1 UTILIZING GEOGRAPHIC INFORMATION SYSTEM S AND R EMOTE SENSING TO INVESTIGATE URBANIZATION PROCESSES: IN BOTH THE US AND KUWAIT By MUHAMMAD ALMATAR A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011
2 2011 Muhammad Almatar
3 To my parents, wife, children and family
4 ACKNOWLEDGMENTS I acknowledge and thank my committee members Dr. Jane Southworth, Dr. Youliang Qiu, Dr. Timothy Fik, and Dr. Paul Zwick for their support and for making this dissertation possible. I would like to express my immense gratitu de and appreciation for my adv iser Dr. Jane Southworth, who guided me through my academic experience at UF and was always a great friend. Dr. Sou thworth encouraged me to meet the deadlines of my scholarship and helped me in my research process. I would lik e to thank all of my coll e a g u es and friends in the department who have supported me and helped me I also would like to thank Ms. Price Desiree for her help and support as well as all members o f the UF Geography D epartment for providing me with a fa mily environment during my six years of learning I give m any thanks to any person who has helped me with my educational experience I would like to thank Allah, my parents, my wife, my brothers, and all of my family members for encouraging me to come to the US A to pursue my higher education degree I would l ike to express my deepest appreciation to my country and to Kuwait University for giving me this opportuni ty and for offering me the financial support necessary to further my education. I would like to expre ss my sincere thanks to all of the faculty and administrative staff of the Geography D epartment at Kuwait University for supporting me and helping me whenever I needed it I would like to e specially thank Dr. Obaied Al Otiabie and Dr. Waleed Al Monaies for encouraging me to apply for the scholarship when I was an undergraduate s tudent and for their support in my graduate experience I would also like to show my appreciation to K uwait Municipality for providing me with GIS datasets.
5 I would like to take this opportunity to thank all my friends here in Gainesville for their support and en courag ement to keep working and for making me feel like I am home with my family. I would like to thank my roommate Eisa Al Nashmi, my friend Talal Mah m mied and all of my fri ends for helping me to get back on track wheneve r I was down. I also would like to thank my editor Ms. Ar a go na for supporting and helping me with my writing. Finally, I would like to express my gratitude to every person who has helped, support ed, and taugh t me anything
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREV IATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 13 First Research Paper ................................ ................................ .............................. 15 Second Research Paper ................................ ................................ ......................... 17 Third Research Paper ................................ ................................ ............................. 18 2 UTILIZING GIS AND REMOTE SENSING TO PREDICT LAND USE AND LAND COVER CHANGE IN ALACHUA COUNTY, FL FOR 2013, 2018, AND 2023 ................................ ................................ ................................ ........................ 21 Methodology ................................ ................................ ................................ ........... 27 Study Area ................................ ................................ ................................ ........ 27 Validation ................................ ................................ ................................ .......... 31 Results ................................ ................................ ................................ .................... 32 Discussion and Summa ry ................................ ................................ ....................... 35 3 URBANIZATION AND THE URBAN COOL ISLAND IN THE STATE OF KUWAIT, USING REMOTE SENSING AND GIS ANALYSIS FROM 1986 TO 2010 ................................ ................................ ................................ ........................ 49 Methodology ................................ ................................ ................................ ........... 54 Study Area ................................ ................................ ................................ ........ 54 Data Preparation ................................ ................................ .............................. 56 Image classification ................................ ................................ .................... 57 Change detection analyses ................................ ................................ ........ 60 Thermal band analysis ................................ ................................ ............... 61 Result s ................................ ................................ ................................ .................... 62 Accuracy Assessment Analyses ................................ ................................ ....... 62 Change Detection Analyses ................................ ................................ ............. 62 Thermal Band Analysis ................................ ................................ ..................... 64 Discussion and Summary ................................ ................................ ....................... 65
7 4 UTILIZING GIS AN D REMOTE SENSING TO PREDICT LAND USE AND LAND COVER CHANGE IN THE STATE OF KUWAIT FOR 2015, 2020 2025, AND 2030 ................................ ................................ ................................ ............... 86 Methodology ................................ ................................ ................................ ........... 90 Study Area ................................ ................................ ................................ ........ 90 Data Preparation ................................ ................................ .............................. 91 CA_Markov model ................................ ................................ ..................... 93 Validation ................................ ................................ ................................ ... 95 Results and Discus sion ................................ ................................ ........................... 96 Summary ................................ ................................ ................................ .............. 101 5 CONCLUSION ................................ ................................ ................................ ...... 113 LIST OF REFERENCES ................................ ................................ ............................. 121 BIOGRAPH ICAL SKETCH ................................ ................................ .......................... 133
8 LIST OF TABLES Table page 2 1 Classification accuracy assessments of 1993, 1998, 2003, and 2008 classified images of Alachua County. ................................ ................................ 41 2 2 Predicted models accuracy for each image input pair. ................................ ....... 42 2 3 Future changes in Alachua County, FL from 1998 to 2013, 2018, and 2023. ..... 43 2 4 Future changes in Alachua County, FL from 1998 to 2013, 2018, and 2023 based on 1993 and 1998 pair ................................ ................................ ............. 44 3 1 Supervised classification 1986, 1990, 2000, 2005, and 2010 ac curacy assessment. ................................ ................................ ................................ ...... 73 3 2 T he changes in land use classes in pixels and Square kilometers of 1986, 1990, 2000, 2005, and 2010 classified images. ................................ ................. 75 3 3 Change trajectories of land use and land cover classes of interest for supervised classification images (1986 1990 2000 2005 2010). Built= Built up area and Non= Non built up area. ................................ ................................ 75 3 4 T hermal band analyses of built up and non built up areas for 1986, 1990, and 2000 classified images. BB Temp= Black body temperatures ........................... 76 4 1 Classification accuracy a ssessment of 1986, 1990, 2000, 2005, and 2010 classified images of Kuwait. ................................ ................................ ............. 103 4 2 Predicted models accuracy for each pair model. ................................ .............. 104 4 3 Future changes in Kuwait from 2010 to 2015, 2020, 2025, and 2030. ............. 106
9 LIST OF FIGURES Figure page 2 1 The Study area is Alachua County map that includes major roads and cities name. ................................ ................................ ................................ ................. 45 2 2 The classified remotely sensed images for years: a 1993, b 1998, c 2003, and d 2008. ................................ ................................ ................................ ........ 46 2 3 The predicted land use and land cover maps for years: a 2013, b 2018, and c 2023 based on the input image date 1998 2003. ................................ ............... 47 2 4 The population of Alachua County, FL from 1950 to 2025. ................................ 48 3 1 M iddle Eastern Countries. ................................ ................................ .................. 77 3 2 Kuwait districts and boundar ies. ................................ ................................ ....... 78 3 3 Flowchart of land use and land cover change analyses. ................................ .... 79 3 4 The classified remotely sensed images for years: a 1986, b 1990, c 2000, d 2005, and e 2010. ................................ ................................ ............................... 80 3 5 Change detection analyses for non built up and built up areas. ......................... 81 3 6 The thermal band analysis of Kuwait bulit up and non bui lt up areas for years: a 1986, b 1990,and c 2000. ................................ ................................ ..... 82 3 7 A set of training samples in central of the study area. ................................ ........ 83 3 8 Land use and land cover change of Kuwait in SQ kilometres from 1986 to 2010. ................................ ................................ ................................ .................. 84 3 9 Land use and land cover change in pixels of Kuwait from 1986 to 2010. ........... 85 4 1 M iddle Eastern Countries. ................................ ................................ ................ 107 4 2 Kuwait districts and boundaries. Source: Kuwait municipality. ......................... 108 4 3 Markov chain analyses and CA_Markov model. ................................ ............... 109 4 4 The classified remotely sensed images for years: a 1986, b 1990, c 2000, d 2005, and e 2010. ................................ ................................ ............................. 110 4 5 Future land use changes in Kuwait. ................................ ................................ .. 111 4 6 The predicted land use and land cover maps for years: a 2015, b 2020, c 2025, and 2030 based on the input image date 1990 2005. ............................ 112
1 0 LIST OF ABBREVIATION S CA_Markov Cellular Automata Markov GIS Geographic Information Science Landsat ETM + Landsat Enhanced Thematic Mapper Landsat TM Landsat Thematic Mapper LCS Land Change Science LST Land Surface Temperatures RS Remote Sensing
11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy UTILIZING GEOGRAPHIC INFORMATION SYSTEM S AND R EMOTE SENSING TO INVESTIGATE URBANIZATION PROCESSES: IN BOTH THE US AND KUWAIT By Muhammad Almatar December 2011 Chair: Jane Southworth Major: Geography Urbanization is occurring worldwide due to the tremendous increase of human population that causes land transformation. As the populations of developed and developing countries continue to grow, the development of new urban areas becomes necessary in order to accommodate the new population. Urbanization, therefore, has become one of the dominant research areas within the Land Change Science research community as researchers and policymakers attempt to manage current urbanized areas as well as future urban expansions. Moreo ver, utilizing new technologies such as Remote Sensing (RS) and Geographic Information Systems (GIS) to study urban expansions might minimize potentially negative effects of urbanization. This dissertation incorporates GIS and RS to evaluate the urban expa nsion of two contrasting study areas: Alachua County, Florida and the State of Kuwait. Urban expansion and population growth were investigated within each of these study areas. Nine Landsat TM and ETM+ images were also implemented in this research to monit or, analyze, and predict urban land use change for both Alachua County and the State of Kuwait. Furthermore, thermal band analysis was utilized to examine the differences in
12 Land Surface Temperatures (LST) between the built up and non built up areas of Kuw ait. The findings of this research indicate urban expansion is occurring in both developing and developed countries, and it will continue to occur in the future. Alachua County and Kuwait will both face a great deal of land transformation to urban land use areas, and both Alachua County and Kuwait should take the appropriate measures necessary to minimize unplanned urban expansion. The surprising finding that the LST of built up areas in Kuwait is cooler than that of non built up areas (defined in this study as an Urban Cool Island) is an unexpected deviation from the Urban Heat Island discussed in many previous studies. Therefore, this research suggests: 1) The Kuwaiti government should designate more areas for urban development to accommodate the future population and further investigate the implications of an Urban Cool Island; and 2) The Alachua County government should generate a new policy to limit the amount of urban expansion and to encour age vertical urban dev elopment.
13 CHAPTER 1 INTRODUCTION Humans have played a major role in changing and transforming their surrounding areas since their existence on the Earth, but the speed of transformation has never been as rapid as t oday (Kates et al., 1993). I ncrease d land t ransformation is related to multiple factors but it is primarily linked to human actions. Urbanization, a phenomenon discussed in several fields and numerous studies reflects the relationship between humans and their environment through the transformati on of land cover from rural areas to urban areas for human use Humans obtain needs, like food, from the various landscapes of their surrounding area s causing massive changes including the drainage of wetlands, loss of vegetation, and expansion of urbanize d areas. These activities are considered the main forces of change in the biosphere and in the shape of land cover (Turner and Meyer, 1994). Though there are many types of processes that transform ominant process of current and future land transformation activity. Population increase is known to be one of the primary causes of land transformation and changes to the global environment (Sage, 1994). T he increase of population in urban areas necessitates the development of new urban areas in order to accommodate the new population. Urban areas are defined as the places where the majority of people live and work (Elvidge et al., 2004) and urban population describe s the number of people living in urban areas. The boost of world p opulation occupying urban areas especially within the 20 th century, has increased the process of urbanization significantly across the world As where the urban pop ulation mad e up 5% of the total population in 1900, this percentage increased to 50% by 2005 (Maktav et al.,
14 2005). Furthermore, increasing population has a cumulative impact on the biosphere correlating with an increase in the level of urbanization (Turner and Meyer, 1994). A pot ential problem associated with urbanization is unplanned expansion. The unplanned expansion of urban areas could lead to many problems like natural disaster s poverty, increase s in crime rate, and disease. According to Nechyaba and Walsh (2004), urban expa nsion causes greater amounts of air pollution as the increase of car and vehicle traffic increases gas emissions in urban and suburban areas. Moreover, the many changes in the environment relating to urbanization and suburbanization include greater amounts of precipitation runoff, change s of temperatures, increase s of air pollution, and replacement s of vegetation (Elvidge et al., 2004). These environmental and social problems will continue to rise with the increase of urban expansion and human population. S tudying urban land use, therefore, is important as increased awareness of impending change s enables the existing social structure to circumvent environmental and social problems produced by unplanned expansion (Rahman 2007). To control, minimize, and solv e the social and environmental problems linked to unplanned urban expansion several res earch communities and scientists have started to investigate their causes and effects. Geography is the only social science that examines the relations hip between society and the natural environment (Peet, 1 998). Geographers analyze the environmental system from both physical an d human perspectives, and investigate how natural forces a nd living organisms reshape the surface over time. According to Mannion (2002), there are tw o main types of changes in the E ce: 1) natural changes that take place on the E surf ace resulting from natural ly occurring events such as earthquake s volcano es, and
15 climate changes ; 2 ) human chan ges that occur as land transformation s resulting from human activities like urban and agricultur al developments. Furthermore, Land Change Science (LCS) is a well known research community comprised of various scientific fields interested in investigating the impact of human activities on the surrounding environment. Geographers and LCS researchers effectively utilize remote sensing and GIS to monitor the effects of urbanization on the environment and to provide enhanced future planning for decision makers and urban plan ners (Tatem and Hay, 2004). This dissertation incorporate d Remote Sensing and GIS to explore urb an land use and land cover change in two contrasting study areas. The study areas were selected in consideration of the differing levels of urban expansion between developed and developing countries. The first study area, Alachua County, is located in North Central FL in t he developed country of the USA. T he second study ar ea, Kuwait is situated in the northeastern region of the Middle East This dissertation implemented satellite images to analyze urban expansion in both study areas. Three research papers are included in this dissertation; the first research paper was cond ucted in Alachua County, FL, and the other two were conducted in Kuwait. First Research Paper The first paper, presented in Chapter 2, utilized four Landsat TM images for 1993, 1998, 2003, and 2008 to predict future land use and land cover changes The Mar ch 1993, February 1998, and March 2003 images were classified and tested f or accuracy in a previous study (Almatar, 2008). The April 2008 image was classified and checked for accuracy using accuracy assessment a nalysis. The classified images were grouped i nto pairs (1993 1998, 1993 2003, and 1998 2003) and implemented in to a land change model to forecast change s in Alachua County, FL for 2013, 2018, and 2023. A Cellular
16 Autom ata Mark ov model in conjunction with Markov chain analysi s was employed to answer t he main research questions: 1) What type s of lan d use and land cover changes may occur in Alachua County in the years 2013, 2018, and 2023 ? 2) Which is the best pair of classified images to produce the best prediction model based on the validation process? An alternate scenario was also produced to demonstrate the trend of land transformation before the real estate boom. G iven the real estate crash since 2009 and the booming real estate of the time peri od pre 2009 used to predict this analysis, it was felt this might have affected the resulting trend of future land transformation Therefore, Landsat TM images for 1993 and 1998 were implemented in the Cellular Automata Markov model to predict future change b ased on the much slower real estate growth from the 11) The finding s of the two mo dels therefore, indicate different scenarios of future land use and land cover change in Alachua County and allow for a detailed interpretation of the rates and patterns of predicted changes in urban growth The outcomes of this study provide d information regarding the level of urban land transformation occurring in a city located in the developed world by utilizing remote sensing and GIS techniques More spec ifically, this research utilized Cellular Markov Model, an advanced prediction model, to predict and analyze future land transformations and to ascertain valuable information for future planning. Indicating future urban land transformations is req uired in order to achieve the ultimate goal of this research, which is to compare future urban land transformation in deve loped and developing countries. Such comparison enables geographers and Land Change
17 Science researchers to better understand the relat ionship between human s and the natural environment. This information might also be helpful for urban plann ers and decision makers Second Research Paper The second research paper presented in Chapter 3, employed five cloud free Landsat TM and ETM + images for 1986, 1990, 2000, 2005, and 2010. The satellite images were classified using supervised classification (maximum likelihood algorithm) for built up and non built areas. The classified images were incorporated in change detection analyses and thermal ban d analyses to answer the three main research questions : 1) How has urban land use expanded over other land cover classes in Kuwait from 1986 to 2010? 2 ) How does the increase of population influence the direction of urban expansion in Kuwait? 3 ) Is the increase of urbanization affecting the Land Surface T emperatures (LST) in Kuwait? The classified Landsat TM and ETM+ images were implemented in change detection analyses to determine changes of land use and land cover in Kuwait from 1986 to 2010. T hermal band analysis was undertaken in this paper to examine the impact of land use an d land cover change on the LST The classified image s of 1986, 1 990, an d 2000 were associated with thermal band values, which were converted into degrees Fahrenheit The built up and non bu ilt classes were linked to thermal band pixels values to detect the differences in surface temperatures to investigate the effect of land transformat ion on the surface temperature. The findings of this paper contribute to the overall goal of this research, which is to detect the impact of human actions on the surrounding environment. T his paper determine d the amount of urban land transformation in Kuwait from 1986 to 2010 to be
18 associated with urban population growth. Furthermore, the the rmal band analysis outcomes are useful to understand ing the relationship between urban land transformation and Land Surface Temperature s (LST) Though t hermal band analysis is a well known approach used in several researches to investigate the environ mental effect of urban land transformations on Land Surface Temperature s (LST) this is cons idered to be the first attempt to use thermal band analysis for this purpose in Kuwait Additionally, this methodology was purposefully selected with the intention of examining the connection between urban expansion and the phenomenon in Kuwait. Moreover, t he outputs of this study were implemented in a Cellular Automata Ma rkov model and Markov chain analysis to predict f uture urban land use change s in Kuwait. Third Research Paper The third paper, presented in Chapter 4, utilized five Landsat TM and ETM+ images for 1986, 1990, 2000 2005 and 2010 to predict future chan ges in land use and land cover. The se images were classified and tested fo r accuracy in Chapter 3 The accuracies of the five classified images were ranging from 90% to 94%. The classified images were implemented into pair s (1986 1990, 1986 2000, 1986 2005, 1990 2000, 1990 2005, and 2000 2005 ) in a land change model t o forecast changes of land use and land cover in Kuwait for 2015, 2020, 2025, and 2030 A Cellular Automata Markov model along with Markov chain ana lyses was employed to answer two main res earch ques tions: 1) Based on the current trends of population growth and urban expansion, how will urban land use increase in Kuwait from 2010 to 2030 ? 2) Which pair of dates best predict s current changes and so is most likely to predict future changes with the most confidence according to the validation process ?
19 To compare urban land transformation s in developed and developing countries, the results of this paper predicted the urban land transformation in Kuwait, a developing country, and contrasted the outputs with those of Alachua County, located in a developed country. The comparison of urban land transformation s between develop ed and developing countries was conducted in this research to measure the impact of hu man activities on the proximate environment. Moreover, determining future urban land transformation s in Alachua County and Kuwait could offer important information related to the amount s location s and trend s of future changes. The study utilized the outc omes of this research to recommend additional policies and proactive measures to manage the future urban land transformations of both study areas Employing an advanced land change prediction model also was validated in this study as a powerful model, with in the remote sensing and GIS field s to predict future land use and land cover change s This paper, as the final study, is considered to be the last step in the process of achiev ing the main goal of this dissertation which is to examine the relationship between human activities and the surrounding environment. In the first paper, data was gathered and analyzed in Alachua County to predict future urban land transformations in a developed country. The second paper investigated the change s o f urban land use in Kuwait, a developing country, by utilizing satellite images and GIS data from the recent past. Lastly, the third paper, based on the outputs generated in the second paper, served to predict future urban land transformations in Kuwait to be compared to those of Alachua County predicted in the first paper. Each study was designed with the
20 expectation of further understanding the phenomenon of urbanization as a link between human activity and the environment.
