SAAD ALGHARIB, M.A., August 2008 GEOGRAPHY

SPATIAL PATTERNS OF URBAN EXPANSION IN KUWAIT CITY BETWEEN 1989 AND 2001 (87 PP .)

Director of Thesis: Dr. Jay Lee

Urbanization is a complex phenomenon that occurs during the city’s development from one form to another. In other words, it is the process when the activities in the land use/land cover change from rural into urban. Since the oil exploration, Kuwait City has been growing rapidly due to its urbanization and population growth by both natural growth and inward immigration. Recently, the use of Remote Sensing and Geographic

Information Systems became very useful and important tools in urban studies because of the integration of them can allow and provide the analysts and planners to detect, monitor and analyze the urban growth in a region effectively. Moreover, both planners and users can predict the trends of the growth in urban areas in the future with remotely sensed and

GIS data because they can be effectively updated with required precision levels.

The main purpose of this study is to detect and evaluate the changes in the land use/land cover of Kuwait City area due to the urbanization process between 1989 and

2001. Furthermore, this study assesses if the spatial patterns and process of these changes take place in a random fashion or with certain identifiable trends.

During the study period, the result of this study indicates that the urban growth has occurred and expanded 10% from 32.4% in 1989 to42.4% in 2001. Also, the results revealed that the largest increased of the urban area was occurred between the major

highways after the forth ring road from the center of Kuwait City. Moreover, the spatial distribution of urban growth occurred in cluster manners. SPATIAL PATTERNS OF URBAN EXPANSION IN KUWAIT CITY BETWEEN 1989 AND 2001

A thesis submitted to Kent State University in partial fulfillment of the requirements for the degree of Master of Arts

by

Saad M Algharib

August, 2008

Thesis written by

Saad M Algharib

B.A., Kuwait University at Kuwait, 1997

M.A., Kent State University, 2008

Approved by

______, Advisor

Jay Lee

______, Chair, Department of Geography

Jay Lee

______, Dean, College of Arts and Sciences

John R.D. Stalvey

ii

TABLE OF CONTENTS

Page

LIST OF TABLES ……………………………………………………………………... v

LIST OF FIGURES …………………………………………………………………....vi

ACKNOWLEDGEMENTS …………………………………………………………... viii

CHAPTER

1 INTRODUCTION ...... 1

2 RESEARCH PROBLEMS ...... 5

3 LITERATURE REVIEW ...... 7

3.1 URBANIZATION AND LAND COVER CHANGES...... 9

3.2 URBANIZATION IN ARABIC GULF...... 12

3.3 GIS AND REMOTE SENSING TECHNOLOGY DETECTION OF

CHANGES IN LAND USE AND LAND COVER...... 17

4 STUDY AREA AND HISTORICAL DEVELOPMENT ...... 23

4.1 STUDY AREA...... 23

4.2 HISTORICAL DEVELOPMENT...... 24

5 RESEARCH METHODOLOGY...... 29

5.1 GEOGRAPHIC INFORMATION SYSTEMS (GIS) ...... 29

5.2 REMOTE SENSING...... 30

5.3 RESEARCH DATA...... 31

5.4 METHODOLOGY...... 31

iii

6 DATA PREPARATION, ANALYSIS AND RESULTS ...... 39

6.1 DATA PREPARATION AND REMOTE SENSING ANALYSIS...... 39

6.2 NEIGHBORHOOD OPERATIONS...... 51

6.3 APPLYING NEIGHBORHOOD OPERATION (FILTERING)...... 58

6.4 GIS ANALYSIS...... 62

6.5 THE RESULTS OF URBAN GROWTH IN KUWAIT CITY ...... 64

6.6 EVALUATING THE DISTRIBUTION OF SPATIAL PATTERNS

CHANGING...... 69

6.7 FIRST MORAN’S INDEX (GRID)...... 69

6.8 SECOND MORAN’S INDEX FOR POLYGONS...... 72

7 CONCLUSION...... 78

REFERENCES ………………………………………………………………………… 83

iv

LIST OF TABLES

Page

Table 1: ERROR MATRIX FOR ALL CLASSIFICATION CLASSES 1989...... 49

Table 2: ERROR MATRIX FOR ALL CLASSIFICATION CLASSES 1996...... 49

Table 3: ERROR MATRIX FOR ALL CLASSIFICATION CLASSES 2001...... 49

Table 4: ERROR MATRIX FOR ALL URBAN CLASS 1989 ...... 50

Table 5: ERROR MATRIX FOR ALL URBAN CLASS 1996 ...... 50

Table 6: ERROR MATRIX FOR ALL URBAN CLASS 2001 ...... 50

Table 7: NUMBER OF POLYGONS AFTER TRANSFORMING RASTER TO

VECTOR FORMAT ...... 59

Table 8: SHOWS THE AREAS OF EACH CLASS IN KM² BETWEEN 1989 AND

2001...... 65

Table 9: SHOWS THE VALUES OF MORAN’S INDEX FOR THE URBAN GROWTH

IN KUWAIT CITY ...... 70

Table 10: SHOWS MORAN’S INDEX VALUES FOR ALL THE STUDY AREA

UNITS ...... 75

Table 11: DISPLAYS THE Z SCORE VALUES EQUAL STANDARD

DEVIATIONS ...... 76

v

LIST OF FIGURES

Page

Figure 1: KUWAIT URBAN AREA...... 23

Figure 2: SHOWS LANDSAT IMAGE 1989 OF KUWAIT CITY AND THE

SURROUNDING AREA...... 40

Figure 3: SHOWS LANDSAT IMAGE 1996 OF KUWAIT CITY AND

SURROUNDING AREA. THE CIRCLE SHOWS THAT URBAN AREA IS

FROM LANDSAT IMAGE 1998. THIS IMAGE WAS GENERATED BY

APPLYING MOSAIC...... 41

Figure 4: SHOWS LANDSAT IMAGE 2001 OF KUWAIT CITY SURROUNDING

AREA ...... 42

Figure 5: SHOWS LAND USE AND LAND COVER MAP OF KUWAIT CITY IN

1989 OBTAINED BY UNSUPERVISED CLASSIFICATION METHOD ....45

Figure 6: SHOWS UNSUPERVISED CLASSIFICATION OUTPUT OF 1996 IMAGES

AND THE LAND USE LAND COVER OF KUWAIT CITY...... 46

Figure 7: SHOWS UNSUPERVISED CLASSIFICATION OUTPUT OF

2001 IMAGE ...... 47

Figure 8: ORDER OF MOVEMENT OF A 3-BY-3 WINDOW FOR NEIGHBORHOOD

OPERATIONS. (SOURCE. BOLSTAD. 2005 GIS FUNDAMENTALS: A

FIRST TEXT ON GEOGRAPHIC INFORMATION SYSTEMS, P 363)...... 52

vi

Figure 9: VARY SHAPES OF FILTERS (SOURCE.CHANG. INTRODUCTION TO

GEOGRAPHIC INFORMATION SYSTEMS. P. 264)...... 54

Figure 10: (SOURCE. BOLSTAD. 2005 GIS FUNDAMENTALS: A FIRST TEXT ON

GEOGRAPHIC INFORMATION SYSTEMS, P 364)...... 56

Figure 11: SHOWS AND COMPARES THE DIFFERENCE BETWEEN FILTERING

OPERATION ...... 61

Figure 12: ILLUSTRATES THE URBAN GROWTH OF KUWAIT CITY DURING

THE STUDIES PERIOD ...... 65

Figure 13: SHOWS THE NEW DEVELOPED AREAS BETWEEN 1989 AND 2001 IN

AND AROUND KUWAIT CITY...... 68

Figure 14: SHOWS THE DIFFERENCES OF MORAN’S VALUES BETWEEN EACH

PEIROD...... 70

Figure 15: ILLUSTRATES MORAN’S INDEX VALUES DURING THE STUDIES

PERIOD...... 75

Figure 16: DISPLAYS Z-SCORE OF MORAN’S I FOR THE STUDY AREAS ...... 76

Figure 17: SHOWS THE RESULTS OF THE FIRST PERIOD 1989-1996 BY

APPLYING MORAN’S INDEX POLYGONS FOR THE ENTIRE KUWAIT

CITY...... 77

Figure 18: SHOWS THE RESULTS OF THE SECOND PERIOD 1996-2001

OBTAINED FROM MORAN INDEX POLYGONS TO ENTIRE KUWAIT

CITY...... 77

vii

ACKNOWLEDGEMENTS

I would like to express my great gratitude and appreciation to my committee members, Dr. Jay Lee, Dr. Mandy Munro-Stasiuk and Dr. Chuanrong Zhang of the

Department of Geography at Kent State University. I would like to thank them for reviewing and providing me thoughtful critiques to finish my thesis. Also, their comments have been very helpful for improving the quality of this work. I especially want to thank my thesis advisor, Dr. Lee for his ideas, comments and suggestions which allowed me to complete this work successfully.

I owe a great deal of gratitude to the Department of Geography at Kuwait

University for their financial support which gave me the opportunity to pursue my education in the United States. Also, I extend my appreciation to many people who have helped and encouraged me to further my studies. I especially would like to thank my friend Yukihiro Suzuoki for his assistance.

Finally, I would like to thank all of my family for their support. I especially want to thank my wife for her support, interest and encouragement all the time. I would like to dedicate this thesis to my wife and daughters (Anwar, Aseel and Fajer). Without their love and sacrifice, I would not be able to successfully complete this project.

viii

CHAPTER 1

INTRODUCTION

The industrial revolution had a huge impact on the feasible sizes of cities. Many cities around the world grew rapidly in size after the industrial revolution, especially those in

Western Europe and North America. This is due to continuing rural-to-urban migration in seeking of better socio-economic opportunities. However, the industrial revolution has not had much impact on the size of Kuwait City until very recently because the effect of industrial revolution reached the Arabic Gulf region in which Kuwait City is located only recently.

Kuwait City was a small costal city with a big wall that separated it from the desert. People in this city used to live in small mud houses. The big change in the size of

Kuwait City was at onset of oil exploration in around 1952. Since then, the size of

Kuwait City has increased rapidly. The City became one of the fastest growing cities in the world for the socio-economic opportunities it offers to people in surrounding regions.

The population of Kuwait City was around 200,000 in 1957 but increased to 730,000 by

1970. This trend of increasing population has continued as it reached 2,213,403 in 2005.

The new arrivals in Kuwait City include those people from the surrounding desert while others came from foreign countries, such as India, Egypt and Iran. The capacity of the old Kuwait City was too small to accommodate all of the new immigrants. For this reason, the Kuwaiti government brought in British experts to rebuild the old city and to

1 2

expand the city with new suburbs to cope with the growth. Moreover, the government

devised and adopted a series of master plans that shaped and formed the development of

Kuwait City since the first master plan in 1952.

The process of expanding urbanization areas in Kuwait City is reflected by how

the lands are used in and around the urbanized areas. There are many ways to detect and

analyze the urban growth, such as using geographic information systems (GIS) and

classified remotely sensed images. The use of the GIS and remote sensing technology has

gained much popularity in recent decades in many fields, including urban geography. To

study and plan for urban growth, information such as growth rate, growth, pattern and

extent of urban sprawl would be necessary to support any such efforts.

The main objective of this study is to detect changes in urban land use /land cover

and to examine the changing spatial patterns of urban growth in and around Kuwait City

between 1989 and 2001. In addition, this study also evaluates the spatial patterns of the

changes detected and how they can be related to the spatial configuration of the city.

There have been discussions of the spatial structure of developments in urban growth

among urban geographers and by researchers in related fields. For example, it had been

suggested that urban developments in Gulf region may take one or more of the following

forms: linear development, infill development and scattered development as suggested by

Ali and Taher (2004).

