<%BANNER%>

Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2010-08-31.

University of Florida Institutional Repository
Permanent Link: http://ufdc.ufl.edu/UFE0022396/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2010-08-31.
Physical Description: Book
Language: english
Creator: Almatar, Muhammad
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Geography -- Dissertations, Academic -- UF
Genre: Geography thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by Muhammad Almatar.
Thesis: Thesis (M.S.)--University of Florida, 2008.
Local: Adviser: Southworth, Jane.
Electronic Access: INACCESSIBLE UNTIL 2010-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022396:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022396/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2010-08-31.
Physical Description: Book
Language: english
Creator: Almatar, Muhammad
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Geography -- Dissertations, Academic -- UF
Genre: Geography thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by Muhammad Almatar.
Thesis: Thesis (M.S.)--University of Florida, 2008.
Local: Adviser: Southworth, Jane.
Electronic Access: INACCESSIBLE UNTIL 2010-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022396:00001


This item has the following downloads:


Full Text





USING REMOTE SENSING AND GIS TO STUDY LAND-USE AND LAND-COVER
CHANGE IN ALACHUA COUNTY, FLORIDA FROM 1993 TO 2003





















By

MUHAMMAD ALMATAR


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA

2008

































2008 Muhammad Almatar































To my parents, wife, and children









ACKNOWLEDGMENTS

I acknowledge and thank Dr. Jane Southworth, for her guidance and supervision of this

research. I thank Dr. Binford for providing me with the research satellite images. I thank my

research committee members and Dr. Waylen. I acknowledge my good colleague, Steven

Forrest, for being supportive. I thank my parents and my wife for their support and

understanding.









TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ..............................................................................................................4

L IST O F T A B L E S ..................................................................................................... . 7

ABSTRAC T ...........................................................................................

CHAPTER

1 INTRODUCTION ............... ................. ........... .............................. 10

B ack g rou n d ................... ...................1...................0..........
Study A rea ............................................................................................ 11
U se of R em ote Sensing and G IS .................................... ........... ............. ..........................11

2 UTILIZING REMOTE SENSING AND GIS TO STUDY LAND-USE AND LAND-
COVER CHANGE IN ALACHUA COUNTY, FLORIDA FROM 1993 TO 2003 .............13

In tro d u ctio n ................... ...................1...................3..........
B background ................................................................................................13
L and U se and L and C over C hange...................................................................... ...... 14
Urbanization and Land Use/ Land Cover Change....................................................14
Use of Remote Sensing and GIS .............. ........... ............... 16
M methodology ........... ......................................... ................... ............... 18
Study A rea .............. .. ....... ..................................... .................. 18
Im age Pre-Processing .......................... .......... .. .. .. ......... .. ..... .. 21
C classification Schem e ......................... ................................ .. ........ .... ...........2 1
Training Samples and Fieldwork ............. ..... ........ .................... 22
Im age Classification ........................ ............ .......................22
Normalized Difference Vegetation Index ..................................... ............... 23
T ex tu re an aly sis ................................................. ................ 2 4
Low pass filter .............. ....... ............ ................................ 24
Tasseled cap transformation..................... ....... ............................. 25
T h erm al b an d .................................................... ................ 2 6
N on-spectral data layers .................................... ................... ............... 26
R ule-based classifier ........................................ ................... ..... .... 27
C change D election A naly sis ..................................................................... ..................29
R results ......... ................................. 30
Accuracy Assessm ent ......... .. .................................. ...... ............................. 30
Change Traj ectories....... ...... ...................................................................... ...... 31
D iscu ssio n ......... .. ................................................ ............................... 3 4
Summary of Finding ....... ......................................... .......... 34
Implication of Study .................... .................... ..................35
C o n c lu sio n ........................................................ ................ 4 0









3 C O N C L U SIO N ........ .......... ............................................................ ......................................67

L IST O F R E F E R E N C E S ..................................................................................... ....................69

B IO G R A PH IC A L SK E T C H .............................................................................. .....................73



















































6









LIST OF TABLES


Table page

2-1 Land use and land cover descriptions for classification scheme. ......................................42

2-2 Final decision based rules used in the final classification procedure. ............................43

2-3 Supervised classification 1998 accuracy assessment............... ............... ............... 46

2-4 Rule-based classification 2003 accuracy assessment ....................................................... 46

2-5 Rule-based classification 1998 accuracy assessment ................................ ............... 47

2-6 Rule-based classification 1998 accuracy assessment ................................ ............... 47

2-7 Change trajectories of land use and land cover classes of interest (1993-1998-2003)......48

2-8 Supervised classification 1993 accuracy assessment .....................................................48

2-9 Supervised classification 1998 accuracy assessment .....................................................49

2-10 Supervised classification 2003 accuracy assessment .....................................................49

2-11 Change trajectories of land use and land cover classes of interest for supervised
classification im ages (1993-1998-2003)................................... ............................. ......... 49









LIST OF FIGURES


Figure page

2-1 M ap of study area ....................................................................... .......... ........ .... 50

2-2 Gainesville monthly rainfall from February 1992 to March 1993.................................51

2-3 Gainesville monthly rainfall from January 1997 to February 1998................................51

2-4 Gainesville monthly rainfall from February 2002 to March 2003................................52

2-5 Aerial photos and training samples in central part of study area .....................................53

2-6 Spectral and non-spectral data set analysis layers that created for rule-based
classification .............................................................................. 54

2-7 Sample set of forest extracted training samples with their associated values from all
d ata lay e rs ............................................................................. 5 4

2-8 Process of creating models that have acceptable accuracy in COMPUMINEN's
so ftw a re ................... .......................................................... ................ 5 5

2-9.Tree models with their number of rules and accuracy.......................................................56

2-10 Sample of tree model nodes and branches that used to create classification rules............57

2-11 Set of logic statements that transferred to Knowledge engineering classification tool.....58

2-12 Flowchart of knowledge based classification methodology. .........................................59

2-13 First supervised classification approach of remotely sensed image for Alachua
County, Florida in 1998. ................................................ .... ..................60

2-14 Changes Trajectories chart classes of interest (1993-1998-2003) .............. ................ 61

2-15 Knoweldge based classification changes trajectories map classes of interest (1993-
1998-2003). ...............................................................................62

2-16 Map of stable land cover classes from 1993 to 2003 ...................................................63

2-17 Map of land use and land cover changes from 1993 to 2003 ...........................................64

2-18 Sample of suburban area in Alachua County............................................ .................. 65

2-19 Changes Trajectories map classes of interest for supervised images (1993-1998-
2 003) .......................................................... ................................... 66









Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

USING REMOTE SENSING AND GIS TO STUDY LAND-USE AND LAND-COVER
CHANGE IN ALACHUA COUNTY, FLORIDA FROM 1993 TO 2003

By

Muhammad Almatar



August 2008

Chair: Jane Shouthworth
Major: Geography

This research investigates land use and land cover change in Alachua County, FL from

1993 to 2003. Determining the best classification method to classify satellite images for the

study area was the second objective of this research. Different classification methods were used

in the research in order to improve the classification accuracy. Supervised classification

(Gaussian Maximum Likelihood) was the first approach to classify three Landsat images. The

accuracy assessment of this classification was not acceptable for remote sensing research.