21 CHAPTER 2 U TILIZING GIS AND R EMOTE SENSING TO PREDICT LAND USE AND LAND COVER CHANGE IN ALACHUA COUNTY, FL FOR 2013, 2018, AND 2023 human existence, and there are currently few areas remainin g in their natural state. The rate of change, however, has never been as great as today (Turner and Meyer, 1994). Human activities cause massive changes to various environmental landscapes including the drainage of wetlands, loss of vegetation, and expansion of urbanized ar eas. These changes reflect human social goals (such as the production of food) and directly affect the E resources and the properties of the Earth, the interaction between humans and their surroun dings requires more investigation and policies to sustain the biosphere (Turner and Meyer, 1994). The continuous increase of human population, accompanied by an increase in human consumption of natural resources, causes warmer temperatures, increases green house gas concentration in the atmosphere, and creates various other environmental issues (Elvidge et al., 2004). Urbanization is the shift from a rural to an urban society (Nsiah gyabaah, 2009), and urban land use is characterized by the man made settle ments in which most people live and work (Elvidge et al., 2004). Furthermore, urbanization is defined as the environment altering activities which maintai n and generate urbanized places, such as the construction of buildings, transportation, habitation, a nd water and energy use (Elvidge et al., 2004). In other words, urbanization is the process of converting any land use or land cover areas in to built area s established for human uses. Though land use research, in the previous two decades, demonstrated the importance of deforestation, reforestation, and loss of agricultural land, land use research has more recently
22 exhibited an increased focus on urbanization (Fragkias and Seto, 2007). Several land use and land cover research studies concentrate on studying urban expansion/ developmen t as one of the major factors driving land cover conversions (Gluch, 2002). Suburbanization and urban sprawl are similar processes of conversion that have become noticeable in many countries all over the world, especially in the w estern hemisphere. In general, suburbanization and urban sprawl have very similar representations and specifications despite slightly differing definitions. Suburbanization, beginning in the early twentieth century, is the mass phenomenon in which skilled working families moved to live at a farther distance from their work. According to Johnston et al. (2000), a subu rb is an outer district located within an urban area and urban sprawl is a term used to describe the unplanned expansion of urban land into rural areas along main roads. As suburbanization and urban sprawl each utilize the concept of urbanization as a process of conversion, which expands the existing urban area into the surrounding rural area, suburbanization and urban sprawl will be discussed as one phenomenon. the year 2025 (Gelbard et al., 1999). This future estimation of urban expansion leads several disciplines in social sciences to try to unders tand, analyze, and predict future changes in land use. Doygun and Alphan (2006) studied long term urbanization change and urban land use change on the coastal environ ment of Iskenderun, Turkey, and concl uded urban area s have increase d by 40 times from 1858 to 2002. The unplanned expansion of urban and suburban area s could lead to many problems including natural
23 disaster s poverty, increase s in crime rate s and diseases. According to Nechyaba and Walsh (2004), urban expansion has a strong relation to air pol lution as increased car and vehicle traffic increases gas emissions in urban and suburban area s. Moreover, many changes in the environment relate to urbanization and suburbanization such as greater amounts of precipitation runoff, change s of temperatures, increase s of air pollution, and replacement of vegetation (Elv idge et al., 2004). T his kind of study therefore, is paramount to the prevention of future environmental and social problems linked to unplanned urban expansion (Rahman 2007). Remotely sensed data is one of the richest sources of data used to study spatial and temporal features of land use and land cover changes across (Yang and LO, 2002; Almatar, 2008). Remote sensing and Geographic Information System (GIS) have vastly impr oved the monitoring, analyzing, and predicting of land use and land cover change for a specific area over a period of time. These new technologies, opposed to the manual process of mapping utilized by the historical cartographer, have become invaluable to the m odern assessment of the relationship between humans and the environment. Implementing remotely sensed data in landscape investigations, over spatial and temporal scales, could assist in distinguishing changes that oc cur on the Earth and provide a bett er underst anding of how the land is influenced by human activities (Zhang et al., 2002). The availability of remotely sensed data, which can be observed and collected for large areas within a short period of time, promotes the development of image processi ng and GIS software used within the academic fields.
24 Remotely sensed data is commonly implemented in land use research to study land use and land cover changes based on the classification of satellite images ( Zhang et al., 2002; Kamusoko et al., 2009; Weng 2002; Henriquez et al., 2006; and Lopez et al., 2001). According to Meyer and Turner (1992), land use is a term referring to the land associated with human behavior and interaction. The term land cover refers to the land covered by natural species not re lated to human impact. Land use is exemplified by agricultural and urban development, and land cover is illustrated by forestland and wetland. Changes of land use and land cover are categorized as either modification or conversion. Whereas modification ref ers to a change within the categories of land cover and land use, such as the change from dense forest to mixed forest, conversion refers to a change resulting in the transformation of land from one type to another, such as from forest to bare land (Turner and Meyer, 1994; Almatar, 2008). Conversion, a term referring to the human appropriation of natural resources, therefore, results in the transformation of land cover to land use. For example, an area that was originally covered with trees may be converted to a retail urban plaza (Turner and Meyer, 1994). The significance of urbanization, as a type of conversion, is reflected in the natural and environmental sphere at all geographic scales and l ocations (Herold et al., 2003). As a result of the potential pr oblems of land transformation, many scientific fields have concentrated on studying and monitoring the relationship between humans and a nd producing sustainable polices to protect what remains of our natural resources (Meyer and Turner, 1992). For example, Fragkias and Seto (2007), employed logistic regression analysis to detect urban land use growth in three cities in southern China. They used satellite images to generate classified images for urban and
25 non urban areas from 1988 to 1999 They found that urban land growth increased by 451.6%. Zhang et al., 2002, conducted another study in China; their study was based on analyzing two Landsat images for 1984 and 1997. Change detection with road density and spectral inf ormation were used to study urban built area s They found that the built up area s increased from 385 km 2 in 1984 to 568 km 2 in 1997. These studies can be used for future planning by decision and policy makers to recognize and solve impending problems. In other words, it is ess ential to recognize and understand human impact on the surrounding environment to predict, explain, and respond to potential environmental, economic, and social changes (Turner and Meyer, 1994). Expansion, both urban and suburban, has become one of the most important consequences of human activity that directly and indirectly affects the environment, as well as land use and land cover, for many years. To study and monitor these consequences, scien tists in the social sciences and other fields have studied the causes behind expansion. This paper focuses on population as the major impetus of urban and suburban expansion. The US Census Bureau estimated the increase of urban area from 1960 to 2000 to be 130% in the USA (Alig et al., 2004), and the expected growths will continue with the increase of population. According to Elvidge et al. (2004), the increase of worldwide population creates a trend of concentrating increased populations within human settl ements, which expands the existing settlements past their perimeters. In the United States of America, 80% of the total population lived in urbanized area in the year 2000 (Nechyba and Walsh, 2004), and the ion by 2015. In other words, local and
26 global population increase is considered by many to be the main cause for urbanization and suburbanization. Within land use change science, there are hundreds of models that are used by geographers, urban planners, an d researchers in other fields (Brown et al., 2004). According to Fragkias and Seto (2007), a model is a tool used to organize and describe understanding of interactions within a system to make predictions and to generate and test hypotheses. Land use and l and cover science is one of the disciplines that implements models for analyzing, describing, and predicting changes in the E surface. Models are used within land use and land cover science to: 1) Provide the analyst with information that helps impro ve under standing of changes in land use; 2) Emphasize the importance of empirical diagnostic models using remotely sensed data; 3) Predict future land use changes (Brow n et al., 2004). Cellular Automata Markov (CA_Markov) and Markov chain analyses was impl emented in this paper to predict future land use and land cover changes. According to Pontius and Malonson (2005), Markov analyses are commonly employed for projecting change within categorical data by implementing mathematical formulation s of probability transitions, which were invented by the Russian Mathematician Andrei Markov (1907). In this study, population will be studied as the main factor driving urban ization and suburbanization relating the population increase in Alachua County, Florida to land us e and land cover changes. Based on the findings of previous research conducted in Alachua County, Florida, which studied land use and land cover changes from 1993 to 2003, (Almatar, 2008), this paper employs a probability model to predict future land use a nd land cover changes for 2013, 2018, and 2023. Moreover, the study will conduct pre
27 post economic boom analysis in the study area to predict change in landscape before the real estate economic boom, which occurred here from 2000 2008 approximately, and th e current economic situation, which is one of real estate collapse and stagnation in values. Given this change in the real estate market, which occurred during the time period of analysis being conducted, both analyses were considered, one a continuation o f land use changes based on the real estate boom this state and county had seen, and 1998 land use changes, or pre real estate boom scenario. The Cellular Automated Markov model is on e of the dominant tools that predict changes that will likely take place in the future assuming past land cover trends are a suitable indication of future trends (Li and Reynolds, 1997; Silverton et al., 1992; Kamusoko et al., 2009). The objective of this paper will be used to understand future potential land use and land cover changes, which may help decision and policy makers in future planning, within this region. The Markov model along with remotely sensed data of Alachua County were implemented to proj ect future changes in land use and land cover for 2013, 2018, and 2023. GIS and remote sensing software were employed in order to achieve the main objective of this research and answer two research questions. What type of land use and land cover changes ma y occur in Alachua County in the years 2013, 2018, and 2023? Which is the best pair of the classified images that will produce the best prediction model, based on the validation process? Methodology Study Area Alachua County is located in the north central part of Florida State (Figure 2 1). The county covers about 2,486 km 2 with nine cities and towns; Alachua, Archer,
28 Hawthorne, High Springs, La Crosse, Micanopy, Gainesville, Newberry, and Waldo (Phillips, 1986). Alachua County is one of the fastest growing counties in the north central part of Florida. The county has experienced a major increase of population and urban growth in the last decade, mainly due to the expansion s of the C ity of Gainesville and the University of Florida. According to the U. population increased from 57,026 in 1950 to 181,600 in 1990 (Smith et al., 2008). This increase in population continues to take place in the county, where the population for the years 2000 and 2007 were 217,955 and 247,56 1 respectively (Smith et al., 2008). The population growth should continue throughout the up coming years, as the U.S. Census Bureau estimated the population to reach 253,400 in 2010, and 296,000 in 2025 (Smith et al., 2008). Such increase s of population a nd urban growth, taking place in Alachua County, necessitate an advanced approach to predicting changes in l and use and land cover that may provide valuable information for urban planners and decisions makers. This paper attempts to provide this informatio n for better future policy and planning by generating a prediction model (CA_Markov) based on a suite of classified satellite images depicting likely future land cover changes assuming either a continued pattern of growth based on trends of the recent past or, in consideration of the current real estate crash, or a regressive growth pattern with the predicted growth returning to a pre real estate boom situation The study employed four Landsat satellite images in the prediction model. For full details regar ding image processing refer to Almatar (2008), but a brief overview of the data used for the prediction model inputs is given below. Two TM Landsat images were obtained in March 1993 and February 1998, and two Landsat ETM+ images, from
29 March 2003 and April 2008 were also used. These images were registered to UTM coordinates with an RMS error less than 10 meters. To minimize the atmospheric variation calibration techniques were performed on all images (Green et al., 2004). Image classification was undertake n on all images creating three land use and land cover classes of: 1) W ater land that consists of an area that is inundated by water suc h as rivers, lakes, and wetland; 2) V egetation that contains different species of bushes and trees with canopy greater t han or equal to 25% of the canopy closure and area that used for agri culture, pasture, and grassland; 3) Built area that includes urban and suburban area that is composed of built materials like airports, re sidential area, and others. The built area class was assumed to stay consistent over the study time frame to prevent a suburban misclassification problem with vegetated land cover (as tree cover ed suburban development look s like forest from above). In other words, once an area converts to built area it will continue to be considered built area without any further land transformation. Understanding the future changes in land use an d land cover in Alachua County w ould be difficult without a predictive model. CA_Markov was used in this study to pr edict futu re changes, which has been discussed as one of the cellular automata models that can be used to predict future land use and land cover changes (Sun e t al., 2007). The Markov chain is a mathematical analysis that generates a prob ability transition matrix th at demonstrates the likelihood of land transformation The CA_Markov model employs the transition probabilities to determine the number of cells that changed over time by using cellular dynamics (Verburg et al., 2004). The model relies on spatial transitio ns, where probabilities and transition areas are created from
30 two classified remotely sensed images to forecast future changes at a specific point of time (Mubea et al., 2009). Markov analysis requires the determination of a time period between the two inp ut images and the number of years for future prediction. The Markov analysis produces three outputs; a probability transition matrix, a probability transition area, and a suitability raster group layer that presents each land use and land cover class with its probability values. The probability values represent a number, which ranges from zero to one, where the number indicates the possibility of each class to change to another class in the assigned year. These probability matrixes are used to generate land use and land c over maps of future projection. Therefore, the four classified Landsat images were implemented into a CA_Markov model in this study. Three of the images were used to generate probability transition outputs in Markov chain analysis that were used for a CA_Markov model. The 1993 and 1998 c lassified images were used to predict a 2008 land use map, and were then validated against the observed data that was classified in ERDAS IMAGINE The same process was done for the pair of (1 998 2003) and (1993 2003). All pairs were then compared to the 2008 classified image for validation. The 1993 image was used as the base map in a Markov chain analysis for (1993 1998) and (1993 2003) models. The classified image classes of 1993 were assigned as a base for land use and land cover classes that are used for generating the probability transition matrix, area, and raster layers for both models. The second input image in the Markov chain analysis was a 1998 classified image in the first model and 2003 in the second model. For the third model (1998 2003), the base classified image was 1998 and the later image was 2003. The probability transition areas of each pair were implemented in the CA_Markov model
31 along with the suitability raster group. The model generated a predicted land use and land cover map from ea ch pair of classified images for the year 2008. The three 2008 maps were validated against the actual 2008 observed and classified image. Based on the accuracy of the predicted map, the most appropriate pair of images was then utilized in the model to proj ect future land use and land cover surface of Alachua County, FL, for 2013, 2018, and 2023. An excellent prediction model, accordin g to Brown et al. (2002), must provide: 1) an indication of the amount of land use and land cover change; 2) an indication of the spatial pattern of change; 3) an indication of the location of the predicted changes. Though there are many prediction models used in the land use and land cover field that provide some of the indications required of an excellent prediction model, mos t do not provide them all. Logistic regression, for example, is usually employed to predict the amount of future changes in land use and land cover classes as well as the location of these changes but fail s to indicate the pattern of changes. The CA_Markov model is an appropri ate technique for this research because it provides the analyst with outputs that determine the pattern of changes in land use and land cover and indicates the area and location of future changes. Validation Validating the predicted la nd use and land cover map is one of the most important steps of a predictive model in order to ensure accuracy. In most of the studies employing CA_Markov model, a validation module in the software packages was utilized to produce an accurate measurement o f the projected map. A validation was performed that measures the degree of agreement between two categorical land use
32 is the predicted map that is created from the CA_Markov model, where the reference map is the classified image of the study area in the same year. The validation attempt s to answer two major questions:1) H ow well do the pair of input maps concur in terms of the location of c ells in each land use class? 2) H ow well do the pair of maps concur in terms of the quantity of cells for each class? (Pontius and Malason, 2005). The validation process generates three Kappa indies; Kno, Klocation, and KlocationStrata. Kno is an index that calculates the overall agreement, Klocation is an index that measures the agreement based on location only, and KlocationStrata calculates the agreement based on quantity (Pontius and Malason, 2005). As mentioned previously, three models were implemented in this paper to predict a future l and use and land cover map and then identify the likely changes that may take place in Alachua County in 2013, 2018, and 2023 based on past trend s All three models (1993 1998), (1998 2003), and (1993 2003) were employed to generate the 2008 predicted map and validated with the classified image of 2008. Results According to the accuracy of the predicted land use and land cover, the 1998 and 2003 classified image pair was implemented in the prediction model to predict changes that might occur in the study ar ea in 2013, 2018, and 2023 ( Table 2 2 ). The probability transitions of Markov chain analysis outputs were used in a CA_Markov model to generate the predicted maps. The 1998 and 2003 pair was selected due to its performing level ( Table 2 2 ), and implemented in the model to predict future changes based on changes in land use and land cover between the two classified images. The predicte d changes are generated according to the probability transitions delivered from the two input images. Five yea r windows were used for the prediction model in order to
33 establish consistent intervals and to provide a better comparative of change across images. After generating probability transition and probability area, CA_Markov was implemented to produce land use predictions for the future dates. Figure 2 3 shows the predicted maps of Alachua County that were created by the model. The maps show the future distribution of land use and land cover in Alachua County in 2013, 2018, and 2023. Moreover, the land use maps were combined with the 2008 classified image (Figure 2 3) in order to highlight changes tha t are predicted to take place in the study area in 2013, 2018, and 2023. Figure 2 3 shows the predicted land use map for the study area and identifies changes to bu ilt areas with the color red. According to the future predicted map, there will be no changes in the water class while vegetated and urbanized area s will experience transformation. For example, the vegetation class will lose 119.49 km 2 which will be conve rted to built area s, by 2013 (Table 2 3). In other words, the study area will not undergo any changes to areas of the water class because the exp ansion of urban and suburban areas will not take place over wetland s and water area s and usually takes place ov er vegetated area s. According to the model Alachua County will likely face more urban and suburban expan sion in the future. Figure 2 3 shows the predicted maps of 2013, 2018, and 2023 based on 1998 and 2003 classified images. The maps indicate that 129.34 km 2 will remain as water land cover from 2008 to 2018, while vegetation will decrease from 1,991.59 to 1,872.1 km 2 (Table 2 3). According to the predicted map of Alachua County, t he transformation of vegetated areas to built areas will be 119.49 km 2 ; b uilt area s occupied 319.36 km 2 in 2008 and will increase to 438.85 km 2 by 2018. Built area land use class will boost from
34 319.36 in 2008 to 480.51 km 2 in 2023. The increase in built area class is equivalent to 161.15 km 2 The changes in 2023 in land use an d land cover were demonstrated in a map that indicates changes from 2008 to 2023 (Figure 2 3). According to CA_Markov model results, Alachua County will experience more urban and suburban expansion in the future ranging from 3% to 7%. This expansion of bui lt area will be over vegetated land cover, which means Alachua County will continue to lose forest and other vegetated area s The prediction model indicates that built area will occupy around 18% of the study area by 2023, which means that urban and suburb an area increase by 5% from 2008 to 2023. On the other hand, vegetated area s will occupy 75% of the total area in Alachua County, indicating more expansion of built area s over vegetated area s and reflect ing the post economic boom. Since the accuracy of pre diction models were above 90%, additional analysis was employed to present change in landscape based on the pre economic boom. This analysis aims to determine the changes in land use and land cover based on the level of growth between 1993 and 1998 (pre r eal estate economic boom). The results of this analysis will provide the reader with two scenarios of change in Alachua County (pre and post real estate economic boom). Table 2 4 shows the outputs of pre real estate economic boom, which generally indicates no land transformation will take place in the study area until 2018. The study area will experienced only limited land transformation from vegetation to built area by 0.01% from 2018 to 2023. As such, given economic conditions at this time, we might well expect a situation of little to no change in terms of land use and land cover within Alachua county based on these model predictions from the pre real estate economic boom conditions.