Furthermore, suburbanization typically was characterized by going through spill- over and then in-fill stages. This study uses this framework to detect the spatial configuration of urban growth in Kuwait City to see what types of development exist

3

there. By answering these questions, and relating them to the location of the CBD and the existing transportation network in Kuwait City, this study offers an insight to the understanding of how the urban growth has occurred during the study period for the purpose of better planning for future development. Also, the study provides information to support policy makers for more informed decision-making for the next development stage.

In order to identify the new urban areas between 1989 and 2001, the study uses satellite images of the study area and remote sensing technology for classifying these images. Unsupervised classification method was applied to classify images to land use and land cover data layers. After finishing the unsupervised classification method, GIS overlay function was applied to the classified images for detecting the locations and patterns of the new urban areas that developed during the study period. GIS was also utilized to evaluate the distribution of the spatial patterns. For example, Moran’s index was applied for all data inputs to examine the urban growth distribution.

The results indicate that there has been rapid urban growth in the study area. The urban growth has taken place in three different shapes. First, infill development occurred in and near the center of Kuwait City. Second, linear development was found along the coastal line from the center of Kuwait City to the south and west sides of the city. This study also identifies a significant development located between the major highways in and around Kuwait City. This identified development appeared between the forth and six roads in the west and in the southwest side of Kuwait City and along the highways to south of Kuwait City. In addition, using the Moran’s index, it was concluded that the

4

distribution of the spatial patterns for the new development areas tends to show a clustered pattern that new developments often occurred in close proximity with each other.

CHAPTER 2

RESEARCH PROBLEMS

Urbanization is a complex phenomenon that involves many factors. Cities, growing from a small population settlement to a mega city, often went through urbanization, suburbanization, counter-urbanization, and re-urbanization. Different cities may have taken different routes in this process. Not all cities progress in the same speed in urbanization, either. As a result, cities around the world stand at different stages of urbanization.

This study focuses on Kuwait City because of its social, economic and political importance in Kuwait. This study asks if the changes in land use/land cover as detected from using remotely sensed data would display similar spatial structures as those seen in the West. Specifically, the study applied GIS and remote sensing technology to detect changes in land use/land cover and to analyze the detected changes to examine their spatial patterns as related to existing urbanized areas in and around Kuwait City.

It is expected that the spatial pattern of the changes in land use/land cover could be significantly different from those seen in cities in the West because of the physical surroundings and cultural traits that are unique in the Arabia Gulf. If, however, similar trends and spatial patterns of land use/land cover changes are concluded, this study will provide a new documentation of the essence of urbanization process that could be generalized to be applicable across cultural and physical divides.

5 6

Specifically, this study will carry out research

• To detect urban land use/cover changes and to examine the changing spatial

patterns of urban growth in Kuwait City between 1989 and 2001.

• To evaluate and examine the distribution of spatial patterns changing and to see if

the changes of the geographical patterns occurred in clusters, dispersion or

randomly.

By answering these questions, the study examines the unique characteristics of the pattern of urbanization in and around Kuwait City. For example, there are many types of development, such as linear development, infill development and scattered development

(Ali and Taher 2004) that we can relate the findings to. In addition, the study used GIS and remote sensing technology to detect the direction of changing spatial patterns of land use/land cover in and around Kuwait City. Finally, the study attempted to link the spatial patterns of changes in land use/land cover to transportation infrastructure in and around

Kuwait City.

CHAPTER 3

LITERATURE REVIEW

Urbanization is a complex phenomenon that refers to the extent of urban growth and development, and lives in many mega cities around the world are influenced by urban growth. Urbanization is a worldwide phenomenon that takes place all over the world and is among the most important forces that drive the formation and changes in the patterns of land use and land cover in and around urbanized areas. In cities with a radiant pattern of transportation networks, the growth of cities occurs around the cities center in various directions such as radial and linear growth, depending on the road access. Also, the development of cities sometimes occurs along the urban fringes and by the side of highways.

Human behavior has a huge impact on the environment at the local, regional and global levels. To that end, changes in the land use/land cover as caused by urban growth are among the most significant consequences resulted from intensive human activities in the region (Tian 2005). There have been many studies which associate and relate the development of cities to the changes in the spatial patterns of the cities’ economy, population and industrialization processes. These studies investigate the relationship between urban growth and demographic and economic factors to see how these factors influence the urban growth. For example, economic and industrial growth encourages people to immigrate from rural to urban areas seeking better social and economic

7 8

opportunities Because of this, the size of the cities continue to increase and still grows rapidly.

While these factors influencing urbanization may be ubiquitous, there are others that are unique to specific political and cultural structures of different localities. For example, cities developed with endogenous forces tend to develop in different speed and result in different spatial forms from those with exogenous forces. Furthermore, development of cities in which capitalism provides the main stimulus tends to follow different paths from those in cities with centralized planning.

Kuwait’s economy has increasingly relied on the income generated from exporting petroleum resources. This income has been among government’s controlled resources. As such, the policy makers and municipal councils play essential roles in the location of new urban areas. For example, in Kuwait and Dubai, the governments represented by municipal council decide where the new urban areas will be located and what kind of land use will be in each area. By studying the urban development in Kuwait and Dubai, one can see that rapid changes in land use/land cover took place due to urbanization in both cities since the beginning of oil exploration. In addition, the only one who has the authority to plan and construct new settlements is the government because government owns most of the developable lands in the cities.

Each local government brought experts from western countries, such as Britain and Austria, to plan future developments. Those experts utilized and applied some urban models developed in the West, such as Concentric Zone Model, Sector Model and

Multiple Nuclei Model that were used to build and reshape the cities. Regardless of the

9

outcome being desirable or not, the spatial forms of some cities in Arabic Gulf have become very similar to those cities in Europe and North America.

As pointed out earlier, urbanization is a complex phenomenon that includes and involves many factors. Analyzing this process requires the support of a large amount of spatial data. For this reason, many researchers integrate, utilize and apply both

Geographic Information System and Remote Sensing to assist in the detection and analysis of the land use/land cover changes and the changes of their spatial patterns as key indicators of how the process of urbanization has evolved in their study areas. These tools allow users and planners to easily identify, examine, and evaluate the changes in land use/land cover that occur due to the urban growth.

3.1. Urbanization and land cover changes:

Greene and Pick (2006) stated that cities change every year and even every day. Signs of changes are obvious in each city - rerouting of streets, construction of new high-rise buildings, businesses starting, businesses failing, and so on. In their study, they focus on the economic, demographic and social changes as occurred in cities. The first area of changes they focus on is the economic changes which the growth of cities depends on the most. Furthermore, they mention factors of agricultural and light manufacturing production and their spatial arrangements in the cities that typically form the foundations for the cities’ economic development, which in turn change the way urban areas developed.

10

They suggested that economic changes and urban changes are firmly connected in

a two-way feedback relationship. Economic changes can often impact urban planning,

policies, and cause spatial changes. The later often dictates the former with policies and

land use regulations. The second area of changes they focus on is demographic changes

in the cities they examined. This sector is essential for urban development because it

determines the size, structure and distribution of cities’ populations. Also, they studied

some fundamental components of changes that influence the city’s population growth and

changes such as births, deaths, net migration and annexation. The last area of changes

they focus on is social changes. The social changes involve changes in education, social

status, language, race, ethnicity, marriage and so forth. They showed that social changes

sometimes lead to socio-economic polarization, which often result in the separation of a

society into groups whose differences we can recognize clearly.

In Firman’s study (Firman 1999), he looked at the issue of land conversion and

changes in land use during the period of economic explosion (from the 1980 ѕ to the mid-

1990 ѕ) and the time period of economic crisis (from 1997 until the present) in the fringe areas of major cities in Indonesia. He found that the change from agricultural land to urban land is estimated to have exceeded more than 106,000 ha during the studied period.

These include 54.7% of residential areas, 15.5% of industrial lands, 4.9% of office uses and 25.3% of other urban land uses. He proved that the urban economic development has affected the changes of urban land cover/land use. He stated that the process of urbanization is a normal part in urban development but it has mostly been uncontrolled in cities in Indonesia. On the other hand, the economic crisis has a huge impact on land

11

development from which the demand on the residential areas and industrial estates was reduced. The land conversion in this period was slowing down and even stopping in fringe areas of large cities.

Antrop (2004) studied the urbanization in Europe and showed that urbanization is the change of land use/ land cover from rural life into urban ones during a complex process. Also, he stated that this urbanization not only influences the growth of cities and towns but it is also the processes of the development in countryside represented by improving its accessibility to other areas. For example, when the access to and the availability of cheap open space are easy and fast, new small industries, commerce and exhibition halls, hotels and restaurants will be attracted to it. However, they tend to slow down as fewer vacant or developable lands during this urbanization process.

To that end, Antrop discussed several stages in the development (urbanization). In the first stage, called urbanization, people migrated from the fringe and preferred life in the city center. The second stage, called suburbanization, still shows growth in which the city center loses population while the urban fringe zone is growing hastily. In the third stage, called counterubanization or disurbanization, the urban population declines in both center and fringe areas. The fourth stage, which is called reurbanization, shows a recovering of population in the cities - first in the city center, and later in the fringe area.

Moreover, he argued that the change from one urbanization stage to another depends on factors such as land quality, land price, accessibility and safety. Finally, he related the urbanization process to the industrialization and economic growth that was caused by industrial revolution.

12

Esbah (2007) studied land use/ land cover changes due to urbanization in the city of Aydin between 1986 and 2002. He used three indicators to recognize the impact of urbanization on land resources: the new urbanization density, the loss of agricultural lands and the loss of natural lands. He also stated that there are three main factors that play significant roles in the development of Turkish cities: the increase of 1) industrialization, 2) population and 3) a transportation network. Due to these factors, the land use expanded and changed to other uses and the total change was 1,806.02 ha during the study period. He found that land use in 1986 was agricultural and natural lands were converted to major urban, military, industrial and mining in the study area by 2002.

In summary, a city may be at different stage of urbanization as indicated by different ways changes in land use and land cover occur. To study urbanization as a process, detecting and analyzing changes in land use and land cover will be an important first step as they may serve as indicators of the direction and speed of urbanization.

3.2. Urbanization in Arabic Gulf:

Abo aysh (1981) studied Kuwait metropolis and the problems that occur from it.

Uncontrolled urbanization may cause many problems that affect a country’s development. For example, it created congestion, pollution and misdistribution of the facilities for public services. The main objective of his study was to persuade the government to build new cities to reduce the negative impacts of a crowding metropolis and absorb the pressure on the center of Kuwait City.

13

Abo aysh divided his study into three sections. In the first section, he discussed

the factors that created this pressure in Kuwait City, such as oil, immigration, and the

economy. He suggested that oil exploration had a huge impact on the development of

Kuwait City. Many sectors were influenced by oil trades and oil production such as the

government facilities sector, commercial sector, industrial sector, and transportation

sector. This indeed has affected the development of Kuwait City and is likely to continue.

The second section in Abo aysh’s study describes the urban development of

Kuwait City. There were four kinds of development in the urban growth of Kuwait City:

the first development was in the middle of the old City after oil exploration 1946. At this

time the government tried to control random urban growth that accrued before oil

exploration. The second development was in the1950 ѕ around the old city and some

industrial and commercial areas appeared; in other words, this was the beginning of

modern Kuwait City.

The third development was in the1960 ѕ, around the second development. The city had all kinds of city problems, such as traffic congestion just like we can find in any big modern city. The fourth development was the 1970 ѕ in which the Kuwait City expanded

to the south and to the west side and became a metropolitan city. The final section in his

study explains why we need new cities as he suggested. He stated that the new cities will

accommodate many people and reduce pressure on the center of Kuwait City. Also, he

recommended developing some cities in south, north and west of Kuwait City.