Knowledge based classification was used to improve the accuracy in all images. The accuracy

increased in all images, however there was misclassification in suburban class due to the mixed

texture of this area. A third classification was performed after removing suburban class, the

results of this classification were acceptable for academic research. The trends of changes in land

use and land cover are taking a place in west and central portions of Alachua County. This

research found that 4.97% of the total area transformed from vegetation to built areas by 2003,

1.07% of the total area transformed from vegetation land cover to built area by 1998, and 2.23 %

of the total area have converted from built structure to vegetation area in 1998.









CHAPTER 1
INTRODUCTION

Background

Among social sciences disciplines Geography is the only science that examines the

relations between society and the natural environment (Peet, 1998). Geography as known is the

study of the earth's surface and how natural force and living organisms reshape this surface.

Analyzing the environmental system from both physical and human perspectives, over time,

allows us to investigate how humans modify the earth's surface and how humans change their

surrounding environment to obtain their essential needs such as food and shelter. According to

Turner and Meyer (1994) the earth has experienced a great deal of change, including species

extinctions, and changes such as expansion of cities and loss of forests, although the world has

never experienced such rapid land changes as it is today. The interaction between humans and

their surrounding environment is resulting in changes on land cover patterns and these changes

could be monitored to recognize human actions that drive them to enable us to better understand,

model, manage and predict further environmental change (Turner and Meyer, 1994). One way to

monitor these changes on the earth's surface is to study land use and land cover changes of a

specific geographic area over a period time.

Land use and land cover change is the approach of monitoring changes on the earth's

surface that occurred as a function of the interaction between humans and their surrounding

environment. Land use refers to the use of the land that reflects human behavior. For example, a

forested land cover could include multiple land uses. It could be used as a resource for logging or

as a park for recreation purposes. Land cover change, may be related to changes in land use, or

could simply be a function of changes in the natural system e.g., wetland area changing to bare

land as a function of changes in the climate system (Turner and Meyer, 1994). Land use and land









cover change plays an important role in recent research in order to understand the environmental

and social consequences of change and to identifying the temporal and spatial extent of changes

on the earth's surface. Today, Remote Sensing and Geographic Information System (GIS) have

been used widely to study and analyze land use and land cover. This study examines the changes

that have occurred in land use and land cover in Alachua County from 1993 to 2003.

Study Area

Alachua County is located in the North Central part of Florida. Alachua County includes

number of cities Alachua, Archer, Hawthorne, High Springs, LaCrosse, Micanopy, Gainesville,

Newberry, and Waldo across 960 square miles (Phillips, 1986). Alachua County's climate differs

from other parts of Northern Florida, due to its unique latitudinal location, the variety of surface

features, and distance from the sea. Gainesville is this region's largest city and most populated,

and is home of the largest public University in the state of Florida (The University of Florida).

Use of Remote Sensing and GIS

According to Yang et al. (2002) land use and land cover change could be monitored by

using remotely sensed date to study spatial and temporal features of these changes. Remote

sensing is the science of gathering information about the earth's surface without touching it

(Digital Mapping Systems, 2006). Analyzing this information over spatial and temporal scales

could be an excellent tool to detect changes that occurred on the earth's surface in order to

provide a better understanding of how humans change their surrounding environment. According

to U.S. Geological Survey, GIS is a computer system and software that used to manipulate,

amass, organize and present geographically referenced data based on their location. It is basically

a database system that used to display geographically referenced data as map images.

Remote Sensing techniques have been used widely in research to classify satellite images

according to the spectral signature of the land covers. These land cover types will be presented as









categories and each category includes a group of pixels that have similar values. Changes in land

use and land cover that occurred in the earth's surface could be obtained from the classified

satellite images.

Our study used Remote Sensing and GIS to examine changes in land use and land cover in

Alachua County from 1993 to 2003. Three Landsat images were obtained from two different

satellites (Landsat TM and Landsat ETM), and used to monitor changes in land use and land

cover. These three satellites images classified by three different techniques, unsupervised

classification, supervised classification and knowledge based classification in order to improve

the classification of remotely sensed images and to detect the trend of land use and land cover

change for Alachua County, Florida from 1993 to 2003.

RQ1: What type of land use and land cover change patterns have occurred in

Alachua County from 1993 to 2003?

RQ2: What type of classification method produces the best results? Why?









CHAPTER 2
UTILIZING REMOTE SENSING AND GIS TO STUDY LAND-USE AND LAND-COVER
CHANGE IN ALACHUA COUNTY, FLORIDA FROM 1993 TO 2003.

Introduction

Background

Land transformations have been going on since the beginning of time. The earth has

experienced the spread and extinction of species, loss of forests, drainage of wetlands, expansion

of cropland and major developments of cities, but the world has never suffered such rapid land

transformations as it is today (Turner and Meyer, 1994). According to many scholars, recent

environmental and land changes are attributed to human activities. The human impact on the

earth has reached its peak, is unprecedented in its scale and is significantly influencing the

resources needed in sustaining the biosphere (Turner and Meyer, 1994). It is therefore important

to recognize these human actions and the forces that drive them to understand, model, manage

and predict further environmental change (Turner and Meyer, 1994). One way to accomplish this

is by studying the land use and land cover changes of a specific geographic area over a period

time.

Today, modern technologies in Remote Sensing and Geographic Information System (GIS)

have made it easier to study and analyze land use and land cover. This research examines the

changes in land use and land cover in Alachua County from 1993 to 2003. Alachua County is

located in the north-central part of the state of Florida. Gainesville is this region's most

populated and largest city, and is home of the largest public University in Florida: The

University of Florida.

One the of the project's objectives is to see what kind of land use and land cover changes

Alachua County experienced in the last decade. Urban planners can use this information to plan

for the future and potentially overcome any projected mishaps. More importantly, this project









aims to showcase the need in using GIS and remote sensing technologies to determine, analyze,

and predict land use and land cover change. The earth is changing ever so rapidly, and land use

and land cover research is essential to providing solutions to overcome such changes (Maktav et

al., 2005).

Land Use and Land Cover Change

Land use and land cover change is the approach of understanding changes on the earth's

surface that are reached by the interaction between humans and their environment. Land use

refers to the use of the land associated with human behavior. A forested land cover, for example,

can have multiple land uses. It could be used for recreation purposes, as a park or even a

plantation forest. Land cover, on the other hand, refers to changes in land cover, not related to

land use. For example, forest land changing to bare land (Turner and Meyer, 1994). Land cover

change can cause either conversion or modification. Conversion refers to changing from one

class to another, like from forest to cropland. Modification involves a change within a land cover

class, like a change in composition of a forest. However, while land cover changes usually

follow those of land use, the opposite is not always true: land cover change could occur even if

the land is unaltered by humans (Turner and Meyer, 1994).