35 The results of the prediction model are more likely to be acceptable for the study area. Alachua County in recent years experienced a lot of expansion in suburban area s such as new residential developments condos and business plazas. Also the model indicates the new development area s will be mostly distributed throughout th e central and northern part s of the county which is where most of the population live s and the majority of developments exist. The location of the future urban and suburban area s was generated by the model based on the trend of changes from 1998 to 2003 in land use and land cover. The location of the new development area seem to be the actua l places that Alachua County experien ced land use and land cover changes in recent years The finding s of this model offer a valuable prediction for future distributi on s and expansion s of built area s over vegetated area s Discussion and Summary The main objectives of this study were to identify the best pair of classified images that will produce a reliable predication model, based on the validation process, and to use this prediction model to predict future changes of land use and land cover in Alachua County for 2013, 2018, and 2023. Furthermore, the study employed 1993 and 1998 images to provide a pre real estate economi c boom scenario, which indicates almost no land transformation will happen in the study area. This scenario assumed the future trend of land transfo rmation will return to pre real estate boom conditions. Looking at the post boom conditions, it was found that the 1998 2003 image pair was the best pair f or determining a prediction model due to an increased level of accuracy. This study identified CA_Markov model as a powerful tool for future prediction and for determining future land use and land cover changes in a particular area over a specified period of time. Based on this model, urbanization will be the main trend of land use and
36 land cover change in the region as urbanized areas develop over vegetated areas. The future land use and land cover maps indicate that most future changes in Alachua County w ill likely take place in the western and central portions of the county. In one study that took place in Austin, Texas, it was found that land transformation trends are influenced by the location of transportation networks, as new development tends to occu r near to existing highways (Zhou and Kockelman, 2008; Almatar, 2008). Since the major highways in Alachua County are located within the central and western parts of the county, transportation could be one of the main causes for future land use changes occ urring in these areas. This future urban and suburban expansion may also be highly related to an increase in the future population of the study area. The results of the prediction model, therefore, provide a valuable estimation of changes that may take pla ce in the study area in the next decade. The population of Alachua County is estimated to be 253,400 in 2010 and 296,000 in 2025 ( Smith et al., 2008 ) The estimated increase of population in 2010 and 2025 estimation that built area wi ll be 394.70 in 2013 and 438.85 km 2 in 2023. In 2008, built area occup ied 319 km 2 which means 13.08% of the total area was used for urban and suburban developments. According to the model, this percentage will increase to 16.1 7% in 2018 and 17.98% in 2023, which indicates more urbanized and suburbanized developments in Alachua County. In precise ter ms, the model predicts built area s will increase by 77.34 km 2 from 2008 to 2013 and by 89.15 km 2 from 2008 to 2018 which may be st rongly related to population increase and urban growth. Alachua County will likely continue to experience land transformation in 2018 and 2023 because the future change to built areas is expected to
37 come from vegetated areas causing the vegetated land cove r to suffer more loss according to the prediction model. In 2008, the vegetated land cover occupied around 1991.59 km 2 which represents 81.61% of the total area. Alig et al. (2004) in their study investigating the relationship between urbanization and pop ulation, income, and other factors using a multiple regression model, found that population has a huge impact on urban expansion. Though population is not the only factor driving urban expansion, it has a major influence on urbanization and the increase of built areas, and though this study did not implement a regression model, population could be considered one of the main factors driving urban expansion in Alachua County. Based on post real estate economic model, the amount of vegetated la nd cover will de crease to 1872.1 km 2 in 2013, which makes 76.71% of the total area, and vegetated land cover will continue to decrease in 2023 to be 1830.44 km 2 which is equivalent to 75% of the total area. Moreover, the study area will not experie nce any land transforma tion exceed ing 1% of the total area, according to the pre real estate economic model. In general, Alachua County might face more expansion in urban and suburban areas over vegetated areas. The predicted map identifies the location of future land transforma tion as tending to occur near existing built areas and transportation networks. As the predicted model was generated based on the current occupation of land use and land cover, future change will be predicted based on the changes between 1998 and 2003 imag es. In other words, the estimated location of future change is related to past change in land use and land cover. Markov analysis is a powerful tool for modelling changes in land use and land cover that are difficult to describe (Opeyemi, 2006). Hence, the models are
38 implementing probability transition matrices to determine future land use and land cover changes for a particular area, based on the previous patterns of land use and land cover. Mubea et al. (2009) utilized Markov model to predict changes in N akuru Municipality in Kenya for 2015. They used Landsat images of 1973, 1986, and 2000 to predict a 2015 land use map. They used 1973 and 1986 classified images in Markov model to predict 2000 and 2015 maps for six classes including water, urban, forest, a gricultural, rangeland, and barren. They found that the urban land use class expanded over the vegetated land cover class by 14.28 km 2 in 2015. They concluded that the Markov model possesses the capabilities of descriptive power and future prediction within land use and land cover investigation based on the results of their analysis (Mubea et al., 2009). Kamusoko et al. (2009) studied fu ture land use change in Zimbabwe by utilizing a CA_Markov model. They predicted future land use up to 2030 from three land use and land cover maps (1973, 1989, 2000, and 2005). They projected future land use and land cover change for 2010, 2020, and 2030. According to predicted maps, Zimbabwe will face severe land degradation with a decrease in woodlands and increase in other land use and land cover classes. They concluded the study with stating the importance of such a model to predict the future regarding land cover. Weng (2002) used satellite images as inputs in remote sensing and GIS software to study changes in urbanized area between 1989 and 1997 to coastal regions of China. The clas sified remotely sensed data was implemented in a Markov model to predi ct future changes. He classified three Landsat TM images based on seven land use and land cover classes including built up, barren land, cropland, horticulture farms, dike pound land, forest land,
39 and water. He used the land use and land cover change data to establish the validity of Markov prediction. The Markov process produced transition probability matrixes for 1989 1997, 1989 1994, and 1994 1997. Based on these transition matric es he calculated the percentage of change in the future. He found that 7.14 % of the land will be built area, 0.58% of the land will be barren land, 11.37% will be cropland, 24.19% will be horticulture farms, 5.1 % will be lake pound, 11.76% will be forest, and 39.39% will be water. As a result, urban and horticulture farms have in creased and cropland has decreased. Like other research employing CA_Markov models to predict future land use and land cover changes, this study found that increased urbanization causes a reduction of vegetated areas. Though the CA Markov model is a powerf ul prediction model high in accuracy, there are two limitations to this type of model. The first limitation is the model predicts future changes by assuming that past land cover trends are a suitable indication of future trends. In other words, new land us e and land cover pattern trends will not be considered in the prediction analyses (Sun et al., 2007). The second limitation is the model does not include any socioeconomic considerations in determining future projections. In order to eliminate or minimize these limitations, another advanced analys i s, such as logistic regression, could be implemented along with the prediction model to increase the variability of the results. Regardless of limitations, CA_Markov as a prediction model is considered one of the most important and advanced tools in the remote sensing and GIS fields to investigate changes in landscape.
40 The ultimate goal of this paper is to utiliz e a remote sensing and GIS advanced model to predict future urban land transformation in Alachua County for 2013, 2018, and 2023 associated with future urban population growth. The results of this paper indicate two scenarios of urban land use change in Alachua County (pre and post real estate economic boom). The first scenario detects future expansion of ur ban and suburban areas ranging from 3% to 7%. These expansions will likely be located in the central and northern parts of the county. The second scenario detected no land transformation in Alachua County from 2008 to 2018 and 0.01% of urban expansion wil l take place from 2018 to 2023. These predictions for future urban land use changes are produced according to trends detected from changes occurring in the past. Both scenarios offer valuable insight into possible urban land transformations for urban planners and decision makers in the Alachua County government.
41 Table 2 1 Classification accuracy assessments of 1993, 1998, 2003, and 2008 classified images of Alachua County. Classified Image Year Overall Classification Accuracy Overall KAPPA Statistic 1993 93.41% 0.8739 1998 97.42% 0.9513 2003 98.24% 0.9689 2008 93.41% 0.8742
42 Table 2 2 Predicted models accuracy for each image input pair. Model Pair Model Type Random Yours Perfect 1993 1998 % Correct 25.00% 93.52% 100.00% Improvement ------68.52% 6.48% Kno: 0.9136, Klocation: 0.9173, and KlocationStrata : 0.9173. 1998 2003 Model Type Random Yours Perfect % Correct 25.00% 95.93% 100.00% Improvement ------70.93% 4.07% Kno: 0.9457, Klocation: 0.9632, and KlocationStrata : 0.9632. 1993 2003 Model Type Random Yours Perfect % Correct 25.00% 95.82% 100.00% Improvement ------70.82% 4.18% Kno: 0.9443, Klocation: 0.9610, and KlocationStrata : 0.9610.
43 Table 2 3. Future predicted changes in Alachua County, FL from 1998 to 2013, 2018, and 2023. Change in pixel number and corresponding area from 2008 to 2013 Class type Classified pixels in year Classified pixels in year Change in km 2 Change in % 2008 2013 Water 143,713 143,713 No change No change Vegetation 2,212,879 2,126,942 Los s of 77.34 3.16 % Built Area 354,849 440,786 Gain of 77.34 +3.16 % Change in pixel number and corresponding area from 2008 to 2018 2008 2018 Water 143,713 143,713 No change No change Vegetation 2,212,879 2,080,111 Los s of 119.49 4.89 % Built Area 354,849 487,617 Gain of 119.49 +4.89 % Change in pixels numbers and corresponding area from 2008 to 2023 2008 2023 Water 143,713 143,713 No change No change Vegetation 2,212,879 2,033,831 Los s of 161.14 6.60 % Built Area 354,849 533,897 Gain of 161.14 +6.60 %
44 Table 2 4. Future predicted changes in Alachua County, FL from 1998 to 2013, 2018, and 2023 based on 1993 and 1998 pair Change in pixels numbers and corresponding area from 2008 to 2013 based on 1993 and 1998 Class type Classified pixels in year Classified pixels in year Change in km 2 Change in % 2008 2013 Water 143,713 143,713 No change No change Vegetation 2,212,879 2,212,815 Los s of 0.057 0.00% Built Area 354,849 354,913 Gain of 0.057 0.00% Change in pixels numbers and corresponding area from 2008 to 2018 based on 1993 and 1998 2008 2018 Water 143,713 143,713 No change No change Vegetation 2,212,879 2,212,708 Los s of 0.153 0.00% Built Area 354,849 355,020 Gain of 0.153 0.00% Change in pixels numbers and corresponding area from 2008 to 2023 based on 1993 and 1998 2008 2023 Water 143,713 143,713 No change No change Vegetation 2,212,879 2,212,550 Los s of 0.296 0.01 % Built Area 354,849 335,178 Gain of 0.296 +0.01 %
45 Figure 2 1 Study area map of Alachua County that includes major roads and cit y name s Source : Alachua County GIS Service Center.
46 Figure 2 2 The classified remotely sensed images for years: a ) 1993, b ) 1998, c ) 2003, and d ) 2008.
47 Figure 2 3 The predicted land use and land cover maps for years: a ) 2013, b ) 2018, and c ) 2023 based on the input image date 1998 2003.
48 Figure 2 4 The population of Alachua County, FL from 1950 to 2025 Source : ( Smith et al., 2008 ) 0 50,000 100,000 150,000 200,000 250,000 300,000 350,000 1950 1990 2000 2007 2010 2025 Population Population
49 CHAPTER 3 U RBANIZATION AND THE URBAN COOL ISLAND IN THE STATE OF KUWAIT, USING REMOTE SENSING AND GIS ANALYSIS FROM 1986 TO 2010 Urbanization is one of the most recognizabl e human activities transforming the Land Change Science (LCS) investigates the impact of urbanization (a s well as other causes of change in land use and land cover) on the from environmental and human perspectives using GIS and Remote Sensing (Turner, 2002; Lambin et al., 2006). Urbanization is defined as any human activities that alter the environment to generate and maintain urban places including habitation, transportation, construction, communication, industrialization, education, and many others (Elvidge et al., 2004). Chen (2006) asserts urbanization refers to the growth of urban popul ation and the expansion of urban areas. Urbanization is usually measured by calculating the percentage of the population occupying urban places from the overall population (Berry, 1980). The urban populations of both developed and developing countries cont inue to grow, and the urbanized areas of each, therefore, continue to expand. In both developing and developed countries, the increase of urban population is attributed to two main sources: net migration and natural increase (Berry, 1990). Net migration is the difference between in and out migrations; natural increase is the difference between the birth rate and the mortality rate. In developed countries urban population is expected to grow despite low natural increase, and the source of population growth i n such countries is primarily net migration. Developing countries have simultaneously high natural increase and high net migration causing the urban population growth to be much higher than that of developed countries. The percentage of urban population in India, f or example, increased from 17.3 % (62.4 million) in 1951 to 30.5 % (306.9 million) in 2001,
50 and the population of urban areas is predicted to be 400 million and 533 million by 2011 and 2021, respectively (Rahman, 2007). Furthe rmore, according to Kasarda (1991), the urban population of developing countries was only 16% from the total po pulation in 1950, boosted to 30 % within thirty five years, and is ex pected to continue to increase. Urbanization has started late in developing M iddle Eastern countries in comparison to the more developed western countries. Developing countries did not experience the transformation of rural areas to urban areas until late in the 20 th Century with the advance of economic development (Vlahov and Gale a, 2002). For example, the process of urbanization began in the Arab world in the late 20 th Century when the population started to increase rapidly and society initiated the building of more urbanized areas that are mostly styled after western cities. Mean while, the urban population in Britain is curren tly 92% where as the urban population was 8% in 1939 and 11% in 1987 (Nicholson Lord, 1987; Douglas, 1994). Though the expansions of urban areas of developed and developing countries are similarly dependent u pon continued population growth, developed countries began to experience the transformation of natural areas to built areas at an earlier date than developing countries as a result of economic and demographic growth (Knox and McCarthy, 2005). Middle Easter n countries are experiencing urban expansion and population increase. In t he developing African countries of the Middle East populations are expected to increase from 10% to 17% by 2013 (Belaid, 2010) due to the immigration from rural areas to urban area s, of populations in pursuit of a better lifestyle. This increase of population associated with urban expansion, includes several Middle Eastern coun trie s such as Egypt and Morocco. North African co untries are the most
51 urbanized o n the whole continent wit h an average urban population reaching 54% (Belaid, 2010). Furthermore, several wealthy countries in the eastern portion of the Middle East, located in western Asia, are experiencing urbanization as a result of the petroleum boom. The discovery of oil well s in Kuwait, Bahrain, Qatar, KSA, and UAE has created an economic boost resulting in an increase in population growth and urban expansion. The economic growth in these countries prompted professional experts and laborers to build in order to provide a bett er life style. In the Arabian Gulf, countries have a level of urbanization reaching an average of 84%. The urban population is 92.2% in Bahrain, 97.6% in Kuwait, and 92.5% in Qatar (U nited N ation P opulation Division 2001; Belaid, 2010). Nowadays, several Arab cities have become some of the world wide top modern cities such as Dubai in UAE, which is well known within the tourism and business industries. The oil discovery in Kuwait and the investment of oil revenue into social and economic projects are the main factors for population increase and urban expansion in the country (Abu Ayyash, 1980). The urban population of Kuwait continues to increase from both natural increase and immigration. The population for the years 1985, and 2010 were approximately 1.7 million and 3.3 million respectively (Kuwait Government Online, 2010), and the estimated population of Kuwait for 2030 is 4.2 million ( Uni ted Nation Population Division 2011). Immigration also plays a very important role in the population g rowth of Kuwait where more than two million of the population is composed of international workers (U.S Department of State, 2010). These international workers are usually from Middle Eastern and Asian countries and work in Kuwait temporarily before returning hom e. The boost of population from 1985 to 2010 in Kuwait caused an
52 increase in urbanized areas, which impacted land transformation in the country within the same period of time. The Kuwaiti Government controls oil revenue and manages the planning and impleme ntation of urban development (Kaganova et al., 2005). Three master plans for urban development were designed and implemented to accommodate the demand for new urban development s as well as the increase of population in Kuwait. The first master plan in 1951 by the British firm Minoprio, Spencely and McFarlane (Mahgoub, 2008) addressed the demand for new urban developments and shaped the current modern urbanized areas. The master plan aimed to construct transportation networks, and residential, industrial, an d commercial zones within a forty year timeframe The government developed second and third master development plans in order to accommodate the increase of population (Abu Ayyash, 1980); the population in 1957 was around 200,000 and this population incre ased to be 955,000 in 1975 (Algharib, 2008; Abu Ayyash, 1980). The increase of urbanized areas worldwide has attracted several researchers to investigate urbanization as a phenomenon that may cause a multitude of social and environmental problems. Poverty, for example, is one of the social problems relating to urbanization, as 50 % of the urban population lives below the national poverty levels in Bangladesh, El Salvador, Gambia, Guatemala and Haiti (Knox and McCarthy, 2005). Furthermore, according to Elvid ge et al (2004) many changes in the environment are strongly related to urbanization including an increase of precipitation runoff, loss of vegetation, and air pollution. According to Al Mutairi and Koushki (2009), air pollution increased and exceeded the
53 1998 to 2004. Additionally, the coastal urbanization of the southern areas of Kuwait resulted in rising sea surface temperatures (Al Rashidi et al 2007). Many social science fields monitor the tr emendous social and environmental changes resulting from urbanization in order to analyze, explain, and predict future changes. from human activities by using remote sensing t o identify the changes taking place over a period of time. According to Turner et al (2007), Land Change Science (LCS) is a research community interested in: 1) observing, detecting, and monitoring chang es in land throughout the world; 2) understanding th ese change s as a part of the human system; 3) modeling land change usin g spatially explicit models; 4) assessing field outcomes such as sustainability or resilience. Remote sensing is a technique of collecting information about objects (using satellites, a irplanes, or vehicles) without physical contact with the monitored objects. Remote sensing employs sensors to collect the electromagnetic energy reflected by an object across the surface of the Earth (Fussell et al., 1986; Jensen, 2005). In other words, La nd Change Science investigates transformations of a place over a period of time by applying a comparative analysis of remotely sensed images to identify differences. There is a limited amount of research implementing remote sensing to study land use and la nd cover changes in Kuwait. Most research utilizes historical data to determine the level of growth in urban areas. By employing remote sensing and GIS, this study investigates recent urban developments in Kuwait and the associated land transformations wit hin a period of twenty five years. Five Landsat TM and ETM + images for 1986, 1990, 2000, 2005, and 2010 were used to identify urban land use changes in
54 Kuwait from 1986 to 2010. In this study, pixel based classification was employed to classify the satelli te images; accuracy assessment analyses were implemented to ensure the reliability of the classified images; change detection analyses were utilized to determine urban land use change; and thermal band analysi s was used to detect change in LST The outcome s of this study could assist developers and policy makers in understanding the trends of urban expansion in Kuwait. Analyzing urban expansion and land transformation in the study area may support the designation of new areas for urban development. The ther mal band analysis result s could provide valuable i nfor mation about the nature of surface t emperature s for different land use and land cover categories. This study attempted to answer three main questions : 1) How has urban land use expanded over other land cover clas ses in Kuwait from 1986 to 2010? 2 ) How does the increase of population influence the direction of urban expansion in Kuwait? 3 ) Is the increase of urbanization affecting the LST in Kuwait? Methodology Study Area This study investigates Kuwait, a country located in the eastern portion of the Middle Eastern region. Kuwait is located south of Iraq and northeast of the Kingdom of Saudi Arabia, and the country is bordered by the Arabian Gulf from the eastern side (Figure 3 1). Extending between latit udes (28 and 49 majority of the land surface (Kwarteng and Chavez, 1998). The desert environment te. The summers are hot, long, and dry with a maximum temperature of 50 C; winters are warm and short with a mean temperature of
55 17.7 C and an annual precipitation average around 100mm (Algharib, 2008; Kuwait Government Online, 2010; and Kwarteng and Chave z, 1998). Kuwait has six governmental districts Al Asma, Hawalli, Al Ahmadi, Al Farwania, Al Jahra, and Mubarak Alkabier (Figure 3 2). These districts vary in terms of size, location, and age, and each district has several suburban and residential areas. A l Asma, for example, is the oldest district in the country and the nearest to the capital. The suburban and residential areas of Al Asma include Addullah al salem, Doha, Faiha, Kuwait city, Mansouriya, Nuzha, Qadsia, Shuwaikh, Shammiya, and Sulaibkhat. In 2010, the country occupied 178 18 km 2 of land with a population approximately 3.3 population is allocated in the six governmental districts along the central and southern coast li nes, which is equivalent to 6% of the total land area. According to the United Nations Population Division (2003), the urban population of Kuwait is 97.6% of the total population placing Kuwait as the second most urbanized country in the world (Mahgoub, 2008). distribution of land for residential, agricultural, industrial, and public use (Kaganova et al 2005). Private ownership of land is mostly limited to Kuwaiti nationals withi n existing developed areas. In other words, the government controls non developed land and is responsible for assigning new areas for development. The development of urban areas, therefore, requires government permission. The area designated for urban deve lopment is concentrated in a very small portion of the country, which is the main study area for this research.