Almnyes (1994), studied different models of urban land use patterns, following

the classic Concentric Zone Model, Sector Model and Multi-nuclei Model. He studied the

14

shape of Kuwait City using the Master plan in 1952. Finally, he compared the shape of

Kuwait City with these models. The results that Almnyes found in his studies are that the models play very important roles in urban geography of Kuwait City. Also, after comparing the shape of Kuwait City with these models, he stated that the shape of

Kuwait City looked like a concentric zone model after the Master plan in 1952. In this period, the shape of the city looked like half circle surrounding the old city and the old city became the central business district. When Kuwait City expanded to the south and west sides during the1960 ѕ and 1970 ѕ, the shape of the city had been changed and looked like sector models. Furthermore, he concluded that the shape of Kuwait City has evolved through more than one model.

In another article, Almnyes (1996), attempted to prove that Kuwait City is a big city by comparing it with other big cities around the world like New York and London.

For example, he used population, crime, and transportation to examine as the factors in his comparison. By studying the social and economic indicators of Kuwait City, he showed that Kuwait City has the characteristics similar to those of any metropolitan cities in the world.

Alkhuzamy (2001) studied the urban development of Kuwait City since 1765 by using cartographic information modeling. In his study, he utilized technical integration between digital aerial photographs, satellite images, and digitized maps along with old historical maps to evaluate urban development in Kuwait City. One of his objectives was to detect the trend of the urban growth in Kuwait City. He stated that urban development grew rapidly in Kuwait City since oil foundation in ways similar to those cities in the

15

Gulf Cooperative Council (GCC). However, he also showed that the historical indication has specific characteristics that play an important role in the identity and the trend of the urban development of the Kuwait City after oil exploration. The trend of the urban development in Kuwait City after oil exploration had been influenced by several elements such as the locations of the oil fields and the locations of the splendid areas for the international oil companies. Furthermore, he declared that the urban growth of Kuwait

City was developing on its own course before the oil exploration whereas afterwards the government directed the urban growth of the city by adopting some plans for modernization.

In yet another study, Almnyes (1998) identified the characteristics and the role of the Central Business District (CBD) with respect to Kuwait City. He did so by using some indicators such as the role of the CBD, the role of structural plans, and the land price away and near CBD. He showed that the Central Business District extended from the walled old city that was established before oil exploration. After the oil exploration, the structural plans played important roles in the development of Kuwait City. For example, the plans designed to expand the city outside the wall and to convert the old city to be the Central Business District proceeded by planned development orchestrated by

Kuwait governments.

Kwarteng and Chavez (1998) detected and mapped changes in Kuwait City and its environments between 1986 and 1993 by using Landsat Thematic Mapper (TM) images. Also, they aimed at proving the effectiveness of Landsat Thematic Mapper images for both surface/spectral mapping and temporal change detection. They stated that

16

Kuwait City expanded rapidly during the study period. Actually, ten new neighborhoods were developed during the time period. Finally, they demonstrated and showed that we can deal with urban development and detect/map temporal changes by using Remote

Sensing data and technology.

Gabriel (1987) studied the urban growth in Dubai and stated the city grew rapidly since the oil exporting of 1969. Before this time, people in the city used to live in houses that were built from coral and clay and the access roads to those houses were narrow and winding. Also, those houses were built randomly, by convenience instead of by plans.

After the oil exploration, the government had the ability to invest much more money to lead the urban development in the city in a more structural manner than before.

Moreover, the entire cost of the planning, construction, and allocation of new settlement areas were covered by the government. The government adopted modernization programs such as preparing and following master plans in1960 and 1971. As a result, these plans converted the city and its traditional lifestyle from a small Islamic city to a modern city similar to those in western countries. Furthermore, these plans include future development and we can notice that the modern buildings and urban development also followed the guidelines.

In short, the development of Kuwait City started out as a small Islamic city growing on its own course until the oil exploration that enabled the government to invest and plan for more structural development. Along with the rural-to-urban migration in seeking of better economic opportunities by Kuwaiti population, Kuwait City has grown to be a big city as it is now. This suggests that, when studying the urbanization of Kuwait

17

City, it is important to keep in mind the role that the old center plays as well as the roles of the newly developed south and west sides.

3.3. GIS and Remote Sensing technology detection of changes in land use and land cover:

Esbah (2007) utilized Geographic Information Systems (GIS) and Remote Sensing (RS) to detect changes in land use/ land cover in western Turkish cities due to the urban growth between 1986 and 2002. The author showed that the GIS and remote sensing technologies are very effective tools for studying urban growth. First, he used satellite images to detect land use changes and then to identify all land use types: agricultural, industrial, urban, and so on. Supervised and unsupervised classifications were applied in classifying the satellite images to land cover maps. Second, he utilized GIS to delineate and analyze the changes in all of the land use types.

Similar to Esbah, Weng (2001) examined the spatial patterns of urban growth and studied the urban land cover changes by applying Remote Sensing and Geographic

Information Systems to the Zhujiang delta in southern China. He also stated that the combination of Remote Sensing (RS) and Geographic Information Systems (GIS) demonstrates and proves to be a useful tool for detecting and analyzing urban development. By using remote sensing, we can collect timely data and turn it into information from which we can understand and monitor the processes of urban land development. Also, he utilized Landsat Thematic Mapper (TM) images from remote sensing to build urban land cover data sets. Then he utilized GIS to enter, examine,

18

display, and analyze the changing of spatial patterns of urban growth. He found that urban, or built-up, lands in his studied area have expanded by 47.68% from 1989 to 1997 while croplands and horticulture farms were the most affected.

Qiong Wu and others (2006) utilized the integration of the Remote Sensing (RS) and Geographic Information Systems (GIS) technologies to detect and analyze land use/land cover changes in Beijing municipal areas over three time periods: 1986-1991,

1991-1996 and 1996-2001. They stated that the major force that impacts on land use changes is urbanization, or the phenomenon of urban development. They used remote sensing satellite images to obtain multi-spectral and multi-temporal data that can be utilized to detect and quantify the land use changes. In addition, GIS was utilized to display, store and analyze these images because GIS can offer a more flexible environment for processing these functions.

For the study period, they found that cropland declined by 44,970.2 ha and that most of these changes were by the lands that were converted to urban land use. Also, they revealed that most of the urban development and loss of agricultural land took place in and around the city.

Alphan (2003) detected and analyzed the urban development and land use/land cover changes in Adana city, Turkey, between 1984 and 2000. In his study, he utilized

Remote Sensing to determine land use/land cover changes by using two different satellite images acquired in 1984 and 2000. He applied supervised classification and divided the land use/land cover into five different classes: (1) semi natural, (2) vegetation cover, (3) agricultural (4) urban and (5) water. He found that the urban area grew rapidly from

19

4,872 ha to 9,400 ha during the study period. In other words, urbanized areas were doubled in size while agricultural and semi-natural areas declined by 23% and 34% respectively. In terms of the urbanization process, he related the urban growth in the city of Adana to socio-economic factors and the outcome of local policies. Furthermore, he mentioned that there are some factors which play significant roles in the urban growth and residential development such as the migration of rural people.

Ali and Taher (2004) studied the urban growth of Riyadh in Saudi Arabia between

1987 and 2001 by using Geographic Information Systems and Remote Sensing.

Supervised classification was applied to the study and the accuracy of the assessment test in remote sensing was assessed to be 81% for this method. After supervised classification was done, the authors of the study used these results as input data in raster format in their subsequent analysis with GIS. They used spatial analysis, buffer analysis, and spatial proximity to measure the urban growth. The authors found that the city grew rapidly during this period. The study also shows that the size of the Riyadh increased 91% between 1987 and 2001. Furthermore, the development of urban areas took some different shapes from the intuitive compact growth. They detected that Riyadh’s growth were multiple facets: linear development, infill development and scattered or leap-frog development. Finally, they confirm that the main roads played important roles in the city’s development.

Xiao and others (2004) also used Remote Sensing and Geographic Information

Systems to detect urban growth. The main concepts of this article are the exploration of the temporal and spatial characteristics of urban expansion from 1934 to 2001 and the

20

land use/land cover change from 1987 to 2001 for Shijiazhuang City in China.

Additionally, the authors proved that the urban expansion goes through fast- and slow-

growth stages, with the high-speed growth districts shifting to the east and west sides of

the city (Xiao 2004). Moreover, they categorized the spatial patterns of urban expansion

into three typical types: (1) special objective oriented type in the war age (1934 - 1947),

in which the military influenced the urban expansion; (2) socio-political intervention

type, in this period the urban expansion was influenced by the political situations; and (3)

normal growth type when the economic development and population growth had a huge

impact in the urban growth over the past two decades.

Sudhira and others (2003) use GIS and remote sensing data with collateral data to

identify the pattern of sprawl and subsequently predict the nature of future sprawl. Also,

the study attempts to describe some of the landscape metrics required for quantifying

sprawl in Mangalore, Udupi region in Karnataka State in India. With statistical analytical

techniques, such as logistic regression, they found the percentage of changes in built-up

areas over the period of nearly thirty years was 145.68% and by 2050 the built up area in

the region will rise to 127.7 km².

Tian and others (2005) use both Geographic Information Systems (GIS) and Remote

Sensing (RS) to analyze spatio-temproal characteristics of urban expansion in China during three periods:1990/1991, 1995/1996 and 1999/2000. The authors of the study calculate the urban land percentage and urban land expansion index for every 1 km².

Furthermore, they use Geographic Information Systems and Remote Sensing to analyze urban growth expansion. They conclude that economic growth environments and macro

21

urban development policies are the most important factors that influence land use

expansion.

Abbaspour and Gharagozlou (2005) discuss how an urban development model can be

created and used to apply environmental information in a wider context, with the help of

geographic information system and remote sensing. “In planning urban development, it is crucial to use proper models for determining the environmental characteristics of the region. Planners should utilize methods that can help determine ecological and social capacities, and outline the existing environmental conditions in the area .” (Abbaspour and Gharagozlou, 2005).

Mundia and Aniya (2005) mention that there are planning problems in Nairobi and misconceptions exist for how urban planning should work there. The main purpose of their study was to examine land use/land cover changes and the dynamics of urban growth in Nairobi. The authors use two basic approaches such as post classification comparisons and simultaneous analysis of multi-temporal data. Their study shows that

Nairobi expanded rapidly between 1976 and 2000, which created some problems. In addition, there are many factors that influence the expansion, such as rapid economic development, urban population growth, traffic infrastructure, and the physical setting.

The study collected its data from satellite images. Also, digitization was done for sets of layers for which data were not available. Major roads, contours, and administrative boundaries were digitized into layers from the published topographical maps. There are advantages and disadvantages of the two basic approaches and these

22

approaches are post-classification comparisons and simultaneous analysis of multi- temporal data.

This study adopts a modified version of the Anderson classification system. The study shows that the urban areas increased rapidly from 15 km² in 1976 to 62 km² in

2000 while forests have declined rapidly from 100 km² to 23 km² over the same period.

Furthermore, the agricultural fields increased from 49 km² to 88 km² during the 24 years of the study period. “Like many African cities, Nairobi’s rate of changes has been very fast mainly because of the big number of immigrants from rural areas” (Mundia and

Aniya 2005).

In conclusion, it appears that the integration of remote sensing and GIS provides a feasible approach to studying urbanization in terms of how urban lands expand. With remotely sensed data from different time periods classified into land cover data layers, it is possible to use GIS to detect and analyze locations of where land cover changes and how these changes distribute spatially.

