been radio-metrically corrected to the Level 1 standard and is ready for use by end users.
Prominent studies have been carried out without using all these parameters. They utilize the land surface emissivity value representing the brightness temperature in Kelvin. This is informed by the knowledge that thermal electromagnetic energy is emitted by any object at a temperature that is above absolute zero (Kelvin). On the basis of this principle, Zhang et al (2006) presented a method that converts the signals obtained by the thermal sensors (TM/ETM+) into at-sensor radiance. Thereafter, using the thermal calibration constants provided by the Landsat Project Science Office, these radiance values are transformed to radiant surface temperature (National Aeronautics and Space Administration 2013). Subsequent literature works have employed the use of this method including Kwarten and Small (2005), Kumar et al., (2012), Adinna et al (2009) etc., where the brightness temperature derived was used to represent LST.
In summary, different studies have used various techniques and algorithms to retrieve LST, with the use of Landsat's thermal band.
2.3: Related Literature on Landsat-based Land Surface Temperature Retrieval
Several works have noted the occurrence of urban cool islands in various parts of the world, especially those regions that share similar climatic conditions with that of the area being studied. This phenomenon is unique to desert, arid and semi-arid environments. This is in contrast with the popular notion of urban heat islands found in many case studies urban areas.
Brightness temperature, it has also been observed, is the most popular means of retrieving land surface temperature, especially in studies that relate urban development with surface temperature. Landsat TM and ETM+ data are widely used in these studies, as they meet the requirements for land use/land cover change analysis. Furthermore, there is a general consensus that a negative correlation exists between land surface temperature and NDVI. This can be adduced to the role that the amount of biomass in an area plays in controlling surface temperature.
Xiao and Weng, (2007) examined the land use/land cover changes that have taken place in the southern China province of Guizhou from 1991 to 2001.Landsat TM scenes for 1994 and 2001 were used temperature the study used brighten temperature rat, to acquire the LST values. Results showed that between 1991 and 2001, the land surface temperature increased in the urban areas and this is because of the rapid urban expansion that has occurred in that period. In addition, the conversion of agricultural and forest land into built-up land contributed to the increase in LST and the changes in land use has widened the difference in temperature between the urban and surrounding areas. The study also shows the urban heat island effect in the urban areas of Guizhou and gave forest abundance as a factor influencing LST.
Bounoua et al, (2009) assessed the impact of urbanization on the balance of surface energy, water and carbon through the use of land maps and land surface models. The area studied is the city of Oran situated close to the Sahara desert in Algeria. Landsat ETM+ images for the year 2002 were acquired and supervised classification using texture-colour and radiometric channels were employed. The analysis done in this study showed that there is no great difference between the temperature in the urban areas and non-urban areas especially in rainy seasons, and remarked that urban expansion, when accompanied with an increase of the vegetated areas, may help to reduce the heat and carbon congestion and create a healthy urban environment. The study recommended the integration of comprehensive land use/land cover program with the urbanization processes in the city to prevent environmental damages.
Frey et al, (2006) presented the urban climate of two cities in dry and hot environment and show the possibility for cities to act as daily urban cool islands in contrast with the urban heat island phenomenon often experienced in cities in temperate zones. The two cities studied are Abu Dhabi and Dubai situated in the northwest coast of the United Arab Emirates. Four ASTER satellite images, two for each city, were analysed with regards to the differences in net radiation, emissivity, albedo and surface temperature. Two classes were constructed for the land classification. One represents all urban features (e.g. Industries, parks, city etc.) and the other for rural features (e.g. Agriculture, sand etc..). Results showed that in Dubai, the urban areas had lower temperature than the rural areas because of the large surrounding areas of sand and desert features, while Abu Dhabi showed very little difference between the two areas when compared. The analysis proved that, indeed, these cities have the daily urban surface cooling island effect.
Xian and Crane, (2006) found similar characteristics in the valley of Las Vegas, south of Nevada. Utilizing information from the systems of Landsat 5 and 7, the thermal characteristics of Tampa Bay, Florida and Las Vegas, Nevada were assessed and results showed that the Las Vegas urban surfaces had a daytime cooling effect, also known as an urban cool island or a heat-sink.