56 Data Preparation Five cloud free Landsat TM and ETM + images were obtained for 1986, 1990, 2000, 2005, and 2010 for the study area. Five years wa s the intended time period between satellite images, but the availability of acceptable images changed this time frame. The 1986 TM Landsat image, acquired on February 17 th was the earliest scene. The second Landsat TM scene was acquired on September 8, 1 990. The invasion of Saddam Hussein to Kuwait in 1990 caused burning and spilling oil wells over large areas and created a problem for obtaining clear satellite images. As a result, the 2000 image, obtained on August 18 th was the third available Landsat T M image. Due to the Landsat ETM + (SLC_OFF) problem, the Landsat ETM+ images for 2005 and 2010 were obtained and gap filled through a private agency using adaptive localized histogram matching and specific scene based alterations techniques. These technique s utilized a cloud free Landsat ETM + (SLC_ON) from 2003 image as the base image to fill the missing gaps in the 2005 image. The missing gaps in the 2005 image were 27.6% of the study area. The adjusted 2010 image was later used to fill the missing gaps in the 2010 image. The missing gaps in the 2005 image were 24.6% of the study area. The Landsat ETM + 2005 scene was acquired on July 23 rd and the latest image was obtained on March 15 th of 2010. The Landsat TM and ETM + images were registered to UTM coordinat e system Zone 19 N, WGS 1984 datum with RMS error less than 10 meters. Radiometric calibration and correction techniques were performed on 1986, 1990, and 2000 Landsat images to eliminate or minimize image differences and atmospheric variations like solar elevation and atmospheric effects (Kwarteng and Chavez, 1998).
57 The radiometrically calibrated and gap filled images were overlaid with GIS layers to determine a specific area for the research investigation. A vector data set of Kuwait designated areas for development was utilized in ArcMap with the satellite images to determine the areas for the classification process. Furthermore the Arabian Gulf area was eliminated from all satellite images to make the classification process simpler. The assigned areas f or urban development were subset from all images to provide better classification accuracy. The images include all developed areas and areas assigned for future development. The five subset Landsat images were incorporated in the classification process, ac curacy assessment, and change detections. Image c lassification In this study pixel based classification methods were employed to classify all Landsat images. The process of pixel based analysis is to classify the remotely sensed images according to each p ixel value presenting a reflectance signature. This set of analysis techniques includes two widely utilized classifications: supervised classification and unsupervised classification. Supervised classification is defined as the process of determining the identity and location for land use and land cover classes based on either previous knowledge of the study areas or interpretation of aerial photography, map analysis, and personal experience (Hodgson et al 2003; Jensen, 2005). After identifying sets of p ixel values for each land use and land cover class, an algorithm calculation is usually assigned in the software to generate a final output classified image or map. The analyst selects the spectral signature or pixel values from previous observation, and t he computer system then recognizes the pixels with like characteristics to determine their class members hip (Elhadi and Zomrawi, 2010).
58 Depending on the data needing to be analyzed and the study objectives, the analyst selects an algorithm calculation such as maximum l ikelihood, minimum distance, or nearest neighbor (Jensen, 2005). The maximum likelihood algorithm, considered the most commonly used algorithm (Jensen, 2005), calculates the probability of assigned pixels relating to a specific class and assum es that each class in every band is statistically normally distributed (Elhadi and Zomrawi, 2010). To achieve study objectives, Southworth (2004) employed the maximum likelihood algorithm to classify a Landsat TM into se ven classes: 1) urban/roads; 2) wate r sink holes; 3) water/shade; 4) bare/soil; 5) irrigated agriculture; 6) ea rly mid successional forest; 7) mid late successional forest. The study was conducted in Yucatan, Mexico to investigate the ability of thermal band classification to differentiate t he types of forest successional growth stages. The study concluded that thermal band is very useful for land cover class differentiation. Though supervised classification (maximum likelihood algorithm) was the main approach used for classifying images, uns upervised classification was the ini tial classification method used in order to determine the preliminary classification schemes The result s of the unsupervised classification method (ISODATA) indic a ted the simplicity of land cover classes in the study area. In other words unsupervised classification assisted the classification schemes determination process for two land use and land cover classes including built up and non built areas Built up areas located within urban and suburban areas, are charact erized by built structures such as airports, highways, and shopping centers. Non built areas include areas devoid of built structures
59 such as deserts, farms, and bodies of water. GIS layers, Google Earth and prior knowledge of the study area were employe d to classify the satellite images. Unsupervised classification is defined as the process of classifying remote sensing images into land use and land cover classes without prior knowledge of the study areas. The process of classification relies on the comp uter system to group pixel values with similar characteristics into unique classes based on the statistical determination criteria (Duda et al 2001; Jensen, 2005). This type of analysis is usually employed to provide general information regarding cluster s in the image. The analyst utilizes unsupervised classification to understand the nature of pixel values by producing several unsupervised classified images. Unsupervised classification, therefore, assists the analyst in determi ning the classification sch eme. Unsupervised classification is very suitable for obtaining information about land use and land cover from a multispectral remote sensing image (Loveland et al., 1999; Huang, 2002; Jensen, 2005). This classification provides spectral clusters to be de termined with a high level of objectivity through two main steps: grouping the pixel values with similar properties into clusters and labeling the clusters by the analyst based on their characteristics (Yang and Lo, 2002). Both techniques were used to clas sify the remote ly sensed images in this study. The unsupervised classification was used in this research to investigate the nature of land use and land cover types in Kuwait. One of the study remote ly sensed images was classified first for ten classes an d then for five classes to check the level of reflectance of land cover and land use type in the study area. The unsupervised classification indicated that land use and land cover classes in the study area are easily differentiate d from each other.
60 Change detection a nalyses Change detection analyses were implemented in this study to detect land use and land cover changes in Kuwait from 1986 to 2010. Urban remote sensing research generally uses change detection techniques to monitor changes in urban land use areas over a period of time (Weng, 2002; Yagoub, 2004; Shalaby and Tateishi, 2007; Yuan, 2010). Change detection techniques detect the date s the types, the sizes, and the spatial patterns of change (MacLeod and Congalton, 1998; Lu et al., 2004). Post cla ssification change detection is considered to be a straightforward comparison te chnique (Zhang et al., 2002). This method requires two classified images from two different dates for the same study area. These two classified images should have the same land use and land cover classes, which need to be combined in order to determine the from to changes of the classes. Post classification change detection was incorporated in this study to identify changes in urban areas from 1986 to 2010. A change trajectory i mage was created to incorporate the trajectories of built up and non built development at a pixel scale. The 5 image date s were used (1986, 1990, 2000, 2005, a nd 2010) to create a 32 class change trajectory. For example, a non built area that was urbanized in 2005 would have a trajectory recorded in those pixels of Non Non Non Built Built. As such we can record the amount of change at a pixel le vel and the date the change occurred. From these 32 change classes the mos t dominant trajectories can be distingui shed The result of the change detection analyses identifies changes in built and non urban development area from 1986 to 2010 (Figure 3 3).
61 Thermal band a nalysis Thermal bands analysis was also undertaken in this research to invest igate the bands measure the infrared energy emitted from the Ear appropriate for analyzing surface temperatures and therma l activities (Jensen, 2000). This analysis examined the temperatures of non built and built up areas to determine potential differences in temperature. In this study, thermal bands of Landsat TM and ETM+ were analyzed for 1986, 1990, and 2000 to i Su rface T emperatures (LTS) The Landsat ETM + images were eliminated from this analysis due to the sensor problem (SLC_OFF) in thermal band. The extracted thermal bands were combined with the classified images to determine the temperatures of the ace based on non built and built up classifications. The temperatures of the two classes were converted to black body temperatures in degrees Fahrenheit with in the IDRISI Taiga software package. The results of this analysis determine d the effect of urban ization on surface temperatures in Kuwait. Several studies within the remote sensing field have either conducted thermal band analysis independently to examine surface temperatures or integrated thermal band analysis into the process of classifying remote l y sensed images to increase the level of accuracy In a study conducted by Southworth (2004) thermal band analysis was used along with a maximum likelihood classification for Landsat TM images in order to link the thermal band of the land cover classes. Th e study found a significant relationship between land cover class and resultant blackbody surface temperatures Moreover, Jiang and Tian (2010) conducted a study in Beijing, China using thermal band analyses to investigate the influence of urban developmen t on LST They utilized
62 three Landsat TM and ETM + scenes to cover Beijing area, which were obtained in1995 and 2000, to retrieve the temperatures of the land surface for forest land, cultivated land, water, grassland, built up and unused land. They also ge nerated temperature vegetation index spaces to study the effect of land changes on the temperatures. The study found that the change in land use and land cover classes increased the surface temperatures in the Beijing area. The increase of LST was related to the loss of forest land and the expansion of urban development. Results Accuracy Assessment Analyses The five Landsat TM and ETM + images were classified in ERDAS IMAGINE using a maximum likelihood algorithm technique based on initial unsupervised clas sification clusters (Figure 3 4). The classified images were incorporated in the accuracy assessment analyses to measure the level of agreement between the classified pixels and the ground surface. The analyses employed Kappa coefficient to calculate overa ll accuracy for all classified images. A minimum overall accuracy of 90% was used to determine the acceptability of the images within the urban remote sensing field. The 300 training points obtained from GIS and Google Earth and prior knowledge of the study area were implemented in the accuracy tests. The accuracy Kappa statistic. Table 3 1 shows results for the accuracy assessment analyses for all classified images. Change Detection A nalyses The classified images were analyzed to detect changes in land use and land cover classes in Kuwait from 1986 to 2010. A table was generated to show the classified
63 pixels and the in formation about the cov erage in square kilometers for built up and non built areas (Table 3 2) Based on the change of land use and land cover classes in pixels and square miles, the study area experienced an expansion of urban development from 1986 to 2010. The general trends o f change in land use and land cover classes in Kuwait indicate an increase of urban development within a time period of 24 years. Change detection analyses were undertaken for all classified images to analyze these changes. Table 3 2 details information ab out land transformation in Kuwait from 1986 to 2010 in built up and non built areas. The classified images of built up and non built areas were implemented in post classification change detection analyses. The classified images were combined to determine t he states of from to changes in the classes. The change trajectories for two land use and land cover classes for 1986, 1990, 2000, 2005, and 2010 may indicate the amount of change taking place in Kuwait within the study time frame. The changes in land use and land cover in the study area were calculated in pixels, percentage and km 2 (Table 3 3). The change detection analyses were also visualized in a land use and land cover map to provide information regarding the location of change (Figure 3 5). The chang e detection table and figure indicate six change trajectories, that higher than 1% of change, within the study area: 1) Non Non Non Non Non (never urbanized); 2) Non Non Non Non Built (recent urbanization); 3) Non Non Non Built Built (urbanized since 2000) ; 4) Non Non Built Built Built (urbanized since 1990); 5) Non Built Built Built Built (urbanized since 1986); 6) Built Built Built Built Built (urbanized prior 1986). NON NON NON NON NON Indicates t he exclusively non built areas from 1986 2010. Non built areas occupied 515.5 km 2 and the number of classified pixels for non built areas was 573,704. The percentage of non built areas was 58.7% of the total study area. Figure 3 5 highlights the non built areas in brown.
64 NON NON NON NON BUILT. Shows t he transformation of non built areas to urban development areas since 2005. Table 3 3 shows 5.3% of the total study area converted to urban development area. This land transformation covered an area of 46.6 km2 located in the southern and western portions of the study area, with 51,894 classified pixels. Figure 3 5 highlights the non built areas converted to urban development areas in green. NON NON NON BUILT BUILT. Indicates t he transformation of non built areas to urban development areas since 2000. Table 3 3 shows this land transformation to be 3.5% of the total study area. It covers an area of 31 km2 with 35,588 classified pixels. The land transformation is located in the central and western portions of the study area. Figure 3 5 highlights the land trans formation areas in red. NON NON BUILT BUILT BUILT. Indicates t he transformation of non built areas to urban development areas since 1990. Table 3 3 shows the land transformation to occupy 3.8% of the total study area covering 33.6 km2 with 39,334 classifi ed pixels. The transformation of land from non built areas to built up areas was located in the central portion of the study area. Figure 3 5 highlights the land transformation areas in blue. NON BUILT BUILT BUILT BUILT. Shows t he transformation of non bu ilt areas to urban development areas since 1986. Table 3 3 shows the land transformation area was 4.7% of the total study area with 45,014 classified pixels. By this time, 41.4 km2 areas for urban development were identifiable. The transformation of land f rom non built areas to built up areas was located in the central portion of the study area. Figure 3 5 highlights the land transformation areas in purple. BUILT BUILT BUILT BUILT BUILT. Indicates t he transformation of non built areas to urban development areas since before 1986, which included areas that built up before and after 1986. Table 3 3 shows the land transformation area was 24% of the total study area at the beginning of study time series. The number of classified pixels for built up areas was 23 3,324 equal to 209.79 km2 Figure 3 5 includes built up areas that existed in the study area before 1986. These areas are located in the northern part of Kuwait, where the capital and business services are located. Kuwait experienced urban expansion in sev eral parts of the designated land for urban developments. Figure 3 5 highlights the existing urban area in yellow. Thermal Band Analysis Three classified Landsat TM images were implemented in thermal bands analysis for 1986, 1990, and 2000 to examine the i n surface temperatures in Kuwait. Table 3 4 shows the average black body temperatures for non built and built up areas of each cl assified image. The black body temperatures of built
65 up areas of the 1986 image was 59.20 F, and the black body temperatures for non built areas was 60.49 F. The black body temperatures of built up areas of the 1990 image was 104.51 F, and non built areas had a temperature of 106.51 F. The black body temperature of the 2000 image was 111.63 F, and the temperature for non built areas was 115.44 F. Figure 3 6 shows built up and non built up areas associated the black body temperatures. According to the outcomes of the thermal band analysis, temperatures of built up areas are cooler than non built areas in 1986 by 1.29 F. The deference increased to 2. 00 F in 1990 and 2000 by 3.81 F. The lower temperatures of built up areas compared to non built areas could be related to the contrasting textures and to the different types of vegetation present withi n the urbanized areas of Kuwait and the neighboring deserts. In the urbanized areas of Kuwait, the texture is influenced by the presence of buildings and the materials utilized in urban development. The buildings commonly include white flat roofs and brigh t implanted vegetation in order to provide better air quality. Contrastingly, the sand and rays to be absorbed and the temperature to rise. Discussion and Summary This research aimed to investigate land use and land cover change in Kuwait between 1986 and 2010 to determine the amount of land transformation in urban areas. The utilization of remote sensing and GIS in this study requires a process of validation to ens ure that the outcomes of analyses are acceptable and reliable. The five Landsat images were classified and validated in ERDAS IMAGINE The accuracy assessment analyses produced a level of accuracy ranging from 90.06% to 94.21%, which these
66 levels of accu racy are acceptable within the remote sensing field (Hartter and Southworth, 2009). Change detection analyses were implemented in this study for all classified images to detect land transformation taking place in Kuwait from 1986 to 2010. Within a period o f 24 years, non built areas reduced by 155.3 km 2 Non built areas decreased from 670.8 km 2 in 1986 to 515.4 km 2 in 2010. Meanwhile, built up areas increased from 209.7 km 2 in 1986 to 362.6 km 2 in 2010. The percent of built up areas from total areas escalat ed by 17.4% from 1986 to 2010. According to the change trajectories analyses, 4.7% of the total area transformed to urban development areas between 1986 and1990. This percentage which decreased the percentage of non built areas transformed to built up areas to 3.8%. The percentage of land transformed to non built areas was 3.5% between 2000 and 2005, and the percentage increased to be 5.3% from 2005 to 2010. Urban areas in Kuwait, during the time period of this study, expanded in different directions. The overall change in built up areas was located in the central and southern portions of the country. The development of built up areas resulted in the formation of new towns such as Janob Al sorah, Om Alhiman, Al Zahraa, and many others. The increase of urban development in Kuwait relates to the increased demand on urban areas caused by population growth. The population wa s almost 1.7 million in 1985 and increased to over 3.3 million by 2010 (The World Fact Book, 2010; Kuwait Government in urban areas (Mahgoub, 2008).