CHAPTER 4

STUDY AREA AND HISTORICAL DEVELOPMENT

4.1. Study Area

Kuwait is located in the northwestern part of the Arabian Gulf between

Saudi Arabia to the south and Iraq to the north. Kuwait has a total area of about 17,820 square kilometers

(6,880 square miles). The earth’s surface slopes down smoothly from west to east in this area. In general, the eastern side of the region is flat with the exception of few rocky hills.

The distance between the most northern and southern point of the Figure (1) From Kuwait Municipality state’s boundaries is about 200 kilometers (121 miles); and from east to west is about 170 kilometers (105 miles)

(Ministry of Planning 2005).

23 24

’Kuwait lies between latitudes 28 ,ْ 30’ and 30 ,ْ 06’ north of Equator, and longitudes 46 ,ْ 30

and 49 ,ْ 00’ east of Greenwich. Kuwait’s weather is typical of a desert geographical

region. The population was 2,213,403 million in 2005, including about 1. 3 million non

Kuwait citizens with an annual growth rate of 4.8% and the population density is 124.2.

Figure (1) shows the Kuwait urban areas.

4.2. Historical Development:

According to Kuwait Municipality (1980), the development of Kuwait City has gone

through two main periods since its foundation up to the present time: first early growth

before 1952 and second modern urbanization after 1952.

The first period is early growth; in which the growth of the city took an oval

shape. The city was situated on the peninsula at Ras Ajoze and the city was divided into

three sectors: costal, the commercial, and a residential sector. The prime building

materials that people used to build their houses were sea rocks, wood and mud bricks.

The characteristic of these buildings was that they are often very close to each other and

they were attached to each other with high walls. Also, the streets were narrow. That led

the urban development of the city to occur sporadically. During that time period, there

were no technical employees to study, plan and implement any planned effort to guide

urban growth in the city. The urban development of the city was based on mostly the

recommendations from members of the municipal council, the administrations or other

important people in the country.

25

The second growth period started after oil exploration in the early fifties. In other words, with the commencement of oil exploration, a new period of the country’s development began. Oil played a significant role in the economic development and led to the change from early growth to modern urbanization.

Modern urbanization has passed through seven essential master plans that formed and designed the development of Kuwait City starting in 1952 up to the present time.

During in this period, many projects were applied in both the private and public sectors.

The first master plan in 1952 was planned by Monoprio, Spencely and Macfarlane in which Kuwait City for the first time utilized and applied modern planning standards.

The plan’s aims and objectives covered the city from inside the wall and to the outside.

Furthermore, the plan adopted the concept of a three rings and radial road system around the center of the old city. The plan also paid a great deal of attention to land use and land cover patterns because it identified and allocated all of these areas: commercial, residential, industrial and government offices. The urban development started taking place shortly after applying the plan. This plan established the foundation for urban planning in Kuwait.

The second plan was the municipality development plan in 1967. It was an assembly of different planning studies for different areas. The government adopted the plan to cope with the country’s speedy development in that time. The reason for the increased development was due to population growth by both the natural growth and immigration. The population increased from 206,473 in 1957 to 467,339 in 1965, more than doubling its size in a mere 8-year period. There were some significant achievements

26

of the plan which allocated 40 residential areas and also commercial and industrial areas.

Moreover, the plan added the fifth and sixth ring roads and extended the major radial roads around the Kuwait City center.

In 1970, the second master plan was prepared by Colin Buchanan and Partners.

The plan was prepared to meet developmental requirements for the future increase in population. The main aims and objectives of the plan were to set the long term strategy, the national physical plan, a short term master plan for urban areas and the plan for

Kuwait City. The plan also examined and identified the best location for the land use that was needed for urban growth in the future and divided it into three stages. In this plan, the first stage aimed at accommodating 525,000 people along the coastal regions between

Salmiya and Ahmadi; and at Sulaibikhat and Jahra.

The second stage in this plan was to build a new town to the south and to the west of Kuwait City that could provide new accommodations when the population reached two million people. The last stage studied the areas that could accommodate another million.

The plan also paid significant attentions to transportation, the road traffic and all the facilities that can be offered by the government.

During the period between 1970 and 1977, the change of the increased national income caused unanticipated changes in social life and in population numbers. The circumstance encouraged the government to adopt the first review for the second master plan in 1977 by Shankland Cox and Partnership. The main objectives of that plan were to plan for the metropolitan structure, city center structure, and national physical

27

infrastructure. In the beginning, the plan recommended to build new cities in the south and to the north of Kuwait City to accommodate one million people each by 2000.

Other master plans, second review for the second master plan in 1983, the third master plan in 1997, and the first review for the third master plan in 2005, maintained the principles of the urban structure that was determined in the first master plan and its first review. They also completed the urban development along the coastal line to the south and west of Kuwait City between Mishrif and Shuaiba and at Jahra. These plans also intended to stimulate the urban growth in the lands north of the sixth ring road. For example, these master plans recommended and added six new industrial sites in the metropolitan area. In addition, they planned to establish a new city out of the metropolitan area in the south, in the west, and in the area north of Kuwait City. For example, the first review for the third master plan suggested building five new cities that can accommodate 400 thousand people. The planers also aimed for each city to be independent.

In summary, oil represents the most important natural recourse of the country and the main source of its national income. Due to the situation of oil exploration, the rapid growth of Kuwaiti income has led to rapid urban development. The development of

Kuwait City can be divided into two different periods: the first period in which the houses and buildings were built randomly by convenience instead of by plans whereas the second period the government invests money to lead and direct the urban development of the city by preparing seven master plans. Each one of the master plans

28

contributed to the development of Kuwait City and changed the city from a small Islamic city to a modern city.

CHAPTER 5

RESEARCH METHODOLOGY

To study the patterns of changes in land use/land cover in Kuwait City and how they changed over time, I decided to use remotely sensed satellite images as the source of data from which information of land cover can be extracted. Subsequently, I used geographic information system to detect and analyze any changes of land cover over time and to analyze spatial patterns of those changes. The study integrates remote sensing technology with GIS technology for the purpose of studying land use/land cover data layers.

Geographic Information Systems and Remote Sensing are very important tools in many fields because these tools can provide useful data for users to analyze them easily.

In other words, the integration of the geographic information systems and remote sensing can provide a useful approach to analyzing data from which the users can detect urban growth.

5.1 Geographic Information Systems (GIS):

GIS has been widely used in urban analysis because it offers many functions, such as integrating, storing, analyzing and managing geographic data that can help planners and users manipulate a large amount of data (Juliana, 2006). GIS is well known for having buffering and overlaying functions that users can apply to multiple data layers to detect changes or to formulate spatial patterns of the observed phenomena. GIS has many

29 30

functions that can assist researchers and policy makers who deal with the challenge of analyzing urban environments. GIS has become a very important tool for many who deal with urban planning, transportation, community development, environmental hazards, health care, or a host of other crucial concerns in a contemporary city.

Okunuki (2001) shows that urban analysis can not advance without GIS because many researchers need to manipulate large amounts of spatial data about urban areas to find the distributions (patterns) and processes of spatial phenomena that they investigate.

Furthermore, GIS users can combine data from different sources and connect them in maps of specific areas and this will help users to search, analyze and manipulate them quickly.

5.2 Remote Sensing:

Remote sensing can provide much data that cover large areas and may contain both high level of spatial details and high temporal frequency. Many studies also utilize remote sensing images to detect and monitor land cover changes at various scales (Xiao, 2004).

Recently, the use of Geographic Information System and Remote Sensing became very popular because of the ever improving precision and accuracy. They are important tools in urban studies because this integration of these data sources can provide data that are accurate and precise enough to allow users and planners to detect, monitor and analyze urban growth effectively. Planners can predict the trends of urban growth in the future with remotely sensed and GIS data because they can be effectively updated.

31

This thesis study applied both of these tools for Kuwait City to examine the changing spatial patterns of urban growth in the city and to evaluate the spatial patterns of the changing land use/land cover in and around Kuwait City. Findings from the study provide much needed information for policy makers so that they would know which parts of the city outgrew or fell behind others. Similarly, the resulting information will form the basis for future planning or further reviews of master plans. In general, the study contributes to the development of Kuwait City and the wellbeing of its residents. Finally, due to these reasons, I believe that many cities around the world can use and apply the integration of Geographic Information Systems and Remote Sensing to analyze urban growth.

5.3 Research Data:

Three types of data were used to carry out the study. First, satellite images of the studied areas were needed to be acquired for the time periods being studied. Second, additional

GIS data layers were needed to facilitate the analysis of detected changes in land use/ land cover. Third; hardcopy land use maps of Kuwait City were also needed as a reference and to compare all the results.

5.4 Methodology:

The study first examined and quantified any changes in land cover and land use of

Kuwait City over the study period by using satellite images of the study area and remote sensing technology for the classification of the images into a data layer of land cover.

32

Classification is the method of assigning and deriving information classes from pixels of the remotely sensed images by examining the attribute values of the pixels from multiple bands. For example, the easiest form of classification method is to consider each pixel alone and to assign it to a class based upon its several values calculated in separate spectral bands (Campbell 2007). Moreover, the final results of the analysis that we can acquire from classification of land use/land cover from remotely sensed data are maps similar to images. For information extraction, Jensen (1996) stated that classification is one of the most often applied methods in studies such as this thesis research. According to Campbell, image classification has two fundamental approaches that are widely applied: supervised classification and unsupervised classification. Both of them have some advantages and disadvantages over each other.

First, supervised classification consists of the processes of using pixels that have been previously identified to classify and to identify other unknown pixels. Those identified pixels located within a training site that were delineated on the digital image are the input pixels to the supervised classification. The information from input pixels is utilized to formulate the classification algorithm for allocating pixels to appropriate classes based on specific combinations of spectral values as established by the input pixels (Campbell 2007). Again, this method has some advantages and disadvantages.

The first advantage of supervised classification is that the user and analyst can determine and select the information categories to a particular target and a specific geographic region. The second advantage of supervised classification is that the method is able to identified specific areas which were determined formerly during the process of

33

choosing training areas. Finally, to determine the classification process if it is correct or not, the analyst can detect errors in the classification method by inspecting training data and compare them with other known information or auxiliary data.

In contrast, supervised classification has some disadvantages. First, the classes that the analysts defined previously might not go with the classes from existing data.

Second, when the areas that the analysts want to classify are large and complicated, the training data that are chosen by them might not be representative of conditions encountered. Third, in spite of all resources that are at hand, the process of selecting training data is tedious, expensive and time consuming. Finally, special or unique categories that the analysts do not take into account during the process of choosing training data might not be identifiable by the supervised classification method. This may be because the analysts could not identify them or they might only occupy a tiny area on the image.

In contrast to supervised classification, the second method is unsupervised classification in which “classes are automatically defined based on statistical clustering of pixel values found in the data” (Stan 2005). This means that the analyst allows the computer to group pixels with similar spectral characteristics into unique groups by using numerical operations as defined by established algorithms. Then, the analyst will begin to assign each group of his/her interest into new information classes. However, similar to supervised classification, unsupervised classification also has its advantages and its disadvantages.

34

The first advantage of unsupervised classification is that it is not required to have many details of the region but general knowledge of the study area is needed to better interpret the meaning of the results. The second advantage of unsupervised classification is related to human errors. Comparing to supervised classification, unsupervised classification has the opportunity to reduce human errors. Finally, unsupervised classification allows the distinct units to be recognized as unique classes.

In contrast, unsupervised classification has some disadvantages. First, unsupervised classification limits the user to controlling the menu of information classes.

Second, the relationship that defines for one remotely sensed image will not be applicable for other images. Also, information classes and spectral classes do not have steady relationships. Finally, unsupervised classification classifies and identifies class data that may not be needed or interested in by the users in their categories.