Habib, (2007) conducted a similar study in Dammam, a city in Saudi Arabia. Landsat 5 TM images were utilized. As a result of the great amount of greenery in the central business district of Dammam, the study found out a lower LST there than the surrounding districts. Two major points were arrived at from this study. The first is LST can be reduced and controlled when urban areas are well planned. Secondly, people were recommended to keep constructing buildings in desert areas as well as increase the amount of green space in order to reduce LST.
Kwarteng and Small, (2005) analysed the distribution of surface temperature in the cities of New York and Kuwait. New York has a temperate climate while Kuwait is situated in a desert environment. Landsat TM and ETM+ images of Kuwait City and New York City were both acquired in 2001. Brightness temperature was used in deriving the LST. Due to the lack of substantial vegetation in the surrounding areas of Kuwait City, the urban areas there recorded a lower LST than the surrounding desert areas. In this manner, generally all the land cover in the encompassing zones conveyed higher surface temperature because of the absence of evaporation which prompts change from heat with humidity. Interestingly, the LST was higher in the urban areas of New York City than in encompassing zones as an after effect of the vegetation thickness which surrounded the city, and partook in cooling the air through the process of evaporation of solar radiation. Summarily, the analysis revealed that there is a difference in surface properties and energy flux in the city of Kuwait and New York City which are situated in desert and temperate environments respectively.
Saleh, (2011) analysed and verified the spatial distribution of surface temperature with regards to the land use characteristics in urban areas. The case study area is Baghdad, which is the capital city of Iraq. Landsat 7 ETM+ thermal images were acquired in 2002. Brightness temperature was also used for LST derivation. NDVI values were obtained for the purpose of investigating the relationship between NDVI and surface radiance temperature for each land cover type and the results showed a negative correlation. Unsupervised and supervised classification methods were utilized to create 8 land use classes. The commercial and residential areas, which make up the major part of the city, were found to have high LST, when considering the high surface distribution in the city coupled with a very low vegetation cover. It can be deduced that the reduction of vegetation contributed immensely to the increase in LST in Baghdad city. Surface temperature is shown to rise as a result of urban land developments. The results showed that there was a corresponding rise in surface temperature with urban expansion.
(Kumar, Bhaskar and Padmakumari 2012), investigated the impact that land use types have on the increasing surface temperature in Vijayawada city, India and identify which of them had the most influence. Landsat ETM+ images of 2001 were obtained for the study area and LST was calculated using brightness temperature and the study used supervised classification met. To generate land use maps. It was observed that the highest temperature existed in urban areas and the lowest temperatures were found in the surrounding vegetation areas, clearly demonstrating the urban heat island effect. The study also revealed a strong negative correlation between LST and NDVI.
Weng, (2001) applied remote sensing and GIS techniques to detect urban expansion in the Zhujiang Delta region of South China and assess its impact on land surface temperature. Multi temporal images from Landsat Thematic Mapper for the years 1989 and 1997 were used, while brightness temperature method was used to derive LST. The results showed a considerable increase in urban land area within the periods studied and that these urban areas had the highest surface temperature, followed by the surrounding barren lands. The study also reported a strong negative correlation between surface temperature and NDVI, implying that the greater the land cover's biomasses are, the lower the surface temperature will be.
Through a historical examination of differences in urban and rural air temperature, Weng and Yang (2004) analysed the role of urban development in creating urban heat islands. The study area is the city of Guangzhou city in South China which has gone through rapid urban development since the 1980s. Land surface temperatures were obtained through brightness temperature calculations on Landsat TM images for 1989 and 1997. The expansion of urban land cover made a direct impact on the patterns of surface temperature and heat island intensity dynamics as shown from the result of the study. A positive relationship was found between the continual increase in air temperature and urban development in the 1980s and 1990s. Recommendations were made for taller trees to be planted in the city, having considered the cooling effects of trees for temperature control.