According to Turner and Meyer (1994), land use and land cover are connected by the

"proximate sources" of change, which are the human actions that influence the physical

environment directly. These proximate sources reflect social, economical, political, and cultural

aspects in humans that shape and drive the changes in land use (Turner and Meyer, 1994).

Urbanization and Land Use/ Land Cover Change

The increase of urbanization globally is certainly one of the human driving forces that

directly affects land use and land cover. This significant increase in urban populations and urban

areas is continuously affecting natural and human environments at all geographic scales and









locations (Herold et al., 2003). The world's population is projected to increase from 6.1 billion in

2000 to 7.8 billion in 2025 with the urban share estimated to reach 58 percent. By 2025, the

majority of the world's population will be living in cities or urban areas (Gelbarad et al 1999).

For developing countries this trend can lead to problems like poverty, health issues-increase in

infant mortality and decrease in life expectancy-and natural disasters (Brockerhoff, 2000). For

the more developed world, which this study is more focused upon, challenges are arising less

from population growth and more from the changes in population composition and distribution

(Brockerhoff, 2000).

In the United States, unaffordable housing, inadequate educational opportunities and lack

of public safety in inner cities have all aided increasing poverty and homelessness rates. Also,

the increase of legal and illegal immigration in the United States could be viewed as a great

opportunity for diversity and cultural interaction, but ethnic immigrants living in urban cities

create ethnic clusters that usually cause intolerance and racial tensions between different

immigrant groups.

These same challenges, in addition to low government investments in the infrastructure of

inner cities have motivated the more affluent and middle-class population to relocate to the

suburbs. This movement of people and jobs away from the central cities toward suburbs of major

metropolitan areas is called decentralization, and it is considered as one of the most salient trends

of urbanization in the United States since the 1960s (Brockerhoff, 2000). The increasing growth

of suburbs, however, results in what is called urban sprawl, which is an outward growth of

subdivisions of land segregated by specific use, including low-density commercial and

residential settlements.









According to Brockerhoff (2000), the future of America's urban and spatial trends will be

determined by international flows of capital, information, labor, technology and decisions made

by individuals and government agencies. But the future needs to be beneficial for the future

generation, and urban planners need to use the resources and mechanisms available to analyze

past and current land use and land cover change, predict future implications, and provide

solutions to prevent those potential outcomes. In other words, effective schemes by urban

planners require up-to-date and complete information regarding past land use changes and their

impacts.

Certainly, the increasing interest in research on land use and land cover stems from a

growing concern over globally occurring environmental changes (Houghton, 1994). As

mentioned above, land use and land cover change is the approach of understanding changes on

the earth's surface that is reached by interpreting the interactions between humans and their

environment. These changes could affect environmental aspects like climate change, hydrologic

processes and extermination of plant and animal species. However, systematic analyses of the

change trends are needed to further comprehend and understand the drivers behind these

transformations (Turner el al., 1995).

Land use and land cover change plays an important role in choosing the correct research

approaches that will lead to identifying the temporal and spatial extent of changes on the earth's

surface. In order to understand the environmental and social consequences of change, researchers

have to observe and monitor between various patterns of change. To do so, researchers have to

utilize the technologies in GIS and remote sensing.

Use of Remote Sensing and GIS

Remotely sensed date can be used to study spatial and temporal features of land use and

land cover change (Yang et al. 2002). Remote sensing is the science of obtaining information









about the earth's surface, usually from aircraft or satellite, using instruments that may use visible

light, infra-red or radar to acquire data (Digital Mapping Systems, 2006). Use of such data over

spatial and temporal scales can detect changes occurring on the earth and can provide a better

understanding of how the land is being influenced by human activity. Having the ability to

observe and collect data for large areas within a short period of time, remote sensing can be a

useful tool for GIS.

According to U.S. Geological Survey, GIS is a computer system and software whose

purpose is to organize, amass, manipulate and present geographically referenced data based on

their location. It is basically a database system that manages spatially referenced data which can

be graphically displayed as map images.

There are several Remote Sensing and GIS techniques that have been employed in research

to classify satellite images based on the spectral signature of land covers. Image classification is

the process of grouping pixels that have similar values into classes or categories, which signify

certain types of land use and land cover. Classified satellite images provide information of land

use and land cover classes, which generally used to identify changes in the earth's surface.

Parametric techniques are commonly used to classify remotely sensed data that are distributed

normally by using map algebra, which clusters like-pixel values in the image statistically.

Maximum likelihood and unsupervised classification are widely used as parametric techniques.

For non-normally distributed remotely sensed data nonparametric techniques are implemented to

classify the imagery such as fuzzy and nearest-neighbor classification. One problem relating to

such optimal statistical clustering of data is that it may not correspond to very meaningful

information classes when using unsupervised methods. On the other hand, spectral signatures









used in supervised classification often overlap and make effective discrimination based solely on

spectral reflectance characteristics very difficult (Daniels, 2006).

The level of difficulty in achieving an accurate land cover classification relates to the

definition of the land cover classes used. For example, urban and suburban areas are not similar

to other land cover types, which are more easily classified such as water and forest, because of

it's heterogeneous nature. Urban and suburban as land use/cover classes usually include other

types of land cover in addition to a built signature, such as grass, water, soil, and trees besides

the built materials. The mixed spectral signatures of this land cover result in a need for advanced

classification techniques to differentiate between them. In this paper, GIS data sets, implemented

with remotely sensed data in a rule based classification technique, engage both spectral and non

spectral information to try to improve classification accuracy of remotely sensed images for

Alachua County, FL and detect the changes that occurred in land use and land cover across the

region, from 1993 to 2003. Study questions addressed are;

RQ 1: What type of land use and land cover change patterns have occurred in Alachua

County from 1993 to 2003?

RQ2: What type of classification method produces the best results? Why?

Methodology

Study Area

Located in the North Central part of Florida, Alachua County covers about 960 square

miles (Figure 2-1). Alachua, Archer, Hawthorne, High Springs, LaCrosse, Micanopy,

Gainesville, Newberry, and Waldo are all part of Alachua County (Phillips, 1986).

Due to its unique latitudinal location, distance from the sea and the variety of surface

features, Alachua County's climate differs from other parts of Northern Florida. The county has

one warm rainy season and a cool dry season. The rainy season begins in the middle of May and









ends in September. The average annual rainfall for Gainesville, which is the largest city in

Alachua, is 35 inches per year, with most of the rain falling between June and July

(Dohrenwend, 1978, City of Gainesville, 2008). The average high temperature in winter is

18.3C and the average temperature in the summer is 31.6C (City of Gainesville, 2008).