After comparing and studying both supervised classification and unsupervised classification, I decided that the unsupervised classification method be used in this study for classifying images into land cover data layers because of its advantages such as potentially reducing human error and recognizing unique classes as distinct units.

Moreover, the unsupervised classification method may assist the analytical processes by suggesting the specific numbers of classes to be selected by the actual distribution of the pixel values. Three new classes such as urban, agriculture, and barren land were chosen as the land cover categories. It should be noted, however, that, to evaluate and assess the information that obtained from remotely sensed images, an accuracy assessment test was

35

utilized. The test examined the accuracy of the final result that assists the analyst to avoid errors that will lead to undesirable consequences.

After finishing the unsupervised classification, GIS was utilized to evaluate and detect the urban growth during the study period. The study used some of the spatial analysis functions which involve using data from one or more layers to produce a new output layer by overlays. Spatial analyses have the ability to do this. They are the most often applied tool in GIS to solve problems such as identifying high crime areas and selecting the best locations for new businesses (Bolstad 2005).

Specifically, the study applied the most powerful tool in spatial analysis, which is the overlay and raster calculator functions, to detect changes between overlaid data layers. For example, the overlay and raster calculator functions allow the analyst to obtain new information by combining, multiplying or subtracting data from multiple layers to produce a new single layer and the new information that dose not exist in any single layer. It is also necessary to use a common coordinate system for data layers to avoid errors when you want to use overlay and raster calculator operations. However, to determine the changes and differences in the urban growth, one classified layer was subtracted from the other by using raster calculator functions.

Finally, the global measurement of spatial autocorrelation, Moran’s I, was utilized to measure, quantify and examine the spatial distribution of the detected changes and to see if the changes of the geographical patterns occurred in clustered, dispersed, or random fashion. Moran’s I can be applied to variables or data that measured at interval or ratio scales (Wong and Lee 2005). Moran’s I works by testing both feature locations and

36

the attribute values at the same time. It determines and tests if the distribution of the spatial patterns is clustered or dispersed based on the range values. The range values of

Moran’s index are between +1 and potentially -1. For instance, when we find, a polygon pattern, that Moran’s index value is a positive value, we can describe the relationship as having a positive spatial autocorrelation. In contrast, in a pattern of a negative spatial autocorrelation, we can find the Moran’s index value is below its expected value of -1/

(n-1) where n is the number of objects in the distribution of the studied phenomenon. In other words, a positive Moran’s index value means the distribution of the spatial patterns is clustered together whereas the distribution of spatial patterns is dispersed if the

Moran’s index value is a negative. A final point about Moran’s I is that the distribution of the index values is not symmetric to 0 as the index values of correlation coefficients. This is because the expected value of a random pattern (non-cluster, non-disperse) is -1/(n-1), depending on the size of the n. Consequently, it is important to calculate a Z score of the

Moran’s I, using the estimated standard error of the index value given the number of objects in the pattern. In other words, the index values alone are not useful. The index values have to be coupled with their corresponding Z scores so that we can determine if there are statistically significant differences of the measured pattern from a theoretically constructed random pattern.

To conclude, the work of this method, the calculated statistics with their Z-scores are the basis utilized to evaluate the significance of the index value. With the Z values and their corresponding probabilities, it is possible to determine if any clustered or dispersed pattern being statistically different from a random pattern. However, the null

37

hypothesis of spatial autocorrelation assumes that the distribution of a spatial pattern is not significantly different from a random pattern. The critical Z-score values in the case of using 95% confidence level are -1.96 and +1.96. For example, when the Z-score is located between the critical values, we can’t reject the null hypothesis. In contrast, in order to reject the null hypothesis, the Z-score must be located outside the range of the critical values.

To examine if there exists any geographic variation in the pattern of urban growth

(spatial extent, temporal trend, and categorical shifts), I calculated the changes in land use land cover with the extent of Kuwait City at different levels of geographic resolution.

First, the entire city was included in the analysis. Next, the city was partitioned into several smaller geographic units based on administrative boundaries. Results from these two levels of studies were compared to see if results are consistent across geographic resolutions.

While carrying out the research, PCI software was used to do the unsupervised classification of satellite images. The study has acquired TM images for the study area for 1989, 1996, 1998 and 2001. For GIS procedures, ArcGIS 9.2 was used to detect changes and to convert raster images into polygonal structures. GIS was also utilized to evaluate and examine the distribution of spatial patterns changing and to see if the changes of the geographical patterns occurred in clusters, dispersion or randomly.

In summary, the study first classified a series of satellite images for land cover information to categories related to urban growth. The classified images were then overlaid to detect inconsistencies between them. These inconsistencies represent changes

38

in land cover over time. They used to calculate Moran’s index statistics. Between each pair of images, a set of Moran’s index statistics was calculated based on detected changes. Finally, the trend of how the calculated Moran’s index statistics was constructed to show how the urban growth occurred in Kuwait City. These steps were repeated in each geographic resolution level.

CHAPTER 6

DATA PREPARATION, ANALYSIS AND RESULTS

This chapter describes the processes of data preparation and analysis in this study. These processes were used to analyze and detect the urban growth of Kuwait City. Both Remote

Sensing and Geographic Information Systems functions were described in this chapter.

Also, the chapter presents the results that obtained from these analyses.

6.1 Data Preparation and Remote Sensing Analysis:

The main source of data used for the study area includes four Landsat TM (thematic mapper) images acquired on 8/20/1989, 8/7/1996, 4/14/1998 and 5/25/ 2001. Each one of these images has all seven bands, with a spatial resolution of 30 meters per pixel. To process all the imagery, PCI software was utilized. Two images (1989 and 2001) cover the entire Kuwait City and the surrounding area while the other two images (1996 and

1998) have some areas missing. The 1996 image covers everything except Jahra which is a small area on the west side of Kuwait City. I replaced the missing area by applying a mosaic from the 1998 image with PCI software. In other words, the two images (1996,

1998) were mosaiced together which produced a new image that covers all the Kuwait

City. Mosaic is the process that combines multiple adjacent input raster layers into one raster layer. The newly created image is from the 1996 image except Jahra which came from the 1998 image. Figures (2, 3, 4) show the Landsat images of Kuwait City.

39 40

Figure (2) shows Landsat image 1989 of Kuwait City and the surrounding area.

41

Figure (3) shows Landsat image 1996 of Kuwait City and surrounding area. The circle shows that urban area is from landsat image 1998. This image was generated by applying mosaic.

42

Figure (4) shows Landsat image 2001 of Kuwait City surrounding area.

43

Generally, when you download Landsat satellite image data, layers or bands come in separate files. In a typical remote sensing analytical process, all these bands were merged into a single file to obtain useful information. I then clipped Kuwait City and surrounding areas from all the images for two reasons.

1) The images covered a large area that was not needed in my analysis or was in my

study area (part of Saudi Arabia, Iraq and Iran). For this reason, l clipped all images

so that I only retain the parts of Kuwait City for the datasets to be more manageable

in size.

2) Classification accuracy. For example, I clipped The Arabic Gulf from all images by

using an irregular shape polygonal boundary. The irregular shape was based on a

digital map of Kuwait City. The reason I restricted the Arabic Gulf from the

classification processes was that the reflection from the Arabic Gulf or any water

bodies might affect the classification accuracy. Also, the Arabic Gulf is not in the

study area. All images were georeferenced and rectified to 1984-UTM (Universal

Transverse Mercator) Zone-39 N map projection in order to get an accurate analysis

results because my analysis depends and relies on pixel comparisons.

Unsupervised classification method with the Iterative Self-Organizing Data

Analysis Technique (ISODATA) was applied for classifying images into land cover data layers and to define land cover and land use in the study area. During the classification processes, four reflective bands (visible and near infrared bands) were utilized for all images. Each satellite image of Kuwait City from Landsat TM data was automatically

44

divided into 100 clusters. After all pixels of the images were assigned into clusters, I manually assigned each clusters into 3 new land use/cover classes:

1) urban class, which includes industrial, commercial, residential and all other built-up

areas,

2) barren land class, which includes all of the Kuwait Desert and all the empty land that

lies inside or outside Kuwait City,

3) Parks and trees class, which includes all the trees, public and private parks and farms.

In other words, the resulting clusters were assigned into three new classes (urban, barren land and agriculture). See figures (5, 6, 7) that show unsupervised classification outputs.

45

Figure (5) shows land use and land cover map of Kuwait City in 1989 obtained by unsupervised classification method.

46

Figure (6) shows unsupervised classification output of 1996 images and the land use land cover of Kuwait City.

47

Figure (7) shows unsupervised classification output of 2001 image.

48

To evaluate and assess the information that was obtained from remotely sensed images, the accuracy assessment test was done twice for all classification images: first for all image classes and then again for the urban class identified. For first assessment (all classes), 120 randomly selected points were chosen and evaluated. An overall accuracy assessment test are 95% for 1989 classification map, 96% for mosaic classification map

(1996, 1998) and 95% for 2001 classification map. Tables (1, 2 and 3) show the accuracy assessment for all classification images.

The second accuracy assessment test was applied to check the classified maps of urban class alone. Fifty randomly selected points were chosen and assessed. The reason for choosing and assessing the class alone is that the main intention of this study is to detect changes in urban land use/land cover and to examine the changing spatial patterns of urban growth in and around Kuwait City between 1989 and 2001. An overall accuracy for urban class is of 96%, 95% and 98% for 1989 classification map, mosaic classification map (1996, 1998) and 2001 classification map, respectively. Tables (4, 5 and 6) show all accuracy assessment results for the urban class.

49

Table (1) Error Matrix for all classification classes 1989 Reference Data 1989 Urban Barren land Parks Totals Data Urban 14 1 0 15 Barren land 4 98 1 103 Parks 0 0 2 2 Totals 18 99 3 120

Table (2) Error Matrix for all classification classes 1996 Reference Data 1996 Urban Barren land Parks Totals

Data Urban 14 1 0 15

Barren land 3 101 0 104

Parks 0 0 1 1 Totals 17 102 1 120

Table (3) Error Matrix for all classification classes 2001

Reference Data

2001 Urban desert Parks Totals Urban 12 2 0 14 Data Barren land 3 102 0 105

Parks 0 0 1 1 Totals 15 104 1 120

50

Table (4) Error Matrix for all urban class 1989 Reference Data

1989 Urban Barren land Parks Totals Urban 48 1 1 50 Data Barren land 0 0 0 0 Parks 0 0 0 0 Totals 48 1 1 50

Table (5) Error Matrix for all urban class 1996

Reference Data 1996 Urban Barren land Parks Totals Urban 48 2 0 50 Data Barren land 0 0 0 0 Parks 0 0 0 0 Totals 48 2 0 50

Table (6) Error Matrix for all urban class 2001 Reference Data

2001 Urban desert Parks Totals Urban 49 1 0 50 Data Barren land 0 0 0 0 Parks 0 0 0 0 Totals 49 1 0 50

51

6.2. Neighborhood Operations:

Neighborhood operations are among the most important and powerful operations in raster analyses that can be applied to raster data layers for extracting various features during analytical procedures. Typically, a new raster layer would result from the application of neighborhood operations to an input raster layer. The cell values in the output raster depend on how the ‘neighborhood’ is defined. For example, a neighborhood can be defined as a ‘window’ such as a 3-by-3 window or a 5-by-5 window of raster cells that form a neighborhood. Consequently, the output from applying neighborhood operations to a raster layer would be different, depending on the window’s size and also on the window’s configuration.

When applying neighborhood operations, a window is first defined with a size such that a focal cell is at the center and is surrounded by neighboring cells. For example, a center cell in a 3-by-3 window is surrounded by 8 neighboring cells to the East,

Southeast, South, Southwest, West, Northwest, North, and Northeast directions. The configuration of the window refers to how the new value for the focal cell is to be calculated. For instance, the central cell could be assigned a new value that is the average

(or the maximum, the minimum, or other statistics) of the values of the 8 neighboring cells. A detailed discussion of various options for window sizes and window configurations can be found in Chang (2006).