From the foregoing literature works, a certain trend in urban surface temperature analysis has been observed. Cities are either categorized as having either an urban heat island effect or an urban cool island effect. Landsat thermal data is generally used for computing the land surface temperature in the case study areas. NDVI values are derived from the images to test for correlation with LST. Supervised classification is the preferred method for deriving land use classes in the studies. Brightness temperature is also found to be popular among the literature studied, in contrast to other algorithms such as mono-window, split-window, etc.
It is important to relate the outcomes from these literature works to the research questions asked in this study. Each of these studies monitored the spatial variation of land surface temperature over their study areas. In conducting their analysis, they classified their study areas into different land uses/land cover and thereafter took LST readings from each of the land cover areas to test the relationship between land cover and temperature. They basically sought to understand the impact that urban expansion has made in cooling or heating up the surface temperature, hence the creation of the Urban Heat Islands and Urban Cool Islands phenomenon. Urbanization is seen by these authors as a critical factor in LST studies, and a common relationship is made between urbanization and LST increase/decrease. Habib(2007) drew attention to LST-impacting environmental factors that are similar to those found in the study area, and it remains to be seen from the analysis if the results would be similar.
Chapter Three:Data and Methodology
3.1: Study area
The study area is a Sulaymaniyah city in Kurdistan region and it is located in the northern part of Iraq. It is situated between latitudes (35ᵒ 37′ 36″, 35ᵒ 28′ 08″ N) and Longitudes (45ᵒ 17′ 57″, 45ᵒ 29′ 56″ E). It is bordered by Iran to the northeast and to the west by Erbil and Kirkuk Governorate as displayed in figure. Sulaymaniyah governorate has a total land mass of about 11789 Km2 (Fathulah, 2000).
The study area is characterized by a Mediterranean climate. According to Koppen classification it is type CSA*, which characterized by warm and dry summer, wet and cold winter. There are two short seasons (spring and autumn) between winter and summer with moderate climate conditions and a maximum temperature 23ᵒC. Summer lasts from the beginning of May to the beginning of October and winter extends from November until the start of March. In winter, especially in January, the temperature decreases to 5ᵒC or even less than -1ᵒC.The highest temperatures, ranging between from 45ᵒC to 48ᵒC, are usually recorded between June and September. Much of the precipitation occurs from November until April. Generally, the annual precipitation of Sulaimanyah city is (648.6 ml). However, this has decreased in recent times due to climatic and environmental changes (Sulaimanyah weather station’s data).
A survey carried out by the Sulaymaniyah Directorate of Statistics in 2002 shows a population of 1,704,740 people with annual growth rates of 3 per cent (Sulaymaniyah Province, 2008). In 1987, 63 per cent of its population lived in the urban centers while 37 percent lived in the country sides. In the year 2008, urban population increased by 78 per cent while the rural population dropped to 22 per cent and the rates of urbanization was 3 per cent (op cit). This indicates has experienced a rapid rate of rural-urban influx in the last two decades.
*CSA"A climate where the coldest month is warmer than -3°C but colder than +18°C and summers are dry and hot. This climate is usually found inland on the western sides of continents". (Xpeditionconsulting.Koppen- Geiger Climate classificatio n.(Online). http://xpeditiononline.com/datavis/koppenguide.pdf
illustration not visible in this excerpt
Figure (1): study area (Wikimedia Commons: https://upload.wikimedia.org/wikipedia/commons/8/83/As-Sulaymaniyyah_in_Iraq.svg. CC-BY-SA 3.0)
3.2: Data Collection
The study analyses secondary data to identify the effects of urbanization on LST of the study area. Landsat TM images were obtained from glovis (USGS) http://glovis.usgs.gov/ for the 10 July 1984, 23 August 2000 and 02 October 2010 with a spatial resolution 30m. Two Landsat TM images (path 168, row 035) were obtained from theglovis Web site and used as a primary data. All of its bands were used for this study, and in particular Band 6 (thermal band) which is the most appropriate for detecting LST (table 1). The master plan of Sulaimanyah city was obtained from google in order to examine urban growth and data on weather was obtained from the Sulaymaniyah meteorological station in order to have background knowledge of the climate of the study area.
illustration not visible in this excerpt
Table 1: Landsat5TM Data
3.3: Research method
The software used for the study were Arcmap 10, To analyse satellite images, evaluate the results and create maps, several software packages were used including ENVI 4.7, ERDAS imagines 9.1 and ArcGIS 10. In addition, for the purpose of conducting statistical and regression analysis and creating charts and graphs, Microsoft Excel and SPSS were used.