67 The non built areas transformed to built up area s between 1986 and 2010 in various parts of the country. The expansion of new urban areas between 1986 and 1990 included four cities located east of Fahaheel between two highways (King Fhad and Fahaheel highways): Al Qurain, Al Adan, Al Qusor, and Mubarak Alkabier (Figure 3 5). New urban areas took an infill shape just south of Kuwait City in the central portion of the study area. From 1990 to 2000, the construction of new residential areas such as Al Shoada and Al Sadieg occurred east of Khaittan and west of Fahaheel, and infill urban developments were located south and southwest of Kuwait City. Between 2000 and 2005, urban areas expanded with the construction of new residential developments in the area south of Al Jahra City, the area west of Khaittan, and in southeastern areas. Most of the urban expansion between 2005 and 2010 was located in the southern part of the study area reflecting the development of four new residential cities: Al Mangaf, Sobah Alahimad, Al Wafra, and Fahad Alahmad. The overall patt erns of urbanization indicate that Kuwait experienced a great amount of land transformation from non built to built up ar eas within the 24 years period. Thermal band analysis was implemented in this study to inves tigat e differences between the surface t emp eratures of built up and non built areas. The study found that non built areas had higher temperatures than bui lt up areas with a difference raged from 1 .29 F to 3.81 F. The ther mal analysis detected that surface temperatures decrease with increases of urb an areas (Figure 3 6). Kwarteng and Small (2005) also found in their study of thermal environments in New York City and Kuwait City that urban developments in Kuwait City are cooler than non built areas. Utilizing Landsat ETM+ images of Kuwait City an d New York City to analyze surface temperatures, they
68 concluded urban surface temperatures were higher in New York City than the surrounding areas and urban surface temperatures in Kuwait City were cooler than the surrounding areas. This interesting finding is a function of the climate in Kuwait and the surrounding non built land s which are generally deserts The thermal analysis of this study concluded that more urban areas i n Kuwait would decrease the surface temperature. The finding of this study that urban expansion (2005). The decrease of temperatures in areas of urban development could possibly be explained by the texture of residential areas. Developers incorporate vegetati on in their plans for new residential areas to minimize environmental problems like air pollution, which may decrease surface temperatures in a des ert environment such as Kuwait. This unexpected outcome is interesting in consideration of various studies re garding the many temperate regions (Shimoda, 2003; Lo and Quattrochi, 2 003; Lozada et al., 2006). Despite the positive effect urban expansion may have on surface temperature s in desert environments such as Kuwait, urbanization may st ill lead to a multitude of other social and environmental problems. The overall outcome of this study, that Kuwait experienced tremendous amounts of land transformation related to the increase of population from 1986 to 2010, is supported by the results of several previous studies conducted in the Arabian Gulf. The dominant driver of land transformations with the urban built area increasing from 209.8 km 2 (23.4% of area) in 1986 to a total of 362.5 km 2 (40.8% of area) by 2010, is the population growth. A study conducted by Algharib (2008) employed two Landsat TM images to detect changes in the land use and land cover of Kuwait between 1989 and
69 2001. The study found urban expansion in the southwest, south and west portions of the study area caused a 10% increase of the total urban area. Al Awadhi (2007) conducted a study in Muscat, Oman to monitor and model urban expansion by utilizing GIS and remote sensing. He used historical land use maps to generate d etailed datasets along with aerial photographs and high resolution satellite imagery. The study found that Muscat expanded by 650% from 1970 to 2003, where the annual urban development growth rate was 20%. Urbanization occurs in the majority of Arabian Gul f countries with the exception of Yemen. Monitoring urban growth in Al research. The study used remote sensing to monitor urban growth from 1976 to 2000, and utilized maps, color aerial photographs, GPS, and GIS The satellite images were obtained for 1976, 1978, 1984, 1994, 1998, and 2000. The study found that the city of Al Ain was expanded in the direction of western and southwestern areas. Built up areas occupied 14, 870 hectares in 1990 and increased to 20, 222 hectares in 2000, which indicated an increase of 36% within a period of 10 years. The percent of change in built up areas reached 70% from 1984 to 2000 proving urbanization has taken place in the study area. Another study in UAE utilized 20 GIS layers to detect changes in Al Sammalyah Island between 1999 and 2005. The study employed overlay analyses for five land cover classes. The study found that built up areas increased by 300% within 6 years, which equivalents total area of 10 hectares. Many studies have been conducted examining urbanization as a phenomenon affecting the developing countries of the Middle East. Al Rawashdeh and Saleh (2006) conducted a study employing remote sensing and GIS to detect changes in urban areas
70 in Amman City, Jordan from 1918 to 2002. The study used aerial photographs, Landsat, and Ikonos to monitor changes of five land cover classes. The study found that built up areas increased by 162 square kilometers, which represents 509 times the existing areas in 1918. They related the ex pansion of built up areas to mass immigration and the relocation of refugees from Palestine. Moreover, Dewidar (2003) conducted change detection analyses and post classification analyses using two Landsat TM images in the northern part of the Nile de lta in Egypt between 1984 and 1997 to determine changes in land use and land cover classes. The study utilized supervised classification methods to classify the two Landsat images. The study found that urbanized areas doubled from 1984 to 1997. The increas e in urban development is a phenomenon that is not just happening in Kuwait and the Middle East but across the entire world, especially in developing countries. Turkey is a country located in the developing world, which is experiencing an expansion in urba n areas with a rise of population from natural increase and immigration. Dougun and Alphan (2006) conducted a study in Iskenderun, Turkey to monitor urbanization from 1858 to 2002. The study incorporated GIS historical datasets, aerial photographs, and Lan dsat ETM+ imagery. The study detected a rise of built up areas equal to 40 times those of 1858, and identified population growth as the main driving force for urbanization. Similarly, Kaya (2007) related urban expansion in Istanbul, Turkey between 1987 and 2001 to population growth. He used two Landsat TM and ETM+ satellite images classified for five land use and land cover classes to determine changes in the study area. The study found that urban areas increased by 1000 hectares per year, and subsequently investigated the relationship between urban
71 expansion and population by comparing urban growth to population growth. Immigration was determined to be the main factor for population growth, which wa s reflected by urban expansion. Urbanization began in the m odern day developing world in the late 20 th Century and the developing world will continue to experience the increase of urban population and the expansion of urban areas. Urban populations will continue to rise as a result of natural increase and immigrat ion from rural areas to urbanized areas in the search for a better life style (Vlahov and Galea, 2002). Urban areas will continue to expand as a result of the population growth. The implementation of remote sensing and GIS technologies to investigate chang es in land use and land cover over periods of time could be a powerful source of information for policy makers. The outcomes of such research provide decision makers with information regarding the trends, the sizes, and the types of changes taking place in the past enabling them to generate better plans for the future. The conclusions of this study regarding land use and land cover activities in Kuwait from 1986 to 2010 support the idea of providing more land for urban developments in Kuwait. By utilizing r emote sensing and GIS technologies, valuable information was obtained that could be useful to urban planners and decision makers in the Kuwaiti Government. Urbanization and population growth are predicted to continue increasing in the study area. The conce ntration of population within a small urban area increases the likelihood of environmental problems associated with urbanization such as air pollution. The designation of additional urban development areas, therefore, is strongly suggested by this study as the construction of sustainable urban developments
72 might assist in the prevention of future environmental problems. Surprisingly the well (Bornstein 1968 ; Price, 1979; Torok et al., 2001; Shimoda, 2003; Lo and Quattrochi, 2003; Lozada et al., 2006) is here an Urban Cool Island, which is a fascinating discovery for Kuwait and for other desert environment countries. Future research will investigate this phenomenon and its import ance in the glob al change arena in more detail.
73 Table 3 1 cy and Users.Acc= Users accuracy. Supervised Classification 1986 1986 Non Built Built Up Classified Totals Users.Acc Non Built 140 16 156 0.897 Built Up 14 141 155 0.909 Reference Totals 154 158 Producers.Acc 0.909 0.892 Overall Kappa Statistics=0.8019, Overall Classification Accuracy = 90.06% Supervised Classification 1990 1990 Non Built Built Up Classified Totals Users.Acc Non Built 144 8 152 0.947 Built Up 9 149 158 0.943 Reference Totals 153 158 Producers.Acc 0.941 0.943 Overall Kappa Statistics=0.8846, Overall Classification Accuracy = 94.21% Supervised Classification 2000 2000 Non Built Built Up Classified Totals Users.Acc Non Built 144 10 154 0.935 Built Up 9 147 156 0.942 Reference Totals 153 158 Producers.Acc 0.941 0.930 Overall Kappa Statistics=0.8718, Overall Classification Accuracy = 93.57%
74 Table 3 1 Continued Supervised Classification 2005 2005 Non Built Built Up Classified Totals Users.Acc Non Built 147 17 164 0.896 Built Up 6 140 146 0.958 Reference Totals 153 158 Producers.Acc 0.960 0.886 Overall Kappa Statistics=0.8463, Overall Classification Accuracy = 92.28% Supervised Classification 2010 2010 Non Built Built Up Classified Totals Users.Acc Non Built 148 20 168 0.881 Built Up 5 137 142 0.964 Reference Totals 153 158 Producers.Acc 0.967 0.867 Overall Kappa Statistics=0.8336, Overall Classification Accuracy = 91.64%
75 Table 3 2 The changes in land use cl asses in pixels and Square kilometers of 1986, 1990, 2000, 2005, and 2010 classified images. Year Non Built Area Built Up Area Pixels SQ_ Kilometers Pixels SQ_ Kilometers 1986 745,534 670.8 233,324 209.8 1990 700,520 629.4 278,338 251.2 2000 661,186 595.7 317,672 284.9 2005 625,598 562.0 353,260 318.6 2010 573,704 515.4 405,154 362.6 Table 3 3. Change trajectories of land use and land cover classes of interest for supervised classification images (1986 1990 2000 2005 2010). Built= Built up area and Non= Non built up area. Class Type Number of classified pixels percentage of area Area in square Ki lometers Non Non Non Non Non 573,704 58.7 515.4 Non Non Non Non Built 51,894 5.3 46.6 Non Non Non Built Built 35,588 3.5 31.0 Non Non Built Built Built 39,334 3.8 33.7 Non Built Built Built Built 45,014 4.7 41.4 Built Built Built Built Built 233,324 24.0 209.8
76 Table 3 4. Thermal band analyses of built up and non built up areas for 1986, 1990, and 2000 classified images. BB Temp= Black body temperatures. Year Built Non built Deference in Temperature Acquisition Date BB Temp BB Temp 1986 59.20 6 0.49 +1.29 Non built 17 02 86 1990 104.51 106.51 +2.00 Non built 08 09 90 2000 111.63 115.44 +3.81 Non built 18 08 00
77 Figure 3 1. Middle Eastern Countries. Source: http://letustalk.wordpress.com/2008/07/19/saturday obama arrives in kabul afghanistan/
78 Figure 3 2. Kuwait districts and boundaries. Source: Kuwait municipality
79 Figure 3 3 Flowchart of land use and land cover change analyses.
80 Figure 3 4. The classified remotely sensed images for years: a ) 1986, b ) 1990, c ) 2000, d ) 2005, and e ) 2010.
81 Figure 3 5. Change detection analyses for non built up and built up areas.
82 Figure 3 6. The thermal band analysis of Kuwait bulit up and non built up areas for years: a ) 1986, b ) 1990,and c ) 2000.
83 Figure 3 7 A set of training samples in central of the study area
84 Figure 3 8. Land use and land cover change of Kuwait in SQ kilometres from 1986 to 2010. 0 100 200 300 400 500 600 700 1986 1990 2000 2005 2010 Square Kilometers Year Non-Built Area Built Up Area
85 Figure 3 9. Land use and land cover change in pixels of Kuwait from 1986 to 2010. 0 100,000 200,000 300,000 400,000 500,000 600,000 700,000 800,000 1986 1990 2000 2005 2010 Classified Pixels Year Non-Built Area Built Up Area
86 CHAPTER 4 UTILIZING GIS AND RE MOTE SENSING TO PRED ICT LAND USE AND LAN D COVER CHANGE IN THE STATE OF KUWAIT FOR 2015, 2020 2025, AND 2030 Urbanization, the shifting of rural areas to built up areas, has been transforming results from human activities and is characterized by commercial and residential settlem ents (Nsiah gyabaah, 2009; Elvidge et al., 2004 ). According to Berry (1980), urbanization is commonly measured through calculating the percentage of urban population from overall population. The enormous increase of population in developed and developing c ountries requires new human settlements expanding current urban developments to their fringes (Elvidge et al., 2004). Therefore, the two main components of urbanization, urban developments and urban population, are interconnected and interactive (Cohen, 2004). The amount of increase and the source of urban population vary between developed and developing countries. Population growth of urban areas is attributed to two major sources: natural increase (the difference between mortality rate and birth rate) and net migration (the difference between in and out migration) (Berry, 1990). The main source of population increase in the developed world is the high trend of net migration whereas natural increase is low. The developing countries are experiencing greater amounts of population growth caused by simultaneous high natural increase and high net migration. It is anticipated, therefore, that future population growth will occur mostly in d evelopin g countries, and grow by 60% by 2030 (Hall, 2003; Girard et al., 2007; Rahman et al., 2011). The high trend of population growth within developing urban areas has made some Middle Eastern countries among the hig hest urbanized areas in the world.
87 Several Middle Eastern and many developing countries have experienced a high rate of urbanization since the late 20th Century, especially the Arabian Gulf region, as a result of the petroleum boom (Vlahov and Galea, 2002) The discovery of oil wells in Arabian Gulf countries such as Kuwait, Bahrain, and UAE generated an economic boost producing an increase of urban expansion and population growth. Subsequently, an increase in urban population growth caused fu rther expansio n of urban areas. In 2001, according to the United Nations Population Division, urbanization reached 97.6% in Kuwait, 92.5% in Qatar, and 92.2% in Bahrain (AitBelaid, 2010 ). urban expansion and associated population growth are related to the gover allocation of oil revenue into the development of public services and resources as well as the improvement (Abu Ayyash, 1980). Population growth in the country is generated from two major sources: birth rate and inte rnational immigration. The population in 1985 was approximately 1.7 million and 2.2 million in 2005 (Kuwait Government Online, 2010). and economic projects requires additional workers and, therefore, causes international imm immigrants often work in Kuwait for a period of time before returning home, and the percentage of intern ational immigrants exceeded 50% of the total population in 2010 (U. S increase, and the estimated overall population is predicted at over 3.5 million (UN Population Division, 2004). Moreover, an increase in urban expansion will be necessary to accommodate future popula tion growth.