When applying a neighborhood operation to a raster layer, the procedure starts with applying the window to the upper-leftmost cell (or the northwest corner) of the raster layer. In sequential order, the window is moved to the right (or east) until the end

52

of the row in the raster is reached. At which stage, the window is moved down by one row and the moving of window is resumed eastward from the western end of this next row. This process is repeated from left to right and from top to bottom until all cells in the raster layer are visited. Figure (8) shows a moving window 3 by 3 that starts from the northwest corner of the raster layer, moves to the right (east), to the bottom (south) until the raster layer is exhausted.

Figure (8) Order of movement of a 3-by-3 window for neighborhood operations. (Source. Bolstad. 2005 GIS Fundamentals: A first text on Geographic Information Systems, p 363).

This operation is also referred to as filtering or applying moving windows.

However, it should be noted that the selections of which shapes and sizes of the windows

53

to be used in this neighborhood operation are often determined by the phenomena being studied. This is because the shape, size, orientation, and composition of raster clusters that describe the studied phenomena often dictate what the most suitable filters may be.

When no knowledge about the studied phenomena is available in priori , a most commonly used form of a filter is a square shape of 3-by-3 cells. Other alternatives include rectangle, circle, annulus and wedge.

A rectangle shape can be a square or a rectangle, depending on the width and height of the cell units. For example, if you select to use a 3-by-3 or a 5-by-5 moving window, the result from neighborhood operations will be a square. Alternatively, if a 3- by-6 moving window is applied, the result will be a rectangle. A circle is defined by its radius from the center cell while an annulus is defined by two radiuses (inner and outer) from the center cell to create two circles. An annulus is also called doughnut shape. A wedge is defined by a part of circle from the center cell. See figure (9) for examples for the alternative shapes of possible moving windows.

54

Figure (9). Vary shapes of filters (source.Chang. introduction to Geographic Information Systems. P. 264).

These types of neighborhood operations may be configured with specific statistical methods that essentially use a moving window to derive the final output in raster format. For example, descriptive statistics such as the mean, the range, the maximum, the minimum or the majority can be defined as the new cell value for the focal cell with the statistic calculated from the values of the neighboring cells. In the case of replacing the value of the focal cell with the mean of the values of the neighboring cells, the mean function assigns the mean value of the neighborhood cells to the target cell by calculating the average of all cell values within the neighborhood. If a 3-by-3 window is applied, the mean function will compute the mean value of the nine cells in the window and assigns the value to the focal cell as the new cell value. The most important benefit

55

from using the mean function is that this process decreases the difference of the cell values among cells in the neighborhood, thereby smoothing the data layer (Bolstad 2005).

Another example of a neighborhood function is that a moving window can be configured for edge enhancement. Edge enhancement helps to increase the contrast between a cluster of raster cells that share similar characteristics and those cells neighboring the cluster. The calculation of cell values in a moving window for edge enhancement is based on the range as calculated from the difference between the largest and smallest value in the window. The range is assigned to the targeted focal cell. In this manner, the resulting raster layer often sees enhanced (exaggerated) differences between cells that form boundaries of cell clusters.

A final example for moving windows based on a neighborhood statistics function is a configuration that uses the minimum or the maximum value in the window for the focal cell. The minimum function computes the smallest value among neighborhood cells whereas the maximum function calculates the largest value among neighborhood cells, subsequently assigning the calculated value to the focal cell. Figure (10) shows all these methods and gives us a clear explanation.

56

Figure (10). (Source. Bolstad. 2005 GIS Fundamentals: A first text on Geographic Information Systems , p 364).

Finally, a very useful type of moving windows is a neighborhood operation that replaces the value of the focal cell with the value that is the majority value of the surrounding cells in the window. In other words, the majority function evaluates all the cells that surround the focal cell and assigns it the value occurs most frequently. For example, if a 3-by-3 window is selected, the majority function will find out the majority value of the nine cells in the window and assigns the value to the target cell as the new value. The largest advantage from applying majority is that the function can be applied to smooth the data variation by reducing the polygon numbers. This function is also called

57

smoothing operation. However, the focal or target cell will be given no data on the final

out put if the majority value within the surrounding cells is more than one.

Huffman (1996) applied a majority filter operation with a 3-by-3 window for remotely sensed land cover data for the purpose of agricultural studies. One of the aims in his study was to increase the classification accuracy of land covers by reducing the polygon numbers and the effect of visual noise. After he classified the remotely sensed data, he obtained raster data format. To achieve his goal, he showed that it is necessary to convert raster to a vector format. He also suggested that filter operation is often applied for classified data before vectorization. He found that the operation creates a new layer that contains one class in each region which means he reduced the polygon numbers.

Furthermore, he stated that the overall accuracy increased about 15% from 75.4% to

89.4% as a result from applying the filtering procedure.

In summary, neighborhood operation, also known as filtering, can be considered one of the most important operations in raster analyses that would assist the analyst to generate a new output layer by smoothing the data. It helps to reduce an excessive number of polygons and to remove visual noise. Neighborhood operation generates a new output by moving window through out the input raster layer with the selected statistical methods. One of the most important statistical methods is using the majority among cell values in the defined neighborhood for each focal cell. This procedure is useful especially when converting data layers from a raster format to a vector format.

58

6.3. Applying Neighborhood Operation (Filtering):

Neighborhood operations have many advantages when applied to raster data layers. For

example, they can be applied to smooth the data variation and reduce the difference

between a cell and its surrounding cells (Bolstad 2005). Specifically, raster data layers

that contain noise from classification procedures, from errors in image detection, from

errors in scanning, etc., can be improved by applying appropriate neighborhood

operations.

For this study, there is apparent noise from unsupervised classification of satellite

images used in the analysis. There are single-cell clusters that are clearly not reflecting

land cover correctly and there are small variations of cell clusters that do not agree with

auxiliary data (aerial photos and surveyed maps). Consequently, it is necessary to apply

neighborhood operations to remove them from subsequent analysis, especially before

vectorization. Vectorization is the process of converting raster dataset from raster format

to vector format. If single-cell clusters are not handled, an excessive number of tiny

polygons would be produced from the process of transforming raster to feature. In this

situation, a majority filter is often preformed for removing noise from classified images

and reducing the number of small polygons after vectorization.

I have utilized one of these statistical methods which are the majority operation. I have applied rectangle (square) type by setting the window as 3 by 3. For the purpose of removing single-cell clusters, a 3-by-3 window is more appropriate than a 5-by-5 or bigger window because larger windows can smooth the data more than smaller windows can. Larger windows can also cause errors by deleting some of the smaller clusters of the

59

urban class in my study area that are larger than single-cell clusters. This would affect the accuracy of the final result.

In addition, neighborhood function or filtering was applied twice in the study area. First, I applied filtering to all the data layers that resulted from classifying remotely sensed images. Second, filtering was also performed to all the new images that generated from the first time filtering. The main reasons for performing and repeating the filtering process the second time are that this process decreased the number of tiny polygons that scattered all over the study area. For example, I noticed that the final output from performing filtering for the first time created a large number of tiny linear polygonal features. They were apparent noise that might lead to additional errors in the subsequent analyses.

Repeating filtering again helped to reduce the number of polygons and smoothed the final output by eliminating the noise further. For this reason, applying second filtering helped to reduce noise further. See the table below that shows the changes in the numbers of polygons after converting raster layers to vector format. It also shows the difference between first and second time filtering.

Table 7: Number of Polygons after transforming raster to vector format

1st time filtering 2nd time filtering 1989 Classified Image 3107 2001 1996 Classified Image 4221 2007 2001 Classified Image 4334 2443 The difference between 1996 and 1989 2537 1633 The difference between 2001 and 1996 3513 2281

60

As can be seen in Table (7), each application of majority filtering resulted a significant reduction in the number of polygons in each image. Upon close inspection and comparison of the results, the removed polygons were found to be mostly single-cell polygons and small linear polygons, both are apparently noise occurred in the unsupervised classification. See the figure (11) shows the difference between the image before and after filtering operation.

61

Figure (11) shows and compares the difference between filtering operation.

62

In conclusion, the majority operation determines the value of focal cell by

evaluating all the surrounding cells and then giving it the value that occurs most often.

There are some reasons why I chose this technique: one is to create a smoother output

from the original output. Second, it is used to remove the pixels from one class that

situated inside another class. In other words, this operation is used to reduce the number

of polygons before transforming the format from raster to vector for subsequent spatial

analysis. The largest benefit from using this technique and all of these processes is that it

allows me to the increase the accuracy of the final result.

6.4. GIS Analysis:

For GIS procedures, ArcGIS 9.2 was used to detect changes and to convert raster images into polygonal structures. GIS was also utilized for the last part of the study by applying

Moran’s index to calculate if the spatial patterns and process of these changes take place in a random fashion or with certain identifiable trends. Some functions were preformed to detect the urban expansions before converting raster to vector format by using spatial analyst tools and toolbox such as mask, raster calculator, reclassify and Moran’s I functions.

First, a mask was utilized to include in the analysis only the classes within the

study area (Kuwait City boundaries). The mask was created by using Kuwait

governorates layer as the guiding boundary. This function gives all the cells that are

located inside the study area a valid value to be included in the analysis whereas the cells

63

that located out of the study area were given value of No data. This function allows the analyst to prevent the cells that out of the study area from further analyses.

Second, one of the most important functions in spatial analyst tools is the raster calculator. The concept of this function is that it is based on cell by cell combination between layers. Raster calculator is also called map algebra that performs Boolean, logical and combinational operations. However, to evaluate and detect the urban growth during the study period, one of the logical operations in raster calculator called DIFF was utilized. The function calculates the difference between two layers and assigns the output into a new layer. For example, the DIFF operation will assign zero for the final output if the cell values on both input layers are similar. Otherwise, the value on the first input will be assigned to the final output. I applied DIFF operation by using a raster calculator to obtain the difference of the land cover and land use between (1996, 1989) and (2001,

1996).

Third, the reclassify function that is located on spatial analyst tools was used.

Reclassify is the process or the function that is used to convert the cell values into new output values. For example, the function is an important process that is used to group many classes into one class such as grouping residential, industrial and commercial areas into one classification such as urban. The function is also utilized to classify one class into many subdivision classes. For example, we can perform the reclassify operation to classify population density maps into three regions such as low, medium and high density. I performed the reclassify function to group two classes into one class.

Specifically, the agriculture class that contains all trees, private and public parks with

64

urban class which contains all built-up areas was grouped together in one class. The main

reason why I combined these classes together Are: First, due to the Kuwait weather, all

the trees and vegetations in the agriculture class didn’t grow naturally. In other words, all

these trees and vegetations were planted by humans. For this reason, I reclassified them

with the urban class to be one class.

Finally, after determining the difference of the urban growth between (1996,

1989) and (2001, 1996), the global measurement of spatial autocorrelation, Moran’s

Index was applied to evaluate and investigate the spatial distribution of urban expansions and to see if the urban growth occurred in clustered, dispersed, or random fashion. For the studies purpose, the function of Moran’s Index was applied twice in the study area one for grid and the other for polygons. The function was applied to compare the urban growth between two different periods. The first period was from 1989 to 1996 and the second period was from 1996 to 2001. The function determined if the distribution of spatial patterns was clustered together or dispersed based on the Moran’s index values.

The Moran’s index values, coupled with the corresponding Z scores, give indications as to whether the observed pattern is a cluster, a disperse, or a random pattern.