In this study, the satellite image analysis was performed using two approaches: the first approach involves the use of supervised image classification techniques and the extraction of NDVI values with the aim of obtaining LST from the images. The second method converts and inputs the raster files into the GIS environment for easy calculation and the numerical results are manipulated through attribute table functions in ArcGIS software. The general workflow for analysing the satellite images is shown in figure (2).
illustration not visible in this excerpt
3.4: Image Pre-processing
As can be deduced from the workflow diagram above, three steps were taken before classification was done on the images, namely, layer stacking, Image subseting, and image enhancement. The Layer stacking step involves creating the color images. According to Horning, (2009) each band has specific wavelengths which are alike to black and white digital photograph in a separate image. For producing the colour image, the bands from different wavelengths should be combined together (Figure 3). The standard False Colour Composite (FCC) combination was used utilizing three bands, namely Band 4-Red,Band3-Green and Band 2 -Near Infra-red. In supervised classification procedures utilizing Landsat TM images, a combination of the bands is very suitable and useful (Lillesand et al 2008); (Saleh 2011). The next step is to subset the area of interest (Sulaymaniyah city), as a normal Landsat image that is downloaded from the website is usually larger in size. In this study, the study site subset is in Longitudes (45ᵒ 17′ 57″, 45ᵒ 29′ 56″ E) and latitudes (35ᵒ 37′ 36″, 35ᵒ 28′ 08″ N) and the area is about (122.76) sq. (see figure 4).
Principal component analysis is one of the techniques of spectral enhancement. Principal component analysis was applied to satellite images which enabled the spectral enhancement particularly for water and built up areas as they have a similar reflectance which made extracting the land cover classes on the study site easy to distinguish. Henry, (1991) stated that the spectral discrimination between reflectance of different features increased byprincipal component analysis. The three spectrally enhanced images were used for the classification and re-classification of the land use and land cover (see figure 5).
illustration not visible in this excerpt
Figure: (3) Outcome of Layer stacking for Land sat SatelliteImage Source: http://glovis.usgs.gov/
illustration not visible in this excerpt
Figure 4:Subset Images for selecting Study area on Satellite Image. Source: http://glovis.usgs.gov/
illustration not visible in this excerpt
Figure 5: Outcome of Principal Component Analysis. Source: http://glovis.usgs.gov/
3.5: Classification scheme
The aim of this study is estimating the effect of urban growth on LST. Consequently the study area is divided into four main land cover classes namely, built up, open land, barren land and vegetation area. Table (2) contains a description each class. In support of the objectives of this study, three images of 1984, 2000 and 2010 were used.
illustration not visible in this excerpt
Table 2: Description of Land Cover Types
The selection of the land cover types above was influenced by the objectives of this study: which is to clarify the relationship between land cover classes and LST, and detect the effect of urban expansion on LST. In addition, the researcher has spent her whole life in this city and has sound knowledge about the area.
3.6: Image Classification Techniques
The study relies on results from applying image classification methods. Remote sensing image classification techniques have become the subject of the majority of researches as it is an accurate, efficient and reliable way for getting information on land surface features when applied to a variety of fields (Lu and Weng, 2006). Also image classification enables to perform the valuable land data investigation and extraction. In other word classification is done in remote sensing to represent earth surface features in digital forms (Rashed and Jugens, 2010). Image classification assigns pixel images to predefined land cover categories. Using Image classification for extracting information is a complex procedure; therefore factors to be taken into account when carrying out this process include the level of image resolution, software/algorithm selection, classification technique selection, appropriate number of training samples, and knowledge/skills possessed by the analyst. Digital image classification is a common method for analysing remote sensing images (Matinfar, 2007). Sorting the pixels in the images is carried out for obtaining meaningful information of reality and using this information for creating thematic maps to represent the various land cover classes.