88 Remote sensing and GIS are widely utilized to monitor, analyze, and predict urban land use expansion. Rahman (2007) stated that implementing remote sensing in urban studies is very useful for: 1) monitoring large s cale urban expansion especially in outlying areas; 2) determining the vacant areas for construction purposes; 3) conducting time series analysis to determine changes in urban areas (change detection); 4) analyzing and monitoring urban landscapes to determi ne urban sprawl; 5) monitoring urban areas pollution such as air or water pollutions; 6) identifying the revenue boundaries of villages on the remote sensing images; 7) monitoring and determining the characterization of urban heat island; and 8) developing systems for traffic manage ment. Investigating and predicting urban expansion may help prevent several social and environmental problems caused by urbanization such as pove rty, air pollution, and rising surface t emperatures ; although the pr evious chapter found an Urban C ool Island for Kuwait Identifying the extent, direction, and location of existing and future urban areas, therefore, would enable urban planners and decision makers to better man a ge urban expansion trends. There are many different remote s ensing models that have been employed to in vestigate urban land use. A model is selected and implemented within the social explaining the relationship between one or more fac tors (Fragkias and Seto, 2007). Shoemaker et al (2004) defined a model in the social sciences as a simple representation of a process or an object of reality, which highlights elements of the object and the connection between them. Models are utilized for studying relationships and interactions amongst system elements and for creating testable hypotheses and
89 predictions in regards to patterns and mechanisms (Fragkias and Seto, 2007). Within urban remote sensing research models are utilized to study urban expansion such as the growth of urban areas, transportation infrastructu res, and topographic con ditions (Henriquez et al., 2006) based on past land use changes determined from remotely sensed images. Cellular Automata Markov (CA_Markov) along with Markov c hain analyses is commonly implemented for predicting change within categorical data sets using mathematical formulation s of probability transitions ( Pontius an d Malonson 2005). Sun et al (2007 ) utilized Markov Chain analysis and Cellular Automata analysis to predict land use change and urban sprawl. The study imp lemented six Landsat TM and ETM + images for the city of Calgary, Canada from 1985 to 2001. Data for road networks w ater bodies, and parks were used along with the remote ly sensed images to produce an object oriented land use classification. ECognition software was employed to classify the images, and t he classified images were later utilized in Markov Cha in and Cellular Automata analyse s to predict future changes. The study impl emented 1985 and 1992 images in the pred iction model to predict 1999; the 1999 image was then compare d to the actual land use map in order to validate the prediction model result. The validated model used the 1990 and 2000 classified images to predict the target year of 2010 and the study areas. The study concludes that will cont inue to increase necessitating the development of urban growth policies. In this study, Cellular Automata Markov (CA_Markov) and Markov chain analyses was employed to predict future urban land use change. Remote sensing research
90 implementing prediction models to project future changes in developing countries is limited, and there are no studi es employing such a model in the Middle Eastern region. Population growth is considered in this study to be the main factor driving urban expansion in Kuwait. The objective of the study is to employ a CA_Markov model to understand future potential urban ex pansion, which can help decision makers in future planning. The model was implemented to predict future urban land use change in Kuwait for 2015, 2020, 2025, and 2030. There are two questions the study aimed to answer: 1) Based on the current trends of pop ulation growth and urban expansion, how will urban land use increase in Kuwait from 2010 to 2030? 2) Which pair of dates best predicts current changes and so is most likely to predict future changes with the most confidence according to the validation proc ess ? Methodology Study A rea Kuwait is situated in the eastern part of the Middle Eastern region, and the Arabian Gulf (Persian Gulf) borders the eastern side of the country. The Kingdom of Saudi Arabia is the western neighbor of Kuwait and Iraq is the northern neighbor (Figure 4 1). With six governmental di stricts Al Asma, Hawalli, Al Ahmadi, Al Farwania, Al Jahra, and Mubarak A lkabier (Figure 4 2), the country is mostly covered by desert type land leaving esert include coastal hills, coastal flat land southern flat desert, and Al Dibdibba gravelly plain (Khalaf et al., 198 4; Kwarteng and Chavez, 1998). Kuwait extends between latitudes (28 and longitudes ( 46 esert Summers are hot and dry, winters are short and warm, and rainfall is minimal. Sum C in the shade,
91 17.7 C and the avera ge annual precipitation is 1 00 mm (Algharib, 2008; Kuwait Government Online, 2010; and Kwarteng and Chavez, 1998). Kuwait is ranked the second most urbanized country in the world, where the percentage of populatio n living in urban areas is 97.6% (United Nations Population Division, 2 003; Mahgoub, 2008). The country covers 17 ,818 km with a population around 3.3 million (The World Fact Book, 2010; Kuwait Government Online, 2010). The urban area of Kuwait is located along the central and southern coastlines in a small area equivalent to 6% of the total land area. The reason for this concentr ation of population is the government, which assigns specific areas for urban use an d designates the rest of the country for petroleum producti on. The country is considered to be one of the wealthiest countries in the world, where pe troleum productions generate 95% of the government income (The World Fact Book, 2010). In terms of the per capita GDP, Kuwait is fourth among Middle Eastern countries after UAE, Qatar and Israel (World Factbook, 2002; Mahgoub, 2008). Data Preparation This study employed five Landsa t (TM, and ETM + ) images obtained for the study area in 1986, 1990, 2000, 2005 and 2010. Though this study attempted to maintain a five year time wind ow between the sate llite images, the availability of images covering the study area was limited As the 1985 image was missing a part of the study area due to a shift, the 1986 Landsat image, obtained on February 17 th was the first image used in the study analyses. The seco nd image Landsat TM was obtained on September 8, 1990. A Landsat TM image was obtained for 1995 in an effort to maintain the five year window between images. This image was eliminated from the analyses to e nsure the quality of classification however, beca use the study area wa s affected by the oil spill
92 caused by Saddam Hussein occupation in 1990. Therefore, the third Landsat TM image w as acquired on August 18, 2000. Landsat ETM + images for 2005 and 2010 were acquired and corrected using an adaptive local ized histogram matching technique with specific scene based alterations through a private agency due to the SLC problem The Landsat ETM + 2003 image was used as the base image in the gap filling process for the 2005 image, which was later used as the base image for gap filling the 20 10 image. The gap filled area in the 2005 image was 27.6% of the study area and 24.6% in the 2010 image. The fo urth Landsat ETM + image was obtained on July 23, 2005, and the last Landsat ETM + image was acquired on March 15, 2010. These satellite images were obtained during the dry season to make sure that all images have a cloud free presentation. The data preparation processes for the images of this study include calibration, registration, and classification. All images were refe renced and registered in ERDAS I MAGINE and the Arabian Gulf bodies of water were masked from the images. Also, a GIS layer indicating urban development areas designated by the government were used to subset all images. They were registered to the U TM coordinate system ( zone 19 N, WGS 1984 datum) with RMS error less than 10 meters. In order to minimize atmospheric variation s such as atmospheric effects and solar elevation and curve, atmospheric calibration met hods were performed on 1986, 1990, and 20 03 images. Image classification was undertaken on all images for two land use and land cover cla sses: built up and non built areas. The built up area s contain urban and suburban area s comprised of built structures such as malls, an airport, residences, and many others. Supervised classification and accura cy assessment were performed on all input
93 images in ERDAS IMAGINE The five classified images accuracies were above 90 % which indicates the classified imag es are acceptable for land use change analyses (Weng, 2002) such as Markov chain analyses and CA_Markov model (Figure 4 3, Figure 4 4, and Table 4 1). CA_Markov m odel Predicting the future changes of the land use classes in Kuwait requires an adapt at ion of a well known prediction model within the urban remote sensing field. The CA_Markov model is a powerful approach to spatial and temporal dynamic modeling that investigates changes in land use and land cover by incorporating GIS and remotely sensed data (Li and Reynolds, 1997; Silverton et al., 199 2; Kamusoko et al., 2009). Moreover, Brown et al (2002), asserted that a reliable prediction model offers: 1) an output specifying the amount of change in land use and land cover, 2) a spatial re presentation of the change in land use and land cover patter ns, and 3) a re presentation of the location for predicted change s. A CA_Markov model, used to predict and detect future spatial variation in patterns of land use and land cover changes (Sang et al., 2010), was selected for this study based on these specifi cations. A CA_Markov model incorporates prior analyses generated from Markov chain analyses, a mathematical tool that produces a set of probability transition matrices. This set of probability matrices indicat es the probability of each pixel to change from one state to another over the given time step In order to generate the se probability transition matric es, an interval of time between the two input data should be determined. The probability transition matrix is the first output of Markov chain analyses, which is a table ranging from zero to one that indicates the possibility of change (Joo et al., 2010). The other outputs are a probability transition area and a suit ability raster group layer
94 representing each land use and land cover class with its probab ility values. These Markov chain analyses outputs are used as the input data for CA_Markov model to generate future projection s for land use and land cover classes. T he probability transition matric es are employed by a CA_Markov model to determine the over time changes in the number of cells by utilizing cellular dynamics (Verburg et al., 2004). According to Pontius and Malanson (2005), the cellular dynamics or cellular automation component of the model is important to generate a probability transition of o ne pixel as a function of neighboring pixels. The predicting process also uses the probability transition area and a suitability raster group layer, created by the Markov chain, to generate a future l and use and land cover map indicating the spatial change s in the study area for a specific point in time (Mubea et al., 2009). The final output of the model is a future land use and land cover map base d on the previous changes between the two input data such as remotely sensed images. T he Landsat classified images of Kuwait for 1986, 1990, 2000, and 2005 were implemented as pairs into a CA_Markov model to predict the future likelihood of changes in land use for 2015, 2020, 2025, and 2030. The pairs used in the initial Markov chain analys es and CA_Markov model include (1986 1990), (1986 2000), (1986 2005), (1990 2000), (1990 2 005), and (2000 2005). T he first pair model used 1986 as the base land cover map and 1990 as the later land cover map in the Markov chain analyses to predict the 2010 map. T he period between the two input images was four years the prediction time frame from t he later image was twenty years, and t he outputs of Markov chain analyses were utilized in a CA_Markov to produce the 2010 map. The 1986 image was the base map in the second pair model and 2000 was t he
95 later image with a period of fourteen years between the inputs and ten years for the projection. The third pair model incorporated the 1986 image as the base map and the 2005 image for the later map with a period of nineteen years between the two input s and five years for the predic tion process. The 1990 image was the base map in the fourth pair model and the 2000 image was the later map with ten years period fo r both time frames. The fifth pair model implemented the 1990 image as the base map and 2005 for the later map, with a period of time of fifteen years between the inputs and five years for future projection. The last pair model included the 2000 and 2005 images with fi ve years window between images and for predi ction. All six pair models gene rated a prediction map for 2010, and the 2010 image was used for the validation and evaluation process es Validation Employing a predic tion model to predict future changes in land use and land cover patterns require s a valida tion process to measure the capability of such a model to pr edict the future. Validation modules can be used to measure the level of agreement between tw o categorical maps: a reference map and a comparison map. The reference map is a classified and validated remotely sensed image for a particular area and year The comparison map is the predicted map that is generated using pairs of images in a CA_Markov model. This study incorporated a validation module to measure the accuracy of the six pair models (1986 1990), (1986 2000), (1986 2005), (1990 2000), (1990 2005), and (2000 2005 ) The predicted maps of 2010 were used as the comparison input in the validation module, and the classified image of 2010 was the reference input.
96 The predicted map s were vali dated against the 2010 classified images in IDRISI Taiga using the val idation tool in order to determine the best pair model A ccording to Pontius and Malason (2005), the re are two major questions this valid ation module attempts to answer: 1) H ow well do the comparison and reference maps agree in term s of the amount of cells for each lan d use class? 2) How well do both maps agree in t erms of the location of cells? I n order to a nswer these two major questions, t he validation module produces three Kappa ind exes including Klocation, Kno, and KlocationStrata, The Klocation indicator measures the level of agreement between the in put data based on location only, Kno indicator measures the overall agreement, and KlocationStrata index measures the level of agreeme nt based on quantity (Pontius and Malason, 2005). T he most accurate pair model was then employed to predict future land use map s of Kuwait for 2015, 2020, 2025, and 2030. Results and Discussion One objective of this research was to determine the best pair model, based on the validation process, to predict future land use and land cover change s in Kuwait for 2015, 2020, 2025, and 2030 The validation process indicated an overall agreement (Kno) range from 0.8958 to 0.9660. Table 4 2 shows the r esults of the validation module for all prediction models. The study determined the most accurate pair model for future prediction based on the values of Kno, Klocation, and KlocationStrata. The most accurate model (1990 2005) was employed to forecast futu re built up and non built up areas in Kuwait for 2015, 2020, 2025, and 2030. Identifying the future changes of land use and land cover in Kuwait required change detection analyses. The future maps were combined with the 2010 classified image in order to provide a clear indication of future patterns of built up and non built
97 areas The change detection analyses generated a table and maps indicating future changes, using the from to technique. Table 4 3 demonstrates the cha nge trajectories of future land use and land cover in Kuwait, and F igure 4 6 demonstrates the new built up areas (highlighted in red) by comparing the 2010 image to the future map According to Table 4 3, the classified pixels of non built areas decreased from 573,704 in 2010 to 532,763 in 2015, which indicates a loss of 37 km equivalent to 7% of the total area. The classified pixels of non built areas will continue to decline in 2020 to 514,471, which indicates an additional loss of 16 km equal to 3% of the total area. The classified pixels of non built areas will decrease to 496,478 in 2025, which indicates a loss of another 17 km equivalent to 3% of the total area. The classified pixels of non built areas will decline in 2030 to 480,620, which indicate s more loss of 14 km equal to 3% of the total area. The loss of non built areas will result from land transformations of non built areas to built up areas. This study determined built up areas are expected to increase from 2010 to 2015. According to chang e detection analyses, built up area s will increase by 37 km from 2010 to 2015, which is equivalent to 7% of the study area. This amount of land t ransformation from non built areas to built up area s in 2015 will take the shape of infill development (Figure 4 6). The map indicates that most of the new dev elopment will be located in Al Asma, Al Jahra, and part of Mubarak Alkabier. Al A sma is the oldest district in Kuwait and the nearest residential district to the capital of Kuwait. The proximity of the distr ict to the business and government activities of Kuwait City makes Al Asma a desirable place to live and increases the demand for and the prices of land. Therefore, increased urban development in Al Asma derives mostly from the
98 redevelopment process wherea s the urban development of the Al Jahra and Mubarak A lkabier districts comes mostly from the establishment of new residential areas. The predicted map of Kuwait for 20 20 highlights future changes to non built and built up area s (Figu re 4 6). The percentage of change is expected to reach 10 % in 2020 as land transformations of non built area s to built up area s continue to increase in Kuwait Non built area s are predicted to decrease by an area of 53 km from 2010 to 2020.These future changes are mainly shown to be high in the northern part of the st udy area, and t he location s of future land transformation s are distributed over the six Kuwaiti districts with some districts experiencing more transformation s than other s such as Al Asma, Al Jahra, and Mubarak A lka bier. More over, new urban development is also anticipated to t ake place in the eastern and southeastern parts of the study area. The trend of increased urban expansion in Kuwai t is likely to continue through 2025 and 2030. The 2025 prediction map indicates an expected change of 70 km since 2010 (Table 4 3), and the built up areas are shown to expand by 3% from 2020 to 2025. The location of these transformations will be similar to those in previous years. It is projected, however, that highways will play a major role in the distribution of f uture built up area s The prediction map of 2025 depicts new urban development s along highways in the eastern and central parts of the study area. According to the 2030 prediction map, changes to built up areas will reach 84 km equivalent to 16% of the total area from 2010, and will be mostly allocated in the c entral and northeastern areas. Furthermore, the study area is expected to experience increased land transformation s distributed throughout all six districts from 2 010 to 2030, and built up area s in Kuwait are expected to exceed the total area of non built up area s (F igure 4 5). The findings of the prediction
99 model indicate Kuwait will continue to experience huge land transformations in the future. This projected urban expansion in Kuwait is possibly related to an estimated population increase of 213,864 between 2010 and 2020 (UN Population Division, 2004). By using the most accurate map, representing land transformations in Kuwait from edictions for the future reflect the changes of the past. Most Arabian Gulf countries and many developing countries have experienced this trend of land transformation, known as urban expansion, since the late 20 th Century. Algharib (2008) conducted a study in Kuwait employing Landsat TM images to monitor and identify changes in land use and land cover between 1989 and 2001. The study detected an expansion of urban areas by 10% of the total area located in the south, west, and southwest portions of the study area. These finding s support the CA_Markov model outputs of this study which similarly indic ate fut ure transformation in the south, west, and southwest areas of Kuwait Moreover, Yagoub (2004) conducted a study in Al Ain, UAE to monitor urban growth from 1976 to 2000. He used remote sensing satellite images and GIS data to meet the study objective s The findings of the study built up areas increased by 5,352 hectares. Furt hermore, a study utilizing remote sensing and GIS to investigate urban expansion in Amman City, Jordan by Al Rawashdeh and Saleh (2006) found an increase of 163 square kilometers in the city from 1918 to 2002, which indicates urban areas expanded by 509%. Though there are multiple studies employing remote sensing and GIS to investigate urban land use changes in the Arabian Gulf and the Middle Eastern region, studies employing models to predict future urban changes in this region are almost non existent.
100 The re are several studies exemplifying the usefulness of employing prediction models to determine future urban changes in developing regions outside of the Middle East. For example, Kamusoko et al (2009) conducted a study in Zimbabwe employing CA_Markov mode l to investigate future land use and land cover changes in a part of the developing world for 2010, 2020, and 2030. The study implemented four land use maps including 1973, 1989, and 2000 maps for the prediction model and a 2005 map for model validation. T hey predicted Zimbabwe will experience severe land degradation as a result of the land transformations caused by increased population and concluded CA_Markov is useful for predicting the future. Likewise, Sang et al. (2010) utilized CA_Markov model to fore cast land use and land cover in Fangshan district, China. The study implemented land use GIS data for 2001, 2006, and 2008 in a CA_Markov model to predict 2015, and found urban areas will expand by 8% of the total area from arable land. This study also con cluded the usefulness of CA_Markov models for the forecasting of future changes. Furthermore, according to Opeyemi (2006), a CA_Markov model is a robust computer modeling technique useful for predicting land use and land cover changes that are hard to desc ribe. Computer modeling utilized to investigate current changes and predict future changes of landscape, has become one of the most important tools in the urban remote sensing field (Herlod et al., 2003). The accessibility of advanced computer software an d the tremendous reso urces of spatial data encourage the implement ation of such The process of computer modeling can be a valuable instrument for sustainable planning and for evaluating social and environ mental consequences of landscape fragmentation (Cabral and Zamyatin, 2006;
101 Brown et al., 2002; Henriquez et al., 2006). Furthermore, implementing remote sensing and GIS models to predict the future, especially in developing countries, may provide valuable information for sophisticated urban development. Despite the effectiveness and usefulness of CA_Markov model as a tool for predicting future land use and land cover change, there are limitations associated with the application of this model. The CA_Markov model predicts future changes based on current and past land transformations. The model assumes a trend of land transformation, which has happened in the past, will persist in the future creating a continuous pattern of land transformation over time (Sun e t al., 2007) Therefore, the CA_Markov model is limited in that it does not incorporate new trends of land transformation or recognize legislative changes of policy. Another limitation of the CA_Markov model is the model does not involve socioeconomic info rmation in the prediction of future land transformations. Incorporating socioeconomic considerations may assist in the production of better future land use and land cover projections. Utilizing logistic regression, object oriented, and other types of land change models may be useful in minimizing or eliminating these limitations. Therefore, future research will incorporate finer resolution remote sensing images along with a more advanced land use model from the land change science field in order to circumve nt these limitations. Summary Markov chain analyses and CA_Markov models were successfully utilized and validated in this study to predict future urban development in Kuwait. The most accurate pair of dates, represented by the validated map for 1990 to 200 5, was implemented in the computer model to predict future urban land use change for 2015, 2020, 2025, and 2030. The general outcome of this model indicates that urban areas will continue to
102 expand through 2030. Based on the current trends of land transformation and population growth in the study area b uilt up area s are expected to exceed the total area of non built up area s by 2030, a significant shift in dominance. The findings of this research, therefore, pro vide valuable information for policy makers and urban planners in Kuwait as well designed urban plans and development policies are needed to manage future urban expansion. For example, if current policies remain in place as the population continues to grow, then the reduced amount of open spa ces used for vegetation will negatively impact environmental conditions. This study subsequently urges the Kuwaiti Government to designate new areas for sustainable urban development. Furthermore, the lack of research employing a predicting model in the Mi ddle East expands the usefulness of this methodology beyond the borders of Kuwait to other Arabian Gulf countries
103 Table 4 1. Classification Accuracy Assessment of 1986, 1990, 2000, 2005, and 2010 classified images of Kuwait. Classified Image Year Overa ll Classification Accuracy Overall KAPPA Statistic 1986 90.06% 0.8019 1990 94.21% 0.8846 2000 93.57% 0.8718 2005 92.28% 0.8463 2010 91.64% 0.8336
104 Table 4 2. Predicted models accuracy for each pair model. Model Pair Model Type Random Yours Perfect 1986 1990 % Correct 33.33% 93.05% 100.00% Improvement ------59.72% 6.95% Kno: 0.8958, Klocation: 0.8897, and KlocationStrata: 0.8897. 1986 2000 Model Type Random Yours Perfect % Correct 33.33% 96.06% 100.00% Improvement ------62.73% 3.94% Kno: 0.9410, Klocation: 0.9375, and KlocationStrata: 0.9375. 1986 2005 Model Type Random Yours Perfect % Correct 33.33% 97.58% 100.00% Improvement ------64.25% 2.42% Kno: 0.9637, Klocation: 0.9632, and KlocationStrata: 0.9632. 1990 2000 Model Type Random Yours Perfect % Correct 33.33% 96.41% 100.00% Improvement ------63.08% 3.59% Kno: 0.9461, Klocation: 0.9563, and KlocationStrata: 0.9563.
105 Table 4 2 Continued 1990 2005 Model Type Random Yours Perfect % Correct 33.33% 97.74% 100.00% Improvement ------64.41% 2.26% Kno: 0.9660, Klocation: 0.9706, and KlocationStrata: 0.9706. 2000 2005 Model Type Random Yours Perfect % Correct 33.33% 97.49 % 100.00% Improvement ------64.16% 2.51% Kno: 0.9624, Klocation: 0.9706, and KlocationStrata: 0.9706.
106 Table 4 3. Future changes in Kuwait from 2010 to 2015, 2020, 2025, and 2030. Change in pixel numbers and corresponding area from 2010 to 2015 Class type Classified pixels in year Classified pixels in year Change in Change in % 2010 2015 Non Built 573,704 532,763 37 7% Built Area 405,154 446,095 37 +7% Change in pixels numbers and corresponding area from 2010 to 2020 2010 2020 Non Built 573,704 514,471 53 10% Built Area 405,154 464,387 53 +10% Change in pixels numbers and corresponding area from 2010 to 2025 2010 2025 Non Built 573,704 496,478 70 13% Built Area 405,154 482,380 70 +13% Change in pixels numbers and corresponding area from 2010 to 2030 2010 2030 Non Built 573,704 480,620 84 16% Built Area 405,154 498,238 84 +16%
107 Figure 4 1. Middle Eastern Countries. Source: http://letustalk.wordpress.com/2008/07/19/saturday obama arrives in kabul afghanistan/
108 Figure 4 2. Kuwait dist ricts and boundaries Source: Kuwait municipality
109 Figure 4 3. Flow chart illustrating steps undertaken for the Markov chain analyses and CA_Markov model creation and validation
110 Figure 4 4. The classified remotely sensed images for years: a ) 1986, b ) 1990, c ) 2000, d ) 2005, and e ) 2010.
111 Figure 4 5. Future land use changes predicted from the model (1990 2005) in Kuwait. 0 100 200 300 400 500 600 2010 2015 2020 2025 2030 Square Kilometers Year Future Land use Changes in Kuwait Non-Built Area Built Up Area
112 Figure 4 6. The predicted land use and land cover maps for years: a ) 2015, b ) 2020, c ) 2025, and d) 2030 based on the input image date 1990 2005.