6.5. The Results of Urban Growth in Kuwait City:

Due to the population growth and development, the results of Remote Sensing and

Geographical Information Systems revealed that the urban areas of Kuwait City increased from 243.6 km ² in 1989 to 318.4 km ² in 2001. The results also indicate that the rapid urban expansion has taken place in Kuwait City and surrounding areas during the study

65

period. It is very obvious that the phenomenon of urban growth occurred in the study area between 1989 and 2001. The built-up areas expanded by 10% from 32.4% in 1989 to

42.4% 2001 of the total areas. Table (8) and figure (12) illustrate the results of GIS analysis.

Table (8) shows the areas of each class in km² between 1989 and 2001

2001 1989 from/to Urban barren land Total 1989 Urban 210.2 33.4 243.6 barren land 108.2 399.2 507.3 Total 2001 318.4 432.5 751.0

Urban Growth in Kuwait City, 1989- 2001

600.0

500.0 400.0 Urban 300.0 barren land Km² 200.0

100.0 0.0 Total 1989 Total 2001

Figure (12) illustrates the urban growth of Kuwait City during the studies period.

66

Furthermore, the results demonstrated that GIS could be an important tool for detecting and identifying the urban growth and land use and land cover changes. Using

GIS functions, a simple map of urban growth was created to show the changes in the study area between 1989 and 2001. See Figure (13) that shows the new urban areas. The map shows that the majority of the urban expansion or land use and land cover changes occurred in the south, southwest and west side of Kuwait City after the forth ring road that surround the old city. The maps also shows that most of urban expansion that occurred in the south side of Kuwait City were located along the major highways such as

King Fahed and Fahaheel highways while the majority of the urban expansions that took place in the west and southwest of Kuwait City was located within the forth ring road and six ring road. The results from GIS pointed out that the major highways have a strong relationship with the urban expansions because many new development areas were allocated along and between these highways. The urban expansions in southwest side were restricted from going farther by Kuwait International airport which is located between ring six and seven roads.

In contrast, the old city and the surrounding areas that located before the forth ring road showed less noticeable urban expansions. The reason why the old city and surrounding areas is having less urban expansions is related to the early master plans or to the earliest urbanization. These master plans directed and allocated the urban growth and the earliest urbanization near and in the old city in the past which means no more land for the future growth. In other words, these master plans left few vacant and limited

67

spaces for new developments in the old city and the surrounding areas. Therefore, we can describe the urban growth that occurred in the old city as infill development.

From this result, it can be concluded that the urban growth of Kuwait City took place in three different stages. First is an infill development, which is located in and near the old city or the CBD area. Second is linear development, which is located along the costal lines to the south and west of old city (CBD) until Shuaiba in the south and Jahra in the west. Finally, clear pattern developments appear between the forth and six roads in the west and southwest side of the city and along the majors highways to the south of

Kuwait City.

68

Figure (13) shows the new developed areas between 1989 and 2001 in and around Kuwait City.

69

6.6. Evaluating the Distribution of Spatial Patterns Changing:

Moran’s Index was applied to test and find out the distribution of the spatial patterns if the new developments grow in clusters, dispersed or random manner. The study period was divided into two different periods. The firs period is from 1989 to 1996 and the second is from 1996 to 2001. Also, the Moran’s index was applied to measure the spatial clusterness of the urbanized pixels in both grid form and polygons. First, the entire

Kuwait City was tested in the analysis. Second, the Kuwait City was partitioned into several smaller geographic units based on administrative boundaries. For example, the

Kuwait City has six governorates to test Ahmadi, Al-Asimah, Farwaniya, Hawalli Jahra and Mubarak-Alkabeer.

6.7. First Moran’s Index (Grid):

For both the first period (1989-1996) and the second period (1996-2001), Moran’s indices are calculated for the entire Kuwait City and the six governorates. All these calculations performed on the spatial data of the Kuwait City are reported in table (9) and figure (14). Please note that the ArcInfo Workstation, a command-driven GIS software, does not offer Z-scores for the calculated Moran’s index values. Because of this limitation, the Moran’s index values are reported here without corresponding Z-scores.

While they cannot indicate if statistical significance exceeded critical values, they nevertheless allow comparisons be made among calculated geographical extents.

70

Tables (9) shows the values of Moran’s index for the urban growth in Kuwait City

Difference between Difference between 1996- 2001-1996 1989 Governs Moran's Index (I) Moran's Index (I) Ahmadi 0.760 0.757 Al-Asimah 0.767 0.867 Farwaniya 0.794 0.776 Hawalli 0.721 0.740 Jahra 0.793 0.876 Mubarak-Alkabeer 0.792 0.894 The entire Kuwait City 0.778 0.865

Moran's Index (Grid)

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

Moran's Index values Index Moran's 0 Ahmadi Al-Asimah Farwaniya Hawalli Jahra Mubarak- The entier Alkabeer Kuwait City Governorates

Difference between 2001 and 1996 Difference between 1996 and 1989

Figure (14) shows the differences of Moran’s values between each period.

71

Moran’s index values show and indicate that positive spatial autocorrelation for all the inputs data in both the first and second study periods. For all the entire Kuwait

City, the first study period (1989-1996) has stronger spatial autocorrelation than the second period (1996-2001). After comparing the Moran’s index values in the first and the second study periods for all governorates, the six governorates were divided into two groups. The index values of first group are slightly different whereas the index values of second group are considerably different.

The first group, which is their values slightly different in each period, consists of three governorates Ahmadi, Farwaniya and Hawalli. Although the urban area was increased in each period, the index value of these governorates shows no big difference between them which means that the urban growth that occurred in each governorate and in each period took place at the same manner. Moreover, this indicates that the distribution of spatial patterns changing and process of these changes in each period took place in a cluster fashion.

The second group containing Al-Asimah, Jahra and Mubarak-Alkabeer has different values between each period. The index values of all these governorates show that the distribution of the spatial patterns changing and the urban development occurred in cluster manner. The index value of these governorates also indicates that the first period has stronger spatial autocorrelation than the second period. The reason for that relates to the new urban areas that developed in Mubarak-Alkabeer governorate south of

Kuwait City between 1989 and 1996. The government built four residential areas at the

72

same time (Aladan, alqusour, alqurayn and Mubarak-Alkabeer) mainly for Kuwaiti people.

Furthermore, more developed urban areas were built in south and west of Al-

Asimah governorate in the first period than the second period. Most of these developments were built in Surra and Qurtobah in the south of Al-Asimah governorate and Doha, Suwaikh Health and Suwaikh Education in the west of Al-Asimah governorate. Finally, Jahra governorate has more index value in the first period because the government developed the areas surrounding it.

In summary, the calculation of global Moran’s I statistics for grid data were utilized to test the spatial patterns of the urban growth in the entire Kuwait City and six governorates indicate that there is a strong positive autocorrelation of all the inputs data during the study period. This also indicates the urban growth of the Kuwait City occurred in cluster manner.

6.8. Second Moran’s Index for Polygons:

After converting raster format to vector format for both the first and the second study periods, the Moran’s index was also utilized to examine the spatial distribution patterns of the urban growth in Kuwait City. The Moran’s index was applied for all the entire

Kuwait City and six governorates in each period. The results showed that all the index values were positive. These positive values indicate that positive autocorrelation for all the inputs data during the study period. In other words, the urban growth that took place in the entire Kuwait City and six governorates was clustered together. Table (10) and

73

figure (15) show the Moran’s index values for both the entire Kuwait City and six governorates into different periods.

To test, the work of this method, the calculated statistics with their Z-scores were utilized to assess the significance of the index values. With the Z values and their corresponding probabilities, it was possible to determine if any clustered or dispersed pattern being statistically different from a random pattern. However, the null hypothesis of spatial autocorrelation assumes that the distribution of a spatial pattern is random. The critical Z-score values in the case of using 95% confidence level are -1.96 and +1.96. For example, when the Z-score is located between the critical values, we can’t reject the null hypothesis. In contrast, in order to reject the null hypothesis, the Z-score must be located outside the range of the critical values. Figure (16) and table (11) display the Z score values.

The results showed that the critical values were significant (2.58) for all input units which means the observed patterns were more clustered than random. After reporting the Z values and their corresponding probabilities, I reject the null hypothesis for the spatial autocorrelation statistics, Moran’s index, and determine that the distribution of index values over the entire Kuwait City and six governorates was not a random pattern. The critical values also indicate that there is less than 1% likelihood that this clustered pattern could be the results of random chance. Figures (17 and 18) show

Moran’s index and Z score values graphically.

In summary, this chapter has focused on the data preparations and analysis for both Remote Sensing and Geographic Information Systems. The chapter has also focused

74

on the results and findings during the study period. The preparations started with analyzing the landsat images by using PCI, remote sensing software. Unsupervised classification method was applied to extract information of land use and land cover to use as source of data. After the works of analysis the satellite images were done, the accuracy assessments for all the outputs obtained from classified images were tested to assess the outputs. After the works of remote sensing were completed, GIS tool was utilized to identify the urban growth and measure the spatial patterns of such changes. First, filtering operation was applied to remove the noise from unsupervised classification of the satellite images. Second, the spatial analyst tools and toolbox were utilized to specify the study area and detect the urban expansion. Finally, Moran’s index was utilized to assess and examine the spatial distribution of urban growth. However, the results and findings from the analysis above show that there was a remarkable urban growth in and around

Kuwait City between 1989 and 2001. Moreover, the Moran’s index indicated that positive spatial autocorrelations for all the inputs units during the study period which is mean the urban growth in Kuwait City developed in cluster manners.

75

Table (10) shows Moran’s index values for all the study area units

Difference between 2001- Difference between 1996- 1996 1989 Moran's Index Moran's Index Ahmadi 0.3 0.68 Al-Asimah 0.21 0.29 Farwaniya 0.03 0.11 Hawalli 0.16 0.36 Jahra 0.12 0.28 Mubarak-Alkabeer 0.11 0.2 The entire Kuwait City 0.14 0.22

Moran's Index Polygons

0.7 0.6 0.5 0.4 0.3 0.2 Moran's Index Moran's 0.1 0 Ahmadi Al-Asimah Farwaniya Hawalli Jahra Mubarak- The entier Alkabeer Kuwait City Governorates

difference bteween 2002-1996 Difference between 1996-1989

Figure (15) illustrates Moran’s index values during the studies period.

76

Table (11) displays the Z score values equal standard deviations.

Difference between 2001- Difference between 1996- 1996 1989 Z-score = Standard deviations Ahmadi 49.59 63.52 Al-Asimah 29.38 31.38 Farwaniya 5.31 13.05 Hawalli 19.46 30.48 Jahra 11.87 24.96 Mubarak-Alkabeer 14.03 14.78 The entire Kuwait City 83.62 85.9

Z-Score of Moran's Index Polygons

100 80 60 40 20 0

li a adi City niya t hm simah a Jahr A rw Hawal l-A a uwai A F r K tie Mubarak-Alkabeeren he T Governorates

Difference between 2001-1996 Difference between 1996-1989

Figure (16) displays Z-score of Moran’s I for the study areas.

77

Figure (17) shows the results of the first period 1989-1996 by applying Moran’s index polygons for the entire Kuwait City.

Figure (18) shows the results of the second period 1996-2001 obtained from Moran index polygons to entire Kuwait City.

CHAPTER 7

CONCLUSION

Urbanization is a complex process that refers to the expansion of the cities. Many mega cities around the world are influenced by this phenomenon. However, after oil exploration in Kuwait, many people immigrated to the Kuwait which changed the size of the Kuwait City from a small Islamic city to a new modern city like those in western countries (Almnyes 1994). Also, the population of Kuwait has increased rapidly since the oil exploration. This further expended the size of the Kuwait City. It encouraged the government to adopt several master plans to help it to control immigration and to rebuild the Kuwait City to be a modern city. After studying the expansion of Kuwait City, it was concluded that the urban growth has taken place in the city between 1989 and 2001. This is because urbanization process is a global phenomenon that takes place all over the world (Sudhira 2004), and the Middle East is no exception.