Remote sensing contains different image classification techniques and their suitability depends on the type and purpose of the land cover map used in a study.; (Lu and Weng, 2007) pointed out that efficiently employing multi-source remote sensing data and selecting suitable classification techniques is helpful in minimizing the errors in classification and improving the accuracy of classification.
Image classification at pixel level might be supervised or unsupervised (Levin, 1999) (Adam, 2010). Supervised classification was used for classifying images to attain the objectives of this study. Generally, researchers prefer supervised classification because it provides a more accurate definition of classes and better classification accuracy than unsupervised classification. The most widely used approach of classification is pixel-based image classification which classifies according to the spatial arrangement of the characteristics of the edge of the local neighborhood (J. IM, et al, 2008)
The supervised classification based on pixel requires the selection of training data for every predefined land cover class. Besides, ground knowledge of the field can assist in obtaining enhanced classification. Three stages should be passed before applying supervised classification including training data, choosing the appropriate classifier type and accuracy assessment (Barandela and Juarez, 2002).
Training data is the first step in this study. For each image, 100 samples are selected and 25 samples were given to each land cover class. The data should contain all classes and it should collect from homogenous areas, because it involves field survey and reference data from different sources. Hence, in selecting a training strategy many factors should be considered including number of pixels used for training, size of the training data, the impact of spatial autocorrelation, differences in the image, time and cost (Saha et al, 2005) (Foody and Mathur, 2002).
Maximum likelihood was used for classified images. The assumption made in maximum likelihood classification, being the most commonly used supervised classification technique, is that each spectral class can be described by a multivariate normal distribution. This technique leverages on the mean vectors and the multivariate spreads of each class and spots classes that are elongated. However, a reasonably accurate computation of the mean vector and the covariance matrix for each spectral class is required to make the maximum likelihood classification effective (Huang et al., 2008).
3.7: Accuracy Assessment
The aim of accuracy assessment is to complete the classification process. It can determine the value of the outcome data and it is essential for processing and analyzing remote sensing data (Hashemian and Fatemi, 2004). A variety of methods have been developed for assessing the accuracy of the land cover maps that is extracted from satellite data, for it becomes an important subject in the remote sensing area (Congalton, 1991; Latifovic and Olthof,2004).(Congalton and Green, 2009) states that positional and thematic assessments are the two main approaches for accessing the accuracy of remote sensing data. Positional accuracy assessment is a measure of the spatial distribution of features on the ground and on the map to discover the levels of distribution of features in reality. In contrast, thematic assessment assigns attributes to the features on the map, and deals with the level of reality of the features in the classified image or the maps. The study uses error matrix because it is the most widely used approach which makes comparisons between classified images and referenced data and it can be carried out by selecting random points in order to test the accuracy assessment (Hashemian and Fatemi, 2004). In addition, for representing the differences among changes in the agreement and the reality, Kappa coefficient was calculated. In the current study, for each image, 100 random points were selected and created in the ERDAS 9.1 for accuracy assessment.
3.8: Normalized Difference Vegetation Index (NDVI)
In LST studies, NDVI is a suitable parameter and it is the most commonly used approach for monitoring vegetation situation, as it is less sensitive than other indices to the changes in atmospheric conditions (Tso and Mather, 2009). In this study the aim of utilizing NDVI was to assess the correlation between NDVI and LST. The basis of NDVI involves calculating two important bands from Landsat-5 TM, which are near infrared (NIR) and red (bands 4 and 3) (LiuandZhang, 2011; Gao, 2009). Band 3 and 4 are used as the green or healthy vegetation, with band 4 (NIR) having a reflectance value of 60% in the wavelength between (0.7-1.3µm), while band 3 (red band) can reflect only 20%in the wavelength between (0.5-0.7µm).According to (Kidwell, 1990) the equation for extracting accurate NDVI values, using Landsat-5 TM bands, is shown below:
[...]
-
-
-
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X.