Monthly rainfall from 1992 to 2003 was provided by Gainesville Airport Weather Station

(Figure 2-1). Figure 2-2 shows the monthly rainfall for Gainesville, Alachua County from

February 1992 to March 1993 and the average rainfall for 1992-2003 fro comparison. The figure

indicates that the rainfall in February 1993, one month before obtaining the Landsat TM 1993

image, was above average by 1.75 inches. This difference in the rainfall made the study area

slightly wetter than normal. Figure 2-3 demonstrates the monthly precipitation for Gainesville

from January 1997 to February 1998 and the monthly rainfall from February 1992 to March

2003. In January 1998 the rainfall one month before the Landsat TM 1998 was obtained, was

more than the average by 8.56 inches. Since Alachua County experienced the highest amount of

rainfall in the beginning of 1998, the area of study was wetter than any time else across the study

period. Figure 2-4 shows the monthly precipitation for Gainesville from February 2002 to March

2003 and the average monthly rainfall from February 1992 to March 2003.The figure indicates

that the rainfall in February 2003, one month before obtaining the Landsat ETM 2003 image,

was below the average by 2.98 inches. The rainfall in February 2003 made the study area drier.

These differences in precipitation may affect the amount of area in the Water land cover class

which can be easily monitored, but should not impact the other classes, due to them being such

broad descriptions of land cover e.g., an area would not convert from any land cover class due to

precipitation difference given here, accept for a temporary increase or decrease in a water class,









due to flooding or drought, and really only 1998 might show significant changes. If there are

potential impacts, it will be discussed in the results section.

In the last decade, Alachua County experienced major population growth, mainly due to

the expansion of University of Florida and the Gainesville vicinity. According to the population

Census Counts, the population in Alachua County increased from 181, 600 in 1990 to 217,955 in

2000, an equivalent of about a 20 percent increase. In 2006, the population was estimated to be

227, 120 (U.S. Census Bureau). Also, the recent success of the University of Florida's basketball

and football teams-National Basketballs champions in 2005 and 2006, and the 2006 football

champions-has given the city of Gainesville media publicity and in essence, increased its

popularity. These athletic successes caused more high-school graduates to apply for the

University of Florida. The successes also brought in more money to the city, which has helped

in its growth.

With this growing population, the land use in Alachua County has certainly experienced

significant changes. Most of the land use change, as implied above, has occurred in Gainesville.

The growth of the University of Florida produced changes in Alachua county and Gainesville.

The downtown area in Gainesville has become a government and professional center, and retail

stores migrated west to Interstate 75. As a result of these changes, population growth has

traveled west, and government officials have trying to maintain a balance between the area's

natural environment and this growth. (Gainesvilletoday.com, 2008)

Because of these rapid changes and growth in population occurring in Alachua County, it

was inferred that studying and analyzing the land use and land cover change of this region would

provide valuable information for urban planners to potentially overcome any upcoming crisis and

to successfully plan for the future.









Image Pre-Processing

Two Landsat TM images were obtained for Path 17, Row 39 in March 1993, February

1998, and one Landsat ETM image in March 2003. Since the most recent Landsat ETM image

that was available was 2003 (pre Landsat 7 problems) this ten year period was determined as the

study period. The imagery was obtained during the dry season, when cloud free images are more

available. However, the March 1993 image had some small areas of clouds that were then

masked out of all images by using subset analysis. The masked areas are assigned as no data

within the study site. In order to reference the images sixty distributed ground control points

(GCP), where gathered with GPS receivers in region. UTM coordinate system (zone 17 N), NAD

83 datum was used to register all images and field checked for accuracy (Binford et al 2006).

The root mean square error (RMS) for all images ranged from 6.8 m to 8.9 m and the average

was 8.0 m. CIPEC atmospheric calibration techniques (Green et al., 2005) were performed on all

images to minimize the atmospheric variation such as solar elevation and curve, and atmospheric

effects.

Classification Scheme

To determine the classification scheme unsupervised classification was performed in 10,

20, and 30 classes for the study area. Based on the clustered classes of unsupervised

classification, five land use and land cover classes were designated and to increase the

classification accuracy some land cover classes grouped under their main types (Table 2-1). Land

use classes related to suburban, urban, pasture. Land cover classes were water and forest. This

proposed classification scheme was based on the main aim of the paper which is land use and

land cover changes in Alachua County from 1993 to 2003, much of which is urban and suburban

development, and loss of more natural vegetation in the process.









Field work was conducted for the study area to experiment with the efficiency of the

classification scheme. Water as land cover, consists of an area that is inundated by water such as

rivers, lakes, and wetland. Forest is area containing different species of trees and bushes with >

25% canopy closure. Pasture as land use, composed of agricultural land, pasture, grassland, and

others. Urban, composed of 70-100% built materials such as the airport and shopping malls.

Suburban, composed of 40-70% built materials which are mixed with vegetation such as

Residential.

Training Samples and Fieldwork

Training samples for the study area were collected with Trimble GPS from the field.

Trimble GPS is one of well known high accurate GPS, which has < Im spatial accuracy. Field

work was conducted from May to August 2007. Around 300 distributed training samples were

gathered from across the study area (Figure 2-5). These training samples were used with aerial

photos, from Florida Geographic Data Library, to determine land use land cover types. The

interpretation of aerial photos increased the number of training samples from 300 to over 1000.

These training samples were used to implement the classification and then to assess the

classification via an accuracy assessment. Moreover, around 800 photos were taken during the

field work for land use and land cover types in the study area.

Image Classification

In the first part of this research traditional classification methodology "supervised

classification using the Gaussian Maximum Likelihood" was performed on all image dates in

order to compare to decision based classification method. The average accuracy assessment

(overall accuracy) of these images was not more than 55%, which is unacceptable for remote

sensing classification. These classes are hard to differentiate from some spectrally similar land









cover classes. As result, it was essential to implement an advanced classification technique

which was rule-based in nature.

Building on this initial classification methodology a rule-based classification scheme was

then created for the study region. For this analysis additional data layers had to be created. A

number of spectral analysis layers were created in the ERDAS Imagine processing software for

use in the classification process. These are the initial classification; an Normalized Difference

Vegetation Index (NDVI), Tasseled Cap Transformation (TCT), Texture Analysis (TA), low

pass filter, and edge detector analysis. Moreover, a set of non-spectral analysis layers were

created in ArcGIS for use in the classification process. These are a Digital Elevation Model

(DEM) at 30 m spatial resolution which was obtained from Florida Geographic Data Library

(FGDL), Distance to home, Distance to roads, and distance to school (Figure 2-6). These

distances were measured by using ArcGIS Spatial Analyst tool; Euclidean Distance. This tool

measures the distance between each pattern and it surrounding pattern and it used to relate land

covers to this distance, for example, it measures the distance between all major roads in Alachua

County.