113 CHAPTER 5 CONCLUSION Several research fields utilize adv anced methodologies and technologies to monitor, analyze, and predict future land use and land cover transformations. Geography and LCS research communities examine land transformations, across a period of time, to produce valuable information regarding th e types, amounts, and locations of changes. GIS and RS are technologies that have been utilized for several urbanization. Identifying changes of land use and land cover may provide the information necessary to answer many questions about the relationship between land transformation, population, and social and environmental problems. The overall goal of thi s dissertation is to investigate the similarities and differences o f urbanization in dev e loped and developing countries. This dissertation designed three different papers to measure the level of future urban land transformation associated with urban population growth in Alachua County and Kuwait. Measuring the differences between these two countries portrays the impact human activities have on the natural environment in both study areas T he utilization of advanced technologies and satellite images, to monitor the expansion of urban areas in Alachua County and Kuwait, enab led this research to analyze past and current land transformations and to predict land transformations of the future. Nine Landsat TM and ETM+ satellite images were classified according to the type of land use and land cover occupation. These classified im ages were tested for accuracy and subsequently incorporated into three techniques of analyzing urban land use change: CA Markov ( a land change prediction model ) post classification change
114 detection, and thermal band analysis. CA_Markov utilized classified images in the first and the last papers to forecast future urban land transformations in Alachua County and in Kuwait. The classified images of Kuwait were then implemented in to post classification change detection in the second paper to determine past cha ng es of urban land use. Next, the classified images for Kuwait were employed in thermal band analysis to detect changes in Land Surface Temperature s (LST) Furthermore, the three interrelated studies of this dissertation utilized these techniques to genera te outputs, which can be u sed to compare urban land transformations in developed and developing countries The first paper of this dissertation employed CA_Markov model to predict urban land use change in Alachua Cou nty. The model incorporated four Landsat TM and ETM+ satellite images for 1993, 1998, 2003, and 2008 into the prediction and validation process es to generate future urban land transformation maps for the selected dates. The validation process determined the 1998 2003 pair model as the best model to predict future land use and land cover changes. The model was used to predict 2013, 2018, and 2023 land use maps based on current land cover patterns. The model indicated future expansion of urban land use areas over vegetation land cover will increase from approximately 3% to 7% from 2008 to 2023. According to this model, the percentage of area that will be converted from vegetated areas to built up areas is 3.16% from 2008 to 2013, 1.73% from 2013 to 2018, and 1.71% from 2018 to 2023. The total amount of land that will transform from vegetation land cover to built up areas, will reach 6.60% by 2023. Furthermore, the prediction model suggests the new urban developments will mostly occur in close proximity to current urbanized areas. The
115 predicted maps i dentify the western and central portions of Alachua County as the projected areas of future land transformation. The second paper of this dissertation was designed to investigate urban land use change s and the effect s of urban expansion on Land Surface Tem peratures (LST) in Kuwait from 1986 to 2010. The paper incorporated five Landsat TM and ETM+ satellite images for 1986, 1990, 2000, 2005, and 2010. The analyses detected that built up areas increased from 209.7 km 2 in 1986 to 362.6 km 2 in 2010. The change trajec tories analyses determined the percent increases of built up areas to be 4.7% from 1986 to 1990, 3.8% from 1990 to 2000, 3.5% from 2000 to 2005, and 5.3% from 2005 to 2010. Based on the change detection map, land transformations were generally locate d in the central and southern portions of Kuwait. This study relates this tremendous amount of Furthermore, t he analysis indicated the surface temperatures of non built areas were higher than bui lt up areas. The built up areas were cooler than non built areas in 1986 by 1 .29 F. The difference between built up and non built areas increased i n 1990 to reach 2 .00 F and 3.81 F in 2000 supported with similar findings in a study conducted by Kwarteng and Small in 2005. The third paper of this dissertation implemented CA_Markov model to predict urban land transformation s in Kuwait. The model included f ive Landsat TM and ETM+ satellite ima ges for 1986, 1990, 2000, 2005, and 2010 into the prediction and validation process es to produce information regarding the best future urban land transformation for the selected dates. The validation process determined the 1990 2005 pair model as the best model to predict the future of land use and land cover change. The best model
116 was identified by testing the level of agreement between the predicted maps of 2010 for all pair models a nd the classified 2010 image. The 1990 2005 model was used to predict land use maps for 2015, 2020, 2025, and 2030 based on current land cover patterns. The model indicates the total future expansion of urban land use areas will range from 7% to 16% from 2 010 to 2030. The percentage of area expected to convert from non built to built up areas from 2010 to 2015 is 7%. The amount of land that will transform from non built areas to built up areas, will reach 10% in 2020 13% in 2025, and 16% in 2030. The 1990 2005 pair model suggested built up areas will exceed non built up areas by 2030 with an estimated land change of 84 km 2 from 2010 to 2030. Furthermore, the prediction model indicates the new urban developments will mainly occur near existing urbanized area s. The northern portion of Kuwait is expected to experience most of the future changes. The eastern and southeastern parts of the country, however, are also expected to experience future land transformation. Th e overall findings of research indicates that urban expansion will continue to happen in Alachua County and Kuwait within the coming decade. Population growth was determined to be the main force of land transformation in both study areas. High rates of population growth from natural increase and immig ration expand existing urban areas and, therefore, cause higher rates of land consumption and land transformation. The future transformations from other land cover classes to urban land use class es need attention from the governments in both stu dy areas. T he research suggests Alachua C ounty government should c onsider buying more vacant land in order to manage the future expansion s of ur ban areas. The government sh ould also encourage the developer to build the new residential dwelling vertically to minimize the amount of
117 land transformation to accommodate more population at the same time. Moreover, the research suggests the Kuwaiti government should designate new areas for urban development in order to avoid the future social and environmental problem s relate d to urban expansion. Thermal band analysis was undertaken in this research to detect the influence of land transformation on the LST in Kuwait for 1986, 1990, and 2000. The outcomes of thermal band analysis were integrated into a map that spatially repres ents the differences between Land Surface Temperatures (LST) Surprisingly, the urban areas were cooler in surface temperatures comparing to the surrounding areas (deserts). This gh urbanization and urban expansion is considered by many researchers to have a research proved the opposite to be true in desert environments. Despite the valuable insight pr ovided by the findings of this dissertation for public authorities in Alachua County and Kuwait, there are limitations associated with the techniques employed. The first limitation is related to the level of spatial resolution of the satellite images, which were classified and incorporated into change detection and the prediction model. Th is dissertation utilized medium spatial resolutions images, based on the capability of this research, where high spatial resolution image s might have increase d the acc uracy of the analyses outcomes The second limitation is linked to the prediction model as it assumes the trend s of land use and land cover change remain at a consistent rate of change in the future (Sun et al., 2007) In other words, the model does not ta ke into consideration new policies or deviating land transformation trends
118 during the prediction processes. Likewise, the CA_Markov model is limited in that it does not include changing social, economic, and environmental factors in the prediction of futur e l and use and land cover changes. T o circumvent these limitations of CA_Markov, alternative advanced methodologies such as LUMOS, PUMA, or DSSM models will be employed in future research. The Land Use Modeling System (LUMOS) is a combination of a cellular automata model and an economic model that is utilized to predict future land use patterns. The Predicting Urbanization with Multi Agents (PUMA) model incorporates social and economic information into the prediction process to predict future urban land tra nsformations. The Dynamic Settlement Simulation Model (DSSM) is a cellular automata model combined with GIS raster datasets. Incorporating economic and social information into the prediction model will increase the reliability of the outputs generated and provide further direction for urban planners. In addition to utilizing alternative advanced methodologies, f uture directions for this research include employing finer satellite images and implementing more detailed GIS datasets to investigate urban land us e changes and urban Land Surface Temperatures (LST). The study area will be extended to consist of Kuwait as well as Saudi Arabia and other Arabian Gulf countries. Furthermore, urban developments will be divided into several additional land use categories such as residential, governmental, and business, to be assoc iated with LST. This future analysis will provide information regarding the type of urban land use causing the greatest effect on LST and will p redict the future LST of urban development based on the past and current value s. Future study will address
119 the environmental effects of land use and land cover changes in order to encourage proactive measures and prevent future environmental problems. Comprised of three studies, this dissertation examines the relationship between human activities and the natural environment, compares urbanization in developed and developing countries, and investigates the correlation between population growth and urban expansion. This research was the first to use a CA_Mark ov prediction model in Alachua County as well as Kuwait T he validation process indicates this model is an appropriate method of investigating the amount and location of future land transformation in both study areas Furthermore, this research is the most recent to use such methodologies as post classification change detection and thermal band analysis to study the impact of urban expansion in Kuwait. Surprisingly, urbanization was found to have a positive effect on Land Surface Temperatures (LST) in Kuwai t An Urban Cool Island was identified contrary to the expectation that urbanization necessarily causes an Urban Heat Island. Summarily, this dissertation determined urban expansion occurs in both developed and developing countries, the rate of change is l inked to population growth (which is substantially greater in developing countries), and additional efforts are necessary to minimize unplanned urban expansions. This dissertation concludes with several suggestions for urban planners, policy makers, and re searchers This study suggests Alachua County should invest more capital in p urchasing vacant land and generate new policies to mit igate unplanned urban expansion. Furthermor e, due to a lack of space, Alachua County should encourage developers to build vertically. Kuwaiti Government should similarly improve urban planning efforts in preparation for rapid population gro wth and future urban
120 expansions. The high rate of Population growth in Kuwait makes urban development necess ary to accommodate the future population. Therefore, the Kuwaiti Government should allot more space for urban development. Further inves tigation for such a phenomenon as Urban Cool Island is highly recommended by this research to understand the connection between urban development and Land Surface Temperatures (LST) in K uwait. Finally, it is suggested to extend the examination of the relationships between population growth, urban expansion, and the environment beyond the borders of Alachua County and Kuwait to study areas possessing similar characteristics.
121 LIST OF REFERENCES ABU AYYASH, A. 1980. Urban development and planning strategies in Kuwait. International Journal of Urban and Regional Research, 4(4), 549 571. ABU AYYASH, A. 1981. Planning development towns in Kuwait. GeoJournal, 5(3), 251 260. AKERLUND, U., KNUTH, L., RANDRUP, T. and SCHIPPERIJN, J., 2006. Urban and Peri Urban Forestry and Greening in West and Central Asia (in contribution to the Forestry Outlook Study for West an d Central Asia FOWECA) workshop, FAO, Rome, 5 7 April 2006, Proceedings [online]. AL RAWASHDEH, S. and SALEH, B. 2006. Satellite Monitoring of Urban Spatial Growth in the Amman Area, Jordan. Journal of urban planning and development., 132(4), 211. AL AWA DHI, T., 2007. Monitoring and Modeling Urban Expansion Using GIS & RS: Case Study from Muscat, Oman. ed. Urban Remote Sensing Joint Event, 11 13 April 2007 1 5. AL MUTAIRI, N. and KOUSHKI, P. 2009. Potential contribution of traffic to air pollution in the state of Kuwait. Am. J. Environ. Sci. American Journal of Environmental Sciences, 5(3), 218 222. AL RASHIDI, B., AMOS, L. and EL GAMILY, I., 2007. Utilization of remotely sensed data to detect anthropogenic impacts on sea surface temperature of Kuwait Bay Kuwait. ed. OCEANS 2007 Europe 18 21 June 2007, 1 5. AL SALEM, S. 2008. An Overview of the PM10 Polluiton Problem in Fahaheel Urban Area. Emirates Journal for Engineering Research, 13(3), 1 9. ALACHUA COUNTY GIS SERVICE CENTER, 2010. Alachua GIS Layer s [online]. Available from: http://arrow.acpafl.org/ServiceCenter/gis_main.aspx. ALGHARIB, S., 2008. Spatial Patterns of Urban Expansion in Kuwait City Between 1989 and 2001. (Master of Art). Kent State University. ALIG, J., KLINE, J. D. and LICHTENSTEIN, M. 2004. Urbanization on the US landscape: looking ahead in the 21st century. ALMATAR, M., 2008. Using Remote Sensing and GIS to Study Land Use and Land Cover Change in ALACHUA County, FL From 1993 to 2003. (Master). University of Florida.
122 BAHR, H., 2001. Image Segmentation for Change Detection n Urban Environments. In: DONNAY, J., BATNSELY, M., LONGELY, P. ed. Remote Sensing and Urban Anlysis. 1ED ed. 29 west 35th street, New York, NY 1001: Taylor & Francis Inc, 95 113. BELAID, M., 2010. Remote Sensing of Desert Cities in Developing Countries. In: RASHED, T. and JRGENS, C. eds. Remote Sensing of Urban and Suburban Areas. Springer Netherlands, 245 265. BERNSELY, M., MOLLER JENSEN, L. and BARR, S., 2001. Inferring Urban Land Use by Spatial and Structural Pat tern Recognition. In: DONNAY, J., BATNSELY, M., LONGELY, P. ed. Remote Sensing and Urban Anlysis. 29 west 35th street, New York, NY 1001,: Taylor & Francis Inc, 115 144. BERRY, B., 1976. Urbanization and counterurbanization. Beverly Hills, Calif.: Sage Publications. BERRY, B. 1980. Urbanization and Counterurbanization in the United States. The Annals of the American Academy of Political and Social Science, 451. BERRY, B. 1990. Urbanization. Urban Ecology 103 119. BERRY, B. 2008. Urbanization. Springer 103 119. BORNSTEIN, R. 1968. Observations of the Urban Heat Island Effect in New York City. Journal of Applied Meteorology 8. BROWN, D., GOOVAERTS, P., BURNICKI, A. and LI, Y. 2002. Stochastic Simulation of Land Cover Change Using Geostatistics and Genera lized Additive Models. PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING, 68, 1051 1062. BROWN, D., WALKER, R., MANSON, S. and SETO, K. 2004. Modeling Land Use and Land Cover Change. Land Change Science, 6, 395 409. CABRAL, P. and ZAMYATIN, A., 2006. Three La nd Change Models for Urban Dynamics Analysis in Sintra Cascais Area. 1st EARSeL Workshop of the SIG Urban Remote Sensing [online]. CHEN, A. 2006. Urbanization in China and the Case of Fujian Province. Sage Publications 32(1), 99 130. COHEN, B. 2004. Urban Growth in Developing Countries: A Review of Current Trends and a Caution Regarding Existing Forecasts. World development., 32(1), 23. DEWIDAR, M. 2004. Detection of land use/land cover changes for the northern part of the Nile delta (Burullus region), Egyp t. International Journal of Remote Sensing, 25(20), 4079 4089.
123 DONNAY, J., BARNSELY, M. and LONGLEY, P., 2001. Remote Sensing and Urban Analysis. In: DONNAY, J., BATNSELY, M., LONGELY, P. ed. Remote Sensing and Urban Anlysis. 29 west 35th street, New Yor k,NY 1001: Taylor & Francis Inc, 3 18. DOUGLAS, I. 1994. Human Settlements. (Workshop). DOUGLAS, I., HUGGETT, R. and ROBINSON, E., 1996. Companion encyclopedia of geography : the environment and humankind. London; New York: Routledge. DOYGUN, H. and ALPHAN, H. 2006. Monitoring Urbanization of Iskenderun, Turkey, and its Negative Implications. Environmental Monitoring and Assessment, 114(1 3), 1 3. DUDA, O., HART, E. and STORK, G. 2001. Pattern Classification. New York: John Wiely & Sons 394 452. ELHADI, E. and ZOMRAWI, N. 2010. Change Detection Analysis By Using Ikonos and QuickBird Imageries. Nature and Science, 7(10), 25 31. ELVIDGE, C., WAGNER, T., RYZNER, R., VOGELMANN, J., GOETZ, S., SMITH, A., JANTZ,C., SETO, K., IMHOSS, M., WANG, Y., MILESI, C. and NEMANI, R. 2004. Urbanization. Land Change Science 315 328. FORD, K., 1979. Remote sensing for planners. New Brunswick,NJ: Rutgers University Center for Urban Policy Research FRAGKIAS, M. and SETO, K. 2007. Modeling urban growth in da ta sparse environments: a new approach. Environment and planning. B, Planning & design., 34(5), 858. FUSSELL, J., RUNDQUIST, D. and HARRINGTON, J. 1986. On Defining Remote Sensing. Photogrammetric Engineering & Remote Sensing, 52(9), 1507 1511. GELBARD, A. HAUB, C. and KENT, M., 1999. World population beyond six billion [online]. Washington, D.C.: Population Reference Bureau. GIRARD, L., CERRETA, M., TORO, D. and FORTE, F., 2007. The human sustainable city: Values, approaches and evaluative tools. In: DEAK IN, M. et al. eds. Sustainable Urban Development 2: The Environmental Assessment Methods. London:Routledge: Hobeken: Taylor & Francis, 65 93. GLUCH, R. 2002. Peer Reviewed Articles Urban Growth Detection Using Texture Analysis on Merged Landsat TM and S POT P Data A two step texture analysis using digitally merged TM and SPOT P data produces a binary "built/non built" image identifying urban growth. Photogrammetric engineering and remote sensing., 68(12), 1283.
124 GREEN, M., SCHWEIK, M. and RANDOLPH, C., 2 005. Retrieving Land Cover Change Information from Landsat Satellite Images by Minimizing Other Sources of Reflectance Variability [online]. GUTMAN, G., 2004. Land change science : observing, monitoring and understanding trajectories of change on the Earth 's surface. Dordrecht; London: Kluwer Academic Publishers. GUTMAN, G., JANETOS, ANTHONY C., JUSTICE, CHRISTOPHER O., MORAN, EMILIO F., MUSTARD, JOHN F., RINDFUSS, RONALD R., SKOLE, DAVID L., TURNER, BILLY LEE., COCHRANE, MARK A., BROWN, DANIEL G., WALKER, ROBERT., MANSON, STEVEN., and SETO, K. C., 2004. Modeling Land Use and Land Cover Change. Land Change Science. Springer Netherlands, 395 409. HALL, P., 2003. The sustainable city in an age of globalization. In: GIRARD, L. et al. eds. The human sustainable city : challenges and perspectives from the habitat agenda. Burlington, VT: Ashgate. HARTTER, J. and SOUTHWORTH, J. 2009. Dwindling resources and fragmentation of landscapes around parks: wetlands and forest patches around Kibale National Park, Uganda. La ndscape Ecology, 24(5), 643 656. HEIDEN, U., SEGL, K., ROESSNER, S. and KAUFMANN, H. 2007. Determination of robust spectral features for identification of urban surface materials in hyperspectral remote sensing data. REMOTE SENSING OF ENVIRONMENT NEW YORK 111(4), 537 552. HENRIQUEZ, C., AZOCAR, G., and ROMERO, H. 2006. Monitoring and modeling the urban growth of two mid sized Chilean cities. Habitat International, 30(4), 945 964. HEROLD, M., GOLDSTEIN, N. C. and CLARKE, K. C. 2003. The spatiotemporal for m of urban growth: measurement, analysis and modeling. REMOTE SENSING OF ENVIRONMENT NEW YORK 86(3), 286 302. HEROLD, M. S., J. and MULLER, A. G., S. 2002. Object oriented mapping and analysis of urban land use/cover using IKONOS data. EARSEL wide integration"Prague [online]. HODGSON, X., JENSEN, J. R., TULLIS, J. A., RIORDAN, K. D. and ARCHER, C. M. 2003. Synergistic Use of LIDAR and Color Aerial Photography for Mapping Urban Parcel Imperviousness. Photog rammetric Engineering & Remote Sensing, 69(9), 973 980. HU, Z. and LO, C. P. 2007. Modeling urban growth in Atlanta using logistic regression. Computers, Environment and Urban Systems, 31, 667 688.