This study demonstrated the benefits of using both Remote Sensing and

Geographic Information Systems to identify and analyze the urban growth during the study period. The Remote Sensing was applied to monitor, examine and identify the growth and changes in land use/land cover because this tool can provide the analyst with primary data whereas the GIS was utilized to detect the urban growth and the distribution of spatial patterns.

78 79

The results of the study indicated that the changes in the spatial patterns in Kuwait

City have taken place in a clustered manner. The results also illustrated the urban growth of Kuwait City as it has taken in three different stages. First, infill development was found to be located in and near the old city. Second, linear development was found along the coastal regions. Third, recent developments can be clearly identified to be between the major highways in the south and west of old city.

In order to remove the noise from the classified images and reduce the number of tiny polygons after vectorization of the original satellite images, neighborhood operation

(filtering) was applied. This operation is one of the most important operations in raster analyses that can be applied to raster data to smooth the data variation and improve the quality of raster images. Filtering was preformed twice. First, filtering was applied to all the data layers that obtained from classifying remotely sensed images. Second, filtering was also applied to all the new outputs that generated from the first time filtering. The second time filtering reduced the number of tiny polygons that scattered all over the study area and smooth the final output more. Applying the filtering second time is really helpful to eliminate the noise further and increase the accuracy of the final result.

Moran’s I is a summary statistics that can be applied to examine the distribution of spatial patterns due to the urban growth by testing both feature locations and attribute values simultaneously. It reveals and determines if the distribution of the spatial patterns is clustered or dispersed based on range of calculated index values. The Moran’s I was applied for all the entire Kuwait City and six governorates for the detected changes in land use/land cover over the study period. The results revealed that all the index values

80

were positive which mean that urban growth in the study area trend to be clustered

together. The findings from applying Moran’s I can help and allow the policy makers to

understand how the urban growth has occurred during the study period.

The findings from the results can help the policy makers to predict and plan the urban growth for future developments because these findings will provide information for policy makers so that they would know which parts of the city outgrew or fell behind others. Moreover, linking these findings with the number of population live in the new developed areas can provide important information for policy makers about the needs of the major social facilities such as health and education in and around the city. Findings from the results can also be as a guideline for the policy makers to determine the general locations and requirements of future traffic demands. Also, the policy maker can use the guideline to construct new roads and improve the existing roads to avoid traffic congestion and delays.

The extent of urban growth and cities development often went through a lengthy process of urbanization. Urbanization is a universal phenomenon that occurs all over the world. It is among the most important forces that influence the configuration and changes in the patterns of land use and land cover in and around urbanized areas. Urbanization involves and associates the development of the cities with some factors such as economy, population and industrialization. These factors can influence the rate of changes and direction of cities development. For example, the economic growth in any city around the world encourages people to immigrate from rural to urban areas and from country to country seeking better social and economic opportunities.

81

However, Kuwait’s economy has increased rapidly since Kuwait began its practice of exporting petroleum resources. For city planning, Kuwait government brought experts from Britain to help construct long term development plans. Those experts applied some of the urban models that developed in the west such as Concentric Zone

Model, Sector Model and Multiple Nuclei Model to plan the future development

(Almnyes 1994). By studying the urban growth of Kuwait City, the findings indicate that

Kuwait City has the characteristics similar to those cities in Europe and North America.

Furthermore, the urban development of Kuwait City since oil exploration is similar to the urban growth of those cities in the Gulf Cooperative Council (GCC) (Alkhuzamy 2001).

For example, Dubai is a city located on the Arabian Gulf that grew rapidly since exporting oil.

The government also adopted some master plans by bringing in some experts from other part of Europe. This gave an indication that city in the Arabian Gulf may be modeling after those cities in western counties (Gabriel 1987). In other words, the spatial pattern of the changes in land use/land cover in Arabian Gulf Cities is significantly similar to those seen in cities in the West. For these reasons, I believe that the research findings from this study can provide to city planning officials in both the west and in the

Arabian Gulf countries how successful or not so the practice of modeling after western cities has been. Furthermore, I expect to find similar trends in other Arabian Gulf countries. In other words, I expect that the urban growth in other cities in Arabian Gulf would be consistent with the way urban growth in Kuwait City has occurred because the only one who has the authority to plan and construct new settlements is the governments

82

and normally government owns most of the developable lands in the cities. After oil exploration these governments process abundant resources for investing and planning for more structural development than before. Consequently, the governments would need findings such as those concluded here as reference for future decision-making.

If I would like to continue this line of research in the future, I will use high resolution images such as IKONOS and Quick Bird to detect changes in land use and land cover. The high resolution images will help me to increase the classification precision and will provide me with more spatial details than those I could achieve in this study. For instance, these images will allow me to determine all the types of land use land cover. Moreover, if I were to continue this kind of research, I would attempt to study the distribution of population in the study area and link it with the distribution of each type of land use land cover. By using GIS, it link will allow me to make good decisions for future planning and for choosing site locations and solve urban problems.

Finally, this research has proven the benefit of using Geographic Information

Systems and Remote Sensing to study the spatial patterns of urban growth and to examine the changing spatial patterns of urban growth. My suggestion to Kuwait government for future planning is to adopt these tools when the policy makers in Kuwait government want to reviewing previous master plans and when they are planning new master plans. Also, I suggest applying these tools to study the urban growth in and around Kuwait city because these tools reduce the cost and time consuming.

REFERENCES

Abbaspour, M and Gharagozlou, A. 2005. Urban planning using environmental modeling

and GIS/RS: A Case Study from Tehran. Wiley inter Science , 63-72.

Abo Aysh, A. 1981. Planning for cities development in Kuwait. Geographic massages

33 .

Al-khuzamy, A. 2001. Automated cartographic molding of the urban development in

Kuwait. Geographic massages 257 .

Ali and Taher 2004. Study the urban growth of Riyadh City between 1987 and 2001 by

using Geographic Information Systems and Remote Sensing. Geographic

Massages 293 .

Almnyes, W. 1996. The Urban structure of the Kuwait City. Journal of Humanities and

Social Sciences 56 (2): 439-488

Almnyes, W. 1998. The characteristics of CBD of Kuwait City. Journal of Humanities

and Social Sciences 21: 67-124.

Almnyes, W. 1994. The structure of the Kuwait City comparing with models city study in

urban Geography . Journal of Arabic Gulf Studies 7: 19-55.

Alphan, H. 2003. Land use change and urbanization of Adana, Turkey. Land

Degradation and Development. 14: 575-586.

Antrop, M. 2004. Landscape change and the urbanization process in Europe. Journal

Land and Urban Planning 67: 9-26.

83 84

Bolstad, P. 2005. GIS fundamentals: A first text on geographic information systems, 2nd

edition . Minnesota: White Bear Lake.

Bonham-Carter, G. 1994. Geographic Information Systems for Geoscientists Modeling

with GIS, First edition . Ontario: Love Printing Service.

Campbell, B. 2007. Introduction to Remote Sensing . New York: Guilford press.

Chang, K. 2006. Introduction to Geographic Information Systems, Forth edition . New

York: McGraw-Hill.

Cho, S., Chen, Z., Yen, S and English, B. 2007. Spatial variation of output-input

elasticities: Evidence from Chinese county-level agricultural Production data.

Papers in Regional Science 86: 139-157.

Demers, M. 2002. GIS Modeling in Raster . New York: John Wiley& Sons, Inc.

Eshah, H. 2007. Land use trends during rapid urbanization of the City of Aydin, Turkey.

Journal Environ Mange 39: 433-459.

Firman, T. 1999. Rural to urban land conversion in Indonesia during boom and bust

Periods. Journal land use policy 17: 13-20.

Gabriel, F. 1987. The Dubai Handbook . Dubai: Dubai Petroleum Company: 79-

108.

Greene, R. and Pick, J. 2006. Exploring the Urban Community a GIS Approach . New

Jersey: Prentice-Hall.

Hathout, S. 2002. The use of GIS for monitoring and predicting urban growth in East and

West St Paul, Winnipeg, Manitoba, Canada. Journal of Environmental

Management 66: 229-238.

85

Huffman, R. Kborram, S and Stallings, C. 1996. Polygon Mode Filter for Remotely

Sensed Agricultural Land Cover Data. Baros, S and Morain, S (eds.) Raster

imagery Geographic Information Systems . New Mexico: On Word Press. 231-

238.

Jensen, R. 1996. Introductory Digital Image Processing . New Jersey: Prentice Hall.

Juliana, M and John, Z. (2006). GIS for the Urban Environment . California: ESRI Press.

Kali, F., Brian, S. and Marvin, L. 2005. Land cover classification and change analysis of

Twin Cities (Minnesota) Metropolitan Area by multitemporal Landsat remote

sensing. Remote Sensing of Environment 98: 317-328.

Kwarteng, A. Y. and Chavez, P. S. 1998. Change detection of Kuwait City and environs

using multi-temporal landsat Thematic Mapper data. Journal of Remote Sensing

19 (9): 1651-1662.

Lee, J. 2003. Geographic Patterns of Urban Residential Development. In B. Boots, A.

Okabe, and R. Thomas (eds.) Modeling Geographical Systems: Statistical and

Computational Applications . Dordrecht: Netherlands: Kluwer Academic

Publishers. 13-31.

Mundia, C. and Aniya, M. 2005. Analysis of land use/cover changes and urban

expansion of Nairobi City using Remote Sensing and GIS . International Journal

of Remote sensing, 26 (13): 2831-2849.

Ministry of planning. 2005. Annual Statistical Abstract, Edition 42 . Kuwait: State of

Kuwait.

Municipality of Kuwait. 1980. Planning and Urban Development in Kuwait . Kuwait:

86

State of Kuwait.

Municipality of Kuwait. 1997. Third Master Plan . Kuwait: State of Kuwait.

Municipality of Kuwait. 2005. The First Review of the Third Master plan. Kuwait: State

of Kuwait.

Okunuki, K. 2001. Urban Analysis with GIS. Geo journal 52 : 181-188.

Price, M. 2008. Mastering ArcGIS . New York: McGraw Hill.

Stan, A. 2005. Remote Sensing for GIS Mangers . California: ESRI Press.

Sudhira, H. Ramachandra, T. and Jagadish, K. 2004. Urban sprawl: Metrics, Dynamics

and Modeling Using GIS. International Journal of Applied Earth Observation

and Geoinformation 5 : 29-39.

Tian, G., Liu, J. and Xie, Y. 2005. Analysis of Spatio-Temporal Dynamic Pattern and

Driving Forces of Urban Land in China in 1990 using TM images and GIS.

Elsevier 6: 400-410.

Xiao, J., Shen, Y., Ge, J. and Tang, C. 2006. Evaluating Urban Expansion and Land Use

Change in Shijiazhuang, China, by Using GIS and Remote Sensing . Landscape

and Urban Planning 75 : 69-80 .

Weng, Q. 2001. Modeling Urban Growth Effects on Surface Runoff with Integration of

Remote Sensing and GIS. Journal Environmental Management 28 (6): 737-748.

Wong, D.W. S. Wong and Lee, J. 2005. Statistical Analysis of Geographic Information

with Arcview GIS and ArcGIS . New York: John Wiley and Sons. 327-387.

87

Wu, Q., Li, H., Wang, R., Paulussen, J., He, Y., Wang, M., Wang, B., Wang, Z. 2006.

Monitoring and predicting land use change in Beijing Using remote sensing and

GIS. Landscape and Urban Planning 78: 322-333.