Normalized Difference Vegetation Index

The Normalized Difference Vegetation Index (NDVI) is a radiometric measure of remotely

sensed imagery that designates activity and relative abundance of healthy vegetation, which

includes percentage of greens cover, leaf area index, green biomass, chlorophyll content, and

absorbed photosynthetic active radiation (APAR) (Jenson 2000). NDVI has been widely

employed in remote sensing research to indicate vegetation classes from other land use and land

cover classes. NDVI analysis was performed based on the calculation of red and near infrared

(NIR) bands:

NDVI = NIR-RED/NIR+RED









NDVI calculation values vary from -1.0 to 1.0. The lower value signifies no vegetation,

while the higher value signifies green and healthy vegetation. In order to differentiate vegetation

classes from non-vegetation classes in this study, NDVI layers were created for each image date.

Texture analysis

Texture analysis techniques are frequently used to indicate the spatial information of

diverse object classes into a classification. Many researchers have revealed that structural and

spatial information can lead to momentous improvements in the accuracy assessment of built up

areas classification (Hsu 1978, Gong and Howarth 1990, Marceau et al 1990, and Puissant et al

2005). Texture analysis outputs may be used as a method to classify an image or as additional

data layer that could be used in the classification. Texture analysis incorporates information

based on the surrounding pixels, so it is involved with spectrally based data. This analysis

method has proven to be especially useful for urban and suburban area differentiation (Puissant

et al 2007).

Texture analysis has several optimal windows sizes such as 3 x 3, 5 x 5, and 13 x 13. The

optimal window size determines the amount of spatial information. If the window size is very

small, inadequate spatial information is extracted to characterize a specific land cover, while

large windows may overlap two types of land cover resulting in erroneous spatial information.

For that reason, image resolution and the content within the image are used to identify window

size for texture analysis (Puissant et al 2007). In this study, texture analysis was performed on all

images using 5 x 5 windows based on resolution and the content within the images.

Low pass filter

The low pass filter was used to attenuate noise and boundary errors, and to enhance

features in the remote sensing imagery, by producing new brightness values after blocking the

high spatial frequency detailed, this filter introduces useful spatial frequency information that









could be extracted from the image (Jenson 2000). Low pass filter has different size of windows,

such as 5 x 5 and 7 x 7. The smaller size windows are usually used to distinguish small features

from their surrounding area. On the other hand, large size windows are used to eliminate small

features in order to recognize large features. These windows are used to generate the average of

pixel values in the box for each pixel value in the image (USGS 2007). Yang and Liu (2005)

used a 3 x 3 filter to decrease the difficulty of the classification process. They performed 3 x 3

filter to reduce classes boundaries error due to the occurrences of intra-pixel spectral mixing and

signal noise. Since this study used remotely sensed data that were obtained from Landsat TM and

ETM with a spatial resolution of 30 x 30 m, a 5 x 5 filter was used in all images. Using a 5 x 5

filter improved the ability of detecting small features within the study area and to reduce noise

and boundary errors.

Tasseled cap transformation

Tasseled Cap Transformation (TCT) is a type of principle components analysis. It

transforms image bands into brightness (to indicate urban and soil), greenness (to indicate

vegetation), and moisture or wetness. According to Jensen (2000) the brightness band in

Tasseled Cap Transformation is used to identify urbanization areas which are particularly evident

in this band. The greenness band is an important source which provides information about

vegetation. Moisture status of the wetland information presents in the wetness band. TCT could

be helpful for use anywhere to disaggregate the quantity of soil brightness, vegetation, and

moisture content in independent pixels in satellite imagery (Jensen 2000). Since this study

attempted to detect changes that occurred in land use such as urban and land cover such as forest,

using this kind of analysis would be helpful to improve the classification accuracy. Tasseled Cap

Transformation analysis was performed on all images for use in suburban area discrimination.









Thermal band

The thermal band was extracted from all images for use in the classification process in

order to improve the accuracy. According to Jensen (2000), this band calculates the quantity of

infrared energy that is released from the earth's surface and it is practical for locating geothermal

activity. Using the thermal band as layer data could be helpful to improve classification accuracy

by indicating the temperature of each land use and land cover classes. Lu and Weng (2006) in

their study of impervious surface in urban land-use classification in Indianapolis/Marion County,

Indiana, USA, used thermal band as input layer to classify urban land use classes, in order to

decrease the number of mixed pixels. Built structures have many different thermal properties

than vegetation area, so using thermal band can be used to discriminate across the classes. As a

result, including the thermal band in the classification as a data layer may help to differentiate

between suburban and vegetation areas.

Non-spectral data layers

Other, non-spectral data layers were also incorporated. A Digital Elevation Model (DEM)

at 30 m spatial resolution was obtained from Florida Geographic Data Library (FGDL) and used

as a data layer for the rule-based classification process. Using a DEM could be useful to identify

the association between land use and land cover classes with the elevation of the landscape. This

data layer could be used to differentiate certain land use types from others. Moreover, distance to

home, distance to roads, and distance to school data layers were also created for the rule-based

classification. The input data for their creation (homes, public schools, and roads) were obtained

from the Department of Growth Management in Alachua County as GIS layers.

According to the Department of Growth Management in Alachua County (2008), homes

layer is a dataset that contains locations and information pertaining of homes, public school layer

contain all of the Alachua county public school, and roads layer is contain all of the Alachua









county roads. Distance to home layer was created from a GIS raster layer which highlights

homes in Alachua County, by using ArcGIS Spatial Analyst tool; Euclidean Distance. This tool

measures distance from the center of a source cell to the center of surrounding cells, the outputs

of this tool is a raster of floating-point distance values (ArcGIS Desktop help). In other words,

this tool measure the distance between each pattern from the surrounding one. For example, it

measures the distance between Gainesville high school and Waldo high school, this distance

could be useful to indicate urban or suburban classes that exits in this range. Using this tool for

homes, major roads, and school could be practical to indicate the influence of these land use

patterns within the study region (Southworth et al. 2002), in order to improve the classification

accuracy.

Rule-based classifier

The study employed GIS and data mining software to generate classification rules that will

be used for knowledge based classification. ArcMap and COMPUMINE's Rule Discovery

Systems (RDS) were used to create sets of rules to classify each land use and land cover class

based on Boolean logic (if/then). All spectral and non-spectral data layers that were created in

ERDAS Imagine and ArcGIS were imported into ArcMap software. Training samples for each

land use and land cover class were also imported into ArcMap as GIS layers. The data layers,

which included spectral and non-spectral information, were associated with training samples as

corresponding values for all variables. To do so, the extraction tool in ArcMap Spatial analyst

tool box was used to parse out each training sample with their associated values from all data

layers. Figure 2-7 shows a sample of the extracted training samples with their values. This table

was imported into COMPUMINEN's software to determine the appropriate model that will be

used to create a set of rules for rule-based classification.