12 5 HUANG, K. 2002. A Synergistic Automatic of Clustering Tech nique for Multispectral Image Analysis. Photogrammetric Engineering & Remote Sensing, 68(1), 33 40. ISSA, S. 2008. GIS Approach for Assessment of Human Footprint and Land Management: an Example in the United Arab Emirates. INTERNATIONAL ARCHIVES OF PHOTOGR AMMETRY REMOTE SENSING AND SPATIAL INFORMATION SCIENCES, 37(2), 591 596. JENSEN, J., 2005. Introductory digital image processing : a remote sensing perspective. Upper Saddle River, N.J: Prentice Hall. JENSEN, R., 2000. Remote sensing of the environment : a n earth resource perspective. Upper Saddle River, N.J.: Prentice Hall. JIANG, J. and TIAN, G. 2010. Analysis of the impact of Land use/Land cover change on Land Surface Temperature with Remote Sensing. Procedia Environ. Sci. Procedia Environmental Sciences 2, 571 575. JOHNSTON, R., GREGORY, D., PRATT, G. and WATTS, M., 2000. The Dictionary of Human Geography 4th ed. Malden, MA: Blackwell Publishing Ltd. JOO, Y., JUN, C. and PARK, S. 2010. Design of a dynamic land use change probability model using spatio temporal transition matrix. Lect. Notes Comput. Sci. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 6016 LNCS(PART 1), 105 115. KAGANOVA, O., AL SULTAN, F. and SPEAKMAN J. 2005. Overview of Real Estate Markets in Kuwait Journal of Real Estate Literature, 13(3), 287 302. KAMUSOKO, C., ANIYA, M., ADI, B. and MANJORO, M. 2009. Rural sustainability under threat in Zimbabwe Simulation of future land use/cover changes in th e Bindura district based on the Markov cellular automata model. Appl. Geogr. Applied Geography, 29(3), 435 447. KASARDA, J. and CRENSHAW, E. 1991. Third World Urbanization: Dimensions, Theories, and Determinants. Annual Review of Sociology, 17, 467 501. KA TES, R., TURNER II, B. and CLARK, W., 1993. The Great Transformation. In: TURNER II, B. et al. eds. THE Earth As Transformed By Human Action. Cambridge, New York, Melbourne: the Press Syndicate of the University of Cambridge, 1 17. KAYA, S. 2007. Multitem poral Analysis of Rapid Urban Growth in Istanbul Using Remotely Sensed Data. Environmental Engineering Science, 24(2), 228 233. KHALAF, F. I. and AL AJMI, D. 1993. Aeolian processes and sand encroachment problems in Kuwait. GEOMORPHOLOGY AMSTERDAM 6(2), 111.
126 KNOX, P. and MCCARTHY, L., 2005. Urbanization: an introduction to urban geography. Upper Saddlw River, NJ 07458: Pearson Education, Inc. KOOMEN, E., STILLWELL, JOHN., BAKEMA, ALDRIK., SCHOLTEN, HENK J.., KOOMEN, E. and STILLWELL, J., 2007. Modelling Land Use Change. In: BARLOW, M. et al. eds. Modelling Land Use Change. Springer Netherlands, 1 22. KUWAIT GOVERNMENT ONLINE, 2010. Population of Kuwait [online]. Available from: http://www.e.gov.kw/sites/kgoenglish/portal/Pages/Visitors/AboutKuwait/Kuwait Ata Glane_Population.aspx KWARTENG, A. and SMALL, C., 2005. Comparative analysis of thermal environments in New York City and Kuwait City. W27. KWARTENG, A. and SMALL, C., 2010. Remote Sensing of Urban Environmental Conditions. In: RASHED, T. and JRGENS, C. eds. Remote sensing of urban and suburban areas. Dordrecht ; London: Springer. KWARTENG, Y. and CHAVEZ, S. 1998. Change detection study of Kuwait City and environs using multi temporal Landsat Thematic Mapper data. International Journal of Remote Sensin g, 19(9), 1651 1662. LAMBIN, F., GEIST, H., VERBURG, P., KOK, K., PONTIUS, R. and VELDKAMP, A., 2006. Modeling Land Use and Land Cover Change. Land Use and Land Cover Change. Springer Berlin Heidelberg, 117 135. LI, H. and REYNOLDS, F., 1997. Modeling effe cts of spatial pattern, drought, and grazing on rates of rangeland: a combine Markov and cellular automaton approach. In: QUATTROCHI, D. A. and GOODCHILD, M. F. eds. Scale in remote sensing and GIS. Boca Raton, Florida: Lewis Publishers, 211 230. LI, X. an d YEH, O. 1998. Principal component analysis of stacked multitemporal images for monitoring of rapid urban expansion in the Pearl River Delta. International Journal of Remote Sensing, 19, 1501 1518. LO, C. and QUATTROCHI, D. 2003. Land Use and Land Cover C hange, Urban Heat Island Phenomenon, and Health Implications: A Remote Sensing Approach Photogrammetric engineering and remote sensing., 69(9), 1053. LOPEZ, E., BOCCO, G., MENDOZA, M. and DUHAU, E. 2001. Predicting land cover and land use change in the urb an fringe. LANDSCAPE AND URBAN PLANNING, 55(4), 271 285. LOVELAND, R., ZHILIANG, Z., OHLEN, O., BROWN, F., REED, C. and YANG, L. 1999. An Analysis of the IGBP Global Land cover Characterization Process. Photogrammetric Engineering & Remote Sensing, 65(9), 1021 1032.
127 LU, D., MAUSEL, P., BRONDIZIO, E. and MORAN, E. 2004. Change detection techniques. International Journal of Remote Sensing, 25(12), 2365 2407. LU, D. and WENG, Q. 200 6. Use of impervious surface in urban land use classification Remote Sensing of Environment, (102), 146 160. MACLEOD, D. and CONGALTON, G. 1998. A quantitative comparison of change detection algorithms for monitoring eelgrass from remotely sensed data. Photogrammetric Engineering & Remote Sensing, 64, 207 216. MAHGOUB, Y. 2004. Globalization and the built environment in Kuwait. Habitat international., 28(4), 505 519. MAHGOUB, Y., 2008. Kuwait: Learning from a Globalized City. In: ELSHESHTAWY, Y. ed. The Evolving Arab City: Tradition, Modernity & Urban Development. New York: Routl edge. MAKTAV, D., 2005. Remote sensing of urban areas : a selection of papers presented at the Third International Symposium on Remote Sensing of Urban Areas, Istanbul, Turkey, 11 13 June 2002. Abingdon: Taylor & Francis. MANNION, A., 2002. Dynamic World : LAND COVER AND LAND USE CHANGE. 198 Madison Avenue, New York, NY10016: Oxford University Press Inc. MATTHEW, R., 2009. Global environmental change and human security. Cambridge, Mass.: MIT Press. MEYER, W. and B., T. 1992. Human Population Growth and Glo bal Land use/Cover Change. Annual Review of Ecology & Systematics, 23(39). MEYER, W. and TURNER, B. 1995. Changes in land use and land cover: A global perspective. Ecology., 76(5), 1682. MORAN, F. and OSTROM, E., 2005. Seeing the forest and the trees : human environment interactions in forest ecosystems. Cambridge, Mass.: Mit Press. MORGAN, P. and CUSHMAN, F. 2005. URBANIZATION AND STREAM ECOLOGY Urbanization effects on stream fish assemblages in Maryland, USA. Journal of the North American Benthologi cal Society., 24(3), 643. MUBEA, K., NGIGI, T. and MUNDIA, C., 2009. Assessing Application of Markov Chain Analysis in Predicting Land Cover Change: A Case Study of Nakuru Municipality. GIS Development [online]. MULLER, R. and MIDDLETON, J. 1994. A Markov model of land use change dynamics in the Niagara Region, Ontario, Canada. Landscape ecology., 9(4), 151.
128 MUNDIA, N. and ANIYA, M. 2005. Analysis of land use/cover changes and urban expansion of Nairobi city using remote sensing and GIS. International Journ al of Remote Sensing, 26(13), 2831 2849. MUNROE, K., SOUTHWORTH, J. and TUCKER, M. 2004. Modeling Spatially and Temporally Complex Land Cover Change: The Case of Western Honduras. PROFESSIONAL GEOGRAPHER, 56, 544 559. NECHYBA, T. and WALSH, R., 2004. Urban sprawl. Cambridge, MA: Lincoln Institute of Land Policy, c2004. NETZBAND, M., STEFANOV, W. and REDMAN, C., 2007. Applied remote sensing for urban planning, governance and sustainability [online]. Berlin; London: Springer. Available from: http://dx.doi.org /10.1007/978 3 540 68009 3 NICHOLSON LORD, D. 1987. The Greening of the Cities. Routledge and Kegan Paul, London 270. NSIAH GYABAAH, K., 2009. Human Security as a Prerequisite for Development. In: MATTHEW, R. et al. eds. Global environmental change and human security. Massachusetts: Massachusette Institute of Technology, 237 259. OPEYEMI, Z., 2006. Change Detection in Land Use and Land Cover Using Remote Sensing Data and GIS (A case study of Ilorin and its environs in Kwara State. Project (Master of Scie nce). University of Ibadan. PHILLIPS, S., 1986. Factors influencing agricultural land use change: a study of Alachua County, Florida (Master). University of Florida. PONTIUS, G. and MALANSON, J. 2005. Comparison of the structure and accuracy of two land ch ange models. International Journal of Geographical Information Science, 19(6), 745 748. PRICE, J. 1979. Assessment of the Urban Heat Island Effect Through the Use of Satellite Data. Monthly Weather Rev Monthly Weather Review, 107(11). RAHMAN, A. 2007. Appl ication of Remote Sensing and GIS Technique for Urban Environmental Management and Sustainable Development of Delhi, India. Applied Remote Sensing for Urban Planning, Governance and Sustainability 165 197. RAHMAN, A., AGGARWAL, S. P., NETZBAND, M. and FAZ AL, S. 2011. Monitoring urban sprawl using remote sensing and GIS techniques of a fast growing urban centre, India. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 4(1), 56 6 4.
129 RANCHIN, T., WALD, L. and MANGOLINI, M., 2001. Improving the spatial Resolution of Remotely Sensed Images by Means of Sensor Fusion: AGeneral Solution Using the ARSIS. In: DONNAY, J., BATNSELY, M., LONGELY, P. ed. Remote Sensing and Urban Anlysis. 1ED e d. 29 west 35th street, New York, NY 1001, 2001: Taylor & Francis Inc, 22 37. RASHED, T., JRGENS, C. and BELAID, M. A., 2010. Remote Sensing of Desert Cities in Developing Countries. Remote Sensing of Urban and Suburban Areas. Springer Netherlands, 245 26 5. RIDD, K. and LIU, J. 1998. A comparison of four algorithms for change detection in an urban environment. Remote Sensing of Environment, 63, 95 100. SAGE, C., 1994. Population and Income. In: MEYER, W. and TURNER, B. eds. Changes in land use and land cov er: A global perspective. Cambridge, New York, Melbourne: the Press Syndicate of the University of Cambridge, 263 285. SANG, L., ZHANG, C., YANG, J., ZHU, D. and YUN, W. 2010. Simulation of land use spatial pattern of towns and villages based on CA Markov model. Mathematical and Computer Modelling, 54(3 4), 938 943. SCHNEIDER, C. and PONTIUS, G. 2001. Modeling land use change in the Ipswich watershed, Massachusetts, USA. Agriculture, ecosystems & environment., 85(1), 83. SCHOWENGERDT, R. A., 1997. Remote Se nsing: Models and Methods for Image Processing. San Diego, CA: Academic Press. SHALABY, A. and TATEISHI, R. 2007. Remote sensing and GIS for mapping and monitoring land cover and land use changes in the Northwestern coastal zone of Egypt. Applied geography ., 27(1), 28 41. SHEN, W. W., J., GRIMM, N. and HOPE, D. 2008. Effects of Urbanization Induced Environmental Changes on Ecosystem Functioning in the Phoenix Metropolitan Region, USA. Ecosystems, 11(1), 138 155. SHIMODA, Y. 2003. Adaptation measures for cli mate change and the urban heat island in Japan's built environment. Building Research & Information 3 4. SHOEMAKER, J., TANKARD, W. and LASORSA, L., 2004. How to build social science theories. Thousand Oaks, CA: Sage. SILVERTON, J., HOLTIER, S., JOHNSTON, J. and DALE, P. 1992. Cellular automaton models of interspecific competition for space the effect of pattern on process. Journal of Ecology, 80, 527 534. SINGH, A. 1989. Digital Change Detection Techniques Using Remotely Sensed Data. International Journal of Remote Sensing, 10(6), 989 1003.
130 SMITH, K. and RAYER, S., 2008. Population projections by age, sex, race, and Hispanic origin for Florida and its counties, 2007 2030. Gainesville, Fla.: Bureau of Economic and Business Research, Warrington College of Business Administration, University of Florida. SOUTHWORTH, J. 2004. An assessment of Landsat TM band 6 thermal data for analysing land cover in tropical dry forest regions. International Journal of Remote Sensing, 25(4), 689 706. SOUTHWORTH, J. and GIBBES C. 2010. Digital Remote Sensing within the Field of Land Change Science: Past, Present and Future Directions. Geogr. Compass Geography Compass, 4(12), 1695 1712. SOUTHWORTH, J., NAGENDRA, H. and TUCKER, C. 2002. Fragmentation of a Landscape: incorporatin g landscape metrics into satellite analyses of land cover change. Landscape Research, 27, 253 270. STABLER, B., MARTIN, A. and STUTZ, C. 2001. Effect of Urban Expansion on Arbuscular Mycorrhizal Fungal Mediation of Landscape Tree Growth. Journal of arboric ulture., 27, 193 202. SUN, H., FORSYTHE, W. and WATERS, N. 2007. Modeling Urban Land Use Change and Urban Sprawl: Calgary, Alberta, Canada. Networks and Spatial Economics, 7(4), 353 376. TOROK, S., MORRIS, C., SKINNER, C. and PLUMMER, N., 2001. Urban heat island features of southeast Australian towns. Australian Meteorological Magazine. 14. TURNER, B. and MEYER, W., 1994. Land Use and Land Cover Change: An Overview. In: MEYER, W. A. T., B ed. Changes in Land Use and Land Cover: A Global Perspective. Cambrid ge, New York, Melbourne: the Press Syndicate of the University of Cambridge, 537. UNITED NATION POPULATION DIVISION 2001. World urbanization perspects 1950 2050 (the 2000 revision). New York: Global Environmental Outlook. UNITED NATION POPULATION DIVISION 2011. Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat, World Population Prospects: The 2007 Revision [online]. United Nations Population Division. Available from: http://esa.un.org/unup/p2k0data.asp [Accessed May 10th 2011]. UNITED NATIONS POPULATION DIVISION, 2003. Demographic trends. [online]. Available from: http://earthtrends.wri.org/pdf_library/data_tables/pop2_2003.pdf U S DEPARTMENT OF STATE, 2010. Kuwait [online]. Available from: http://www. state.gov/p/nea/ci/ku/
131 VELAZQUEZ LOZADA, A., GONZALEZ, J. and WINTER, A. 2006. Urban heat island effect analysis for San Juan, Puerto Rico. Atmospheric environment., 40(9), 1731. VERBURG, P., SCHOT, P., DIJIST, M. and VELDKAMP, A. 2004. Land use change mo deling: current practice and research priorities. GeoJournal, (61), 309 324. VLAHOV, D. and GALEA, S. 2002. Urbanization, urbanicity, and health. Journal of urban health : bulletin of the New York Academy of Medicine, 79(4), 1. WELCH, R. 1982. Spatial reso lution requirments for urban studies. International Journal of Remote Sensing, 3, 139 146. WENG, Q. 2001. A remote sensing?GIS evaluation of urban expansion and its impact on surface temperature in the Zhujiang Delta, China. International Journal of Remote Sensing, 22(10), 1999 2014. WENG, Q. 2001. Modeling Urban Growth Effects on Surface Runoff with the Integration of Remote Sensing and GIS. Environmental Management, 28(6), 737 748. WENG, Q. 2002. Land use change analysis in the Zhujiang Delta of China u sing satellite remote sensing, GIS and stochastic modelling. Journal of Environmental Management, 64(3), 273 284. THE WORLD FACTBOOK, 2002. http://www.wri.org/wri/wr 98 99/pdf/wr98_ud1.pdf. THE WORLD FACTBOOK, 2010. The World Factbook [online]. Central Int elligence Agency. Available from: https://www.cia.gov/library/publications/the world factbook/geos/ku.html 2010]. YAGOUB, M. M. 2004. Monitoring of urban growth of a desert city through remote sensing: Al Ain, UAE, between 1976 and 2000. International Jour nal of Remote Sensing, 25(6), 1063 1076. YANG, X. and LO, C. 2002. Using a time series of satellite imagery to detect land use and land cover changes in the Atlanta, Georgia metropolitan area. International Journal of Remote Sensing, 23(9), 1775 1798. YUAN F. 2010. Urban growth monitoring and projection using remote sensing and geographic information systems: A case study in the Twin Cities Metropolitan Area, Minnesota. Geocarto Int. Geocarto International, 25(3), 213 230. ZHANG, Q., BAN, Y., LIU, J. and HU, Y. 2011. Simulation and analysis of urban growth scenarios for the Greater Shanghai Area, China. Computers, Environment and Urban Systems, 35(2), 126 139. ZHANG, Q., WANG,J., PENG,X., GONG,P. and SHI, P. 2002. Urban buil t up land change detection with road density and spectral information from multi temporal Landsat TM data. International Journal of Remote Sensing, 23(15), 3057 3078.
132 ZHOU, B. and KOCKELMAN, K. 2008. Neighborhood impacts on land use change: a multinomial l ogit model of spatial relationships. The Annals of Regional Science, 42(2), 321 340. ZHU, Y. 2000. In Situ Urbanization in Rural China: Case Studies from Fujian Province. Development and Change, 31(2).
133 BIOGRAPHICAL SKETCH Muhammad Almatar was born in 1979 in Kuwait. In 1997, h e graduated from Salah Shiehab High School in his home country, majoring in science. He received his undergraduate degree as an honor student in G eography with a minor Political S cience from Kuwait University in 2003. He worked as a high school teacher in Salah Shiehab High School. In 2004, Kuwait University offered h im a scholarship to obtain his m D degrees in Geographic Information S ystems. He earned his Master of Science degree fro m the Geography Department at t he University of Florid a, with a specialization in GIS and remote sensing in the summer of 2008 His m aster hesis investigate d suburban areas i n Alachua County, FL utilizing Knowledge Based C lassification (Advanced Classification Method) to differentiate suburban dwellings from vegetation. Suburban residential areas are usually covered with a lot of trees causi ng classification problem. His m aster hesis concluded stating Knowledge Based Classif ication method is no t effective in an area like Alachua County. In Fall 2008, he started his Ph.D. studies at the University of Florida in G eography with a concentration in L and Change S cience and a multi disciplinary minor. He received his Ph.D. from The University of Florid a in fall of 2011.