COMPUMINEN's Rule Discovery System (RDS) is a data-mining software that

classifies information based on RDS criteria. This software is employed to create a tree model

that will be used in rule creation. Using the n-fold cross validation method, ten tree models were

generated for all spectral and non-spectral data layers. Each model built a set of rules from a

number of variables and each model of these ten tree models has an assigned accuracy. This

process was performed many times in order to obtain the highest degree of accuracy. Figure 2-8

shows the process of creating models that have acceptable accuracy in COMPUMINEN's

software.

Figure 2-9 shows ten tree models with their rules number and accuracy, based on accuracy

one of them will be used to create the classification rules. Accuracy of model number nine was

the highest, 81.905(%). Since improving final classification accuracy assessment is one of the

study aims, model number nine was used to create the classification rules. This model has fifteen

rules that were developed from a set of spectral variables; The Normalized Difference

Vegetation Index (NDVI), Tasseled Cap Transformation (TCT), Texture Analysis (TA), and the

low pass filter.

The results of model number nine were presented as tree diagrams which determined the

rules that will be used for rule-based classification. These diagrams included nodes and branches

for each class (Figure 2-10). Each node introduces detail about one particular class with their

values. For example, water node indicates that any pixel = 3.5 in the tasseled cap (band 1), 2 in

the supervised images, <= 9.2566665 in the texture (band 1), and <= 43.5 in the low pass (band

3) will classify as water. Translating model node and branch into rules, which will be

transformed to language for ERDAS Imagine, was performed for all modes and branches. Table

2-2 shows fifteen rules that will be used in the decision based classification.









Multiple rules were created for different land use land cover classes from the optimal tree

model. These rules were transferred to the Knowledge engineering classification tool in ERDAS

Imagine processing software. Transferring rules to the Knowledge engineering classification tool

is an essential step to translate final rules into hypotheses with their value to classify all images.

Figure 2-11 shows a set of logical statements that were used to transfer the fifteen rules to

Knowledge Engineering Classification tool as hypotheses. After transferring all rules to ERDAS

Imagine format the rule-based classification was performed on all images.

Change Detection Analysis

Change detection techniques have been used widely in land use and land cover changes

research. Change detection analysis is a procedure of detecting changes that occur in particular

areas during a period of time. In order to determine the changes in land use and land cover that

occurred in Alachua County from 1993 to 2003, change trajectories were created. This technique

requires the acquisition of remotely sensed images for the same area over two or more time

periods (Southworth et al 2002). In this study three images were classified by performing

decision based classification, and then incorporated into a change analysis. Based on these three

images, change trajectory was implemented to indicate changes occurring across each image

date, by incorporating these changes into a single change image. With three images and five land

use and land cover classes, 125 possible change trajectories classes are feasible, and

interpretation can get confusing (Southworth et al 2002).

These all trajectories were not realistically occurring in the study area, so expert opinion

was also used to find the likely and acceptable options. Since the population growth and the

cities expansion in Alachua County increased last ten years, the land cover conversion is

generally taking a shape of land transformation from vegetation toward urban and suburban

areas. The 125 possible change trajectories were grouped to minimize confusion because









showing all trajectories visually is very difficult especially many of them are spatially limited

and meaningless. For example, urban does not convert to water in area that demands more

urbanization area. To eliminate such confusion some of change trajectories grouped to highlights

the dominant and meaningful conversion.

Results

Accuracy Assessment

Accuracy assessment is considered one of most important elements of the classification

process because without it the classification would be worthless. The Kappa coefficient

assessment method was performed for all images. This method is defined by Lu et al (2005) as a

calculation of overall statistical agreement of a matrix. Aerial photos were take on December

1994 were used to maximize the number of point for the accuracy assessment in order to

distribute the training samples trough the study area within ten years period. Training samples

collected from fieldwork and aerial photo interpretation were randomly divided into two sets.

The first set contained 70% of total training samples, which were used to create the initial

classifications. The other set contained 30% of total training samples, which were held back and

then used for accuracy assessment and validation.

The initial, traditional supervised classification which was performed for the study area,

resulted in unacceptable accuracy levels (Table 2-3 and Figure 2-13) and so knowledge based

classification was used to attempt to improve upon these initial, inferior results. Performing

knowledge based classification increased the accuracy average from 50% to 80 % (Tables 2-4, 2-

5, and 2-6). The accuracy assessment measures that were used in this study included overall

accuracy, user's accuracy, producer's accuracy, and the KAPPA statistic.

In the 2003 image the overall Kappa statistic was 0.7898 and overall classification

accuracy was 85.18%, which means the overall accuracy improved by 35%. In the 1998 image









the overall Kappa statistic was 0.8735 and overall classification accuracy was 90.47%, which

means the overall accuracy improved by 40%. In the 1993 image the overall Kappa statistic was

0.6903 and overall classification accuracy was 76.12%, which means the overall accuracy

improved by 25%. The study established that knowledge based classification improved the

accuracy of the classification significantly.

Change Trajectories

Since indicating the expansion of urban and suburban areas was one of main goals of the

study, pasture and forest were combined to vegetation. This produced a classification with four

land use and land cover classes; water, vegetation, urban, and suburban. In change trajectories

detection for four land use land cover classes (1993-1998-2003), 64 possible trajectories may

take place as land use and land cover changes. This study used change trajectories detection to

identify land use and land cover changes that transformed from one to another type. In order to

minimize trajectories for the study area, only changes that present more than one percent of the

study area will be discussed (Table 2-7, Figures 2-14 and 2-15). Water land cover was

experienced some potential changes due to the increase of the monthly precipitation in 1998 and

decrease in 2003. These changes in water land cover were less than one percent of the study area,

as a result, these changes will not be discussed.

Change trajectory tables indicate 38.27% of the study area occupied by vegetation area

(Veg-Veg-Veg) across the three time periods (1993-1998 -2003). While, suburban area (Sub-

Sub-Sub) that remains without changes since 1993 to 2003 occupied 23.53% of the study area,

urban area occupied just 1.19% within same periods of time. According to table 7, 6.06% of the

study area was covered by water class (Water-Water-Water) from 1993 to 2003 (Figures 2-16).

In 1998, 2.29% of the total area was transformed from vegetation (Veg-Veg-Urban) to

urban area. The trajectory determined that 1.89% of the total area was transformed from









vegetation (Veg-Sub-Sub) to suburban in 1993. Meanwhile, 1.69% of the total area transformed

from urban (Urban-Urban-Sub) to suburban in 1998, 4.77% of the total area transformed from

suburban (Sub-Sub-Urban) to urban area in 1998. 1.05% of the study area occupied by

vegetation area was transformed to suburban area in 1998 (Veg-Sub-Veg) then went back to

vegetation in 2003. These changes in land use and land cover in Alachua County (Figure 2-17)

indicates the area of land use and land cover changes across the study area from 1993 to 2003.

Even though knowledge based classification improved the accuracy assessment in all

images, 17.51% of the total area transformed from vegetation to suburban in 2003 (Veg-Veg-

Sub) and 1.05% from suburban to vegetation (Veg-Sub-Veg) in 1998, which lead to

misclassification between vegetation and suburban. While there was surely an increase in this

suburban class related to the process of development, 17% seems excessive. Han (2002)

demonstrated in his research "an examination of historic land use changes in Alachua County,

Florida: a technological approach" that there are decrease in overall residential and increase in

government property from 1992 to 2003, so from these results the high amount of vegetation

converted for suburban are not acceptable for the study area. In addition, the confusion between

these classes relates to the fact that many suburban developments in Gainesville, do indeed have

a large amount of vegetative cover (Figure 2-18), so the spectral signatures of these classes are

quite close. A vegetated area with some bare soil, may be very similar to heavily vegetated

suburban area with some built structures (built and soil are both low NDVI and greenness, high

in reflectance and brightness, and so similar in these respects). To eliminate this potential

misclassification a third classification was undertaken, a supervised classification with only;

water, vegetation, and built as classes. Built includes urban and suburban structures (buildings,

roads, and paving), which are not mixed with vegetation, so this can be distinguished from this









confusion class of suburban here in Gainesville that is we break out a suburban class into built

verses vegetated, when possible Note that this problem is not just particular to Gainesville, but

it is especially problematic here, due to the almost complete canopy closure in many older

residential neighborhoods making them appear as forest from above ( although there fully

vegetated, full canopy suburban developments, will still appear an vegetated as the satellite

"sees" the trees).

New supervised classifications were performed on all images for these new classes; water,

vegetation, and built areas. Water as land cover, consists of an area that is inundated by water

such as rivers, lakes, and wetland. Vegetation is area containing different species of trees and

bushes with > 25% canopy closure. Built up class includes urban and suburban types, which are

not mixed with vegetation. Accuracy assessment was performed for the new supervised

classification images (Table 2-8, 2-9, and 2-10). Overall Classification Accuracy was 98.24%

and Overall Kappa Statistics was 0.9689 in 1993 image. In 1998 the Overall Classification

Accuracy was 98.57%, while Overall Kappa Statistics was 0.972. Table 2-10 shows that Overall

Classification Accuracy was 98.24% and Overall Kappa Statistics was 0.9689 in 2003 image.

Change trajectories were created for the new simplified supervised classification images.

Since these supervised classification images (1993-1998-2003) are for three land cover only 27

change trajectories are possible as land use and land cover changes. In order to minimize

trajectories for the study area, only changes that present more than one percent of the study area

will be discussed (Table 2-11 and Figure 2-19). The change trajectory indicates 83.64% of the

study area occupied by vegetation area (Veg-Veg-Veg) across the three time periods (1993-

1998-2003), although we know this includes many suburban areas now classified under this

option and with > 25% canopy closure. This is unavoidable however due to their appearing as









forest from above. A GIS layer of suburban developments would really be the only option to

truly differentiate there areas as this research has clearly shown they are not differentiable

spectrally, and this is important in highlighting the need for adaptable land use and land cover

classes, as illustrated in this research From 1993 to 2003 built up area (Built-Built-Built) is

occupied 1.95% of the total study area. According to table 11, 5.54% of the study is area covered

by water class (Water-Water-Water) from 1993 to 2003. In 1998, 4.97% of the total area was

transformed from vegetation (Veg-Veg-Built) to urban area. The trajectory determined that

1.07% of the total area was transformed from vegetation (Veg-Built-Built) to built between 1993

and 1998.

This new classification gives a much better accounting of the changes in built and urban

areas, but we can no longer differentiate between vegetation and most of the suburban classes

due to the canopy closure issues. In this research multiple methods were utilized in order to

understand the dominate changes in land use and land cover. A single method did not suffice.

This has obvious implications for this region, and multiple methods must be used to fully

evaluate the changes in land cover and land use here in Alachua County, as well as in many other

suburban regions across the U.S.

Discussion

Summary of Finding

The objectives of this study were to detect changes in land use and land cover in Alachua

County from 1993 to 2003 and to find the best classification method for the study area. Different

classification methods were used (supervised and knowledge based classification) to determine

changes in land use and land cover that occurred in the study area. The results of these two

classification proved that Alachua County is facing land transformation in terms of cover. This

transformation tends to relate to the increase of population across the study area and particularly









in Gainesville city limits which is the most populated area in the county. Gainesville experienced

major population growth in Alachua County due to the expansion of the University of Florida,

Shands and the popularity of Florida as a location.

The results of the classifications indicate that 4.97% of the total area transformed from

vegetation to built areas by 2003. Moreover, 1.07% of the total area transformed from vegetation

land cover to built area by 1998. This transformation is mostly located in the central and western

portions of Alachua County, which supports trend of population growth requiring more

urbanization. Such changes in land use and land cover relate to the population increase, which

stimulates land owners to transform vegetation areas to built areas often via suburban

developments. In Gainesville, much of this expansion has occurred in the west of the region as

Gainesville has expanded out a long Archer Rd, Newberry Rd, and 39th street. Much of the new

developments include cleaning of previously vegetated land and the building of suburban

developments. These new non-vegetated development may will change when homeowners plant

trees and eventually such neighborhoods may be similar to the older suburban developments in

Gainesville, which appear from above as forest.

Implication of Study

Determining land use and land cover change in Alachua County from 1993 to 2003 for

four classes; water, vegetation, urban, and suburban, was the main objective of this study. As a

result, two different classification methods were used to determine optimum results for the study

area. Supervised classification was the first method that was performed for the study area

remotely sensed images. The results of this method were not accurate enough for remote sensing

research. In order to increase the accuracy of the classification, knowledge based classification

was performed for all images. The results of the classification were accurate when compared

with supervised classification outputs. Even though, knowledge based classification improved









the accuracy assessment in all images, 17.51% of the total area transformed from vegetation to

suburban in 2003, which led to misclassification between vegetation and suburban. While there

was surely an increase in this suburban class related to the process of development, 17% seems

excessive. This results led to the fact that suburban class is hard to differentiate from vegetation

class in area like Alachua County because the confusion between these classes relates to the fact

that many suburban developments, do indeed have a large amount of vegetative cover. Even

using advanced RS classification techniques suburban would not be fully distinguished, so an

additional classification of water, built, vegetation was created to give better understanding of

overall change trends and to remove the suburban class, which proved to be combination of built

(urban) and vegetation, so created problems of spectral seperbility across these classes (urban,

suburban, vegetation). Hence, in order to fully understand this region a simplified classification

of land use land cover classes (built, vegetation, and water) had to be created with more detailed

land cover classification scheme obtained via decision tree classifiers. Only when we used both

classification results together could we fully comprehend and interpret the changes in land use

and land cover classes in Alachua County.

Changes in land cover in Alachua County may relate to the county government policies.

The city of Gainesville government properties such as services building increased especially in

the downtown area in order to provide more public services (Hang, 2002). The government

policies demand more urbanization areas which results more changes in land use and land cover

areas. Transportation played a role in the trends of land use and land cover in the study area.

Zhou and Kockelman (2006) in their study of neighborhood impacts on land use change in the

City of Austin, Texas, found that the use of different models proved that there was a relation

between land use changes and the distance to the nearest highway. In Alachua County the









commercial property uses increased in the western part and it extended along US 441. US 441

magnetized commercial activities which makes it act now as a commercial corridor (Hang,

2002).

Supervised classification techniques have been widely used in remote sensing research to

determine changes in land cover classes. This traditional technique would be a powerful

classification technique for classes that have unique spectral signatures, in other words, for

classes that could be easily discriminated, e.g., water from urban. Some scholars have

incorporate supervised classification with different analysis techniques to study land use and land

cover change. Southworth (2004) used supervised classification in the Yucatan, Mexico, to

successfully differentiate forest growth based on the earth's surface temperature, from Landsat

band 6 data. In areas like Alachua County, it is hard to differentiate between suburban and

vegetation due to the similarity of both spectral signatures, which makes using supervised

classification to study three classes reasonable. Due to the difficulties of classifying suburban

and urban areas, several researchers have created new techniques to eliminate the

misclassification of mixed pixels in remotely sensed images. Zhang and Foody (1998) used a

fuzzy classification approach to evaluate the classification for suburban land cover, in the city of

Edinburgh, UK. They used two remotely sensed images (Landsat TM and SPOT HRV) for an

area of 2 km2, located within the city of Edinburgh. The study area contained suburban structures

such as residential and commercial and it's rich with geographical diversity and appropriate for

fuzzy classification. They found that the Kappa coefficient (an indication of classification

accuracy) more than doubled when they applied the fuzzy technique (i.e., this allows a pixels to

belong to more than one classification, so suburban can be a member of both built and vegetated

classes). Lu and Weng (2006) in their study of impervious surface in urban land-use









classification in Indianapolis/Marion County, Indiana, USA, used land surface temperature data

from Landsat ETM+ and population density to improve the classification of five urban land use

classes. They extracted land surface temperature to get better impervious surface mapping, then

combined this information with population density to improve the classification of their land use

classes (Lu and wand, 2006). The study found that the integration of surface temperature

provided substantially improved impervious surface coverages, which was reflected in the

classification accuracy results. Mesev (1998) in a study of census data in urban image

classification, in the United Kingdom used population sensed data for supervised classification

(linking of urban class from remotely sensed data with urban functional characteristics from

population sensed) in order to improve the classification accuracy of urban classification. While,

Pellizzeri (2002) used Landsat TM and RADARSAT images to decrease the misclassification for

suburban area in their study in Northern Italy. Gluch (2002) merged TM and SPOT-P data to

monitor urban growth of Salt Lake City in Utah. Other classification methods have implemented

in many researches to eliminate the misclassification in urban and suburban areas. This validates

the combined methodologies used in this analysis as appropriate, over traditional classification

techniques, which many researchers have found equally problematic based on the surface types

found in urban and suburban regions (Gluch 2002, Lu and wand 2006, Mesev 1998, Pellizzeri

2002, Southworth 2004, Zhang and Foody, 1998).

Knowledge based classification techniques have been used recently to determine land use

and land cover changes especially for urban regions (Chien and Chou 2000, Haild 1997, Onsi

2003). This technique classifies remotely sensed data based on spectral and non-spectral

information. Some land cover classes have similar spectral signatures; this technique could

differentiate these classes from each other. Comparing knowledge based classification with









traditional classification techniques, knowledge based techniques have given better results than

other techniques. Daniels (2006) has proven that knowledge based techniques improved the

accuracy of classification for tropical land covers relating to the results of supervised

classification. Moreover, knowledge based techniques implemented in several studies have

addressed specifically changes that have occurred in urban areas (Chien and Chou, 2000).

Janssen (1992) used knowledge based classification techniques to improve crop classification

accuracy. This technique was also used to optimize the C-factor mapping in Spain by Folly et al,

(1996). Haild (1997) use knowledge based approaches to link local knowledge, field data and the

spectral land cover classes to indicate changes in urban land use classes.

Knowledge based classification techniques performed in this study greatly improved the

classification accuracy. Spectral and non-spectral data were imported into a data-mining software

in order to assign the best rules for knowledge based classification. Based on the output model of

this software fourteen spectral rules were used for knowledge based classification. The accuracy

assessments of the knowledge based classification were acceptable for remote sensing research;

however, it didn't eliminate the confusion between suburban and vegetation. As a result,

knowledge based classification techniques could not solve the spectral confusions between land

use and land cover classes. This study demonstrates that the spectral signature of a 'suburban

class' (a combination of roads, building and vegetation) in Alachua County and elsewhere, is too

close to the vegetation and urban class signatures (it is in fact a merged class of these two

components), even when using advanced knowledge based classification techniques. For future

research more fieldwork to gather explicit GIS layers of suburban boundaries would help clarify

the classification process and would help differentiate suburban from vegetation and urban.

Using finer spatial resolution satellite images would also be helpful for better classification, as









this would break down the pixels from mixed covers, to individual components, e.g. Quickbird,

0.4 m spatial resolution panchromatic data, although this is very expensive.

In summary, this study found that 4.97 % of total area of Alachua County converted from

vegetation to built area in 2003. These changes in land use and land cover have taken place

mostly in central and western portion of Alachua County. In 1998 1.07 % of the total area

converted from vegetation to built area. Generally, the changes located in the central and western

part of the study area. Gainesville which is the largest city in Alachua County has the most

changes that occurred in the county. Newberry, which is located in the west of Gainesville, is the

second city that has changes after Gainesville. Changes in these cities are reflecting the increase

in population and the expansion of University of Florida in Gainesville. Transportation affects

the land cover conversion especially around US 441 highway.

These results could be used by the county planner as reference for future estimation.

However suburban misclassification could be controlled by developing expertise and

methodologies in order to monitor changes in suburban area in Alachua County. Knowledge

based classification have been proven in several research studies (Chien and Chou 2000, Daniels

2006, Janssen 1992, Haild 1997, Onsi 2003 ) as an advanced technique for land use and land

cover changes, although it was not fully successful in this study, which reveals that no single

method is appropriate for all cover types. This study demonstrated changes in land use and land

cover in Alachua County and these changes have to be monitored for better planning. Suburban

area in Alachua County, like elsewhere, needs advanced analysis techniques and field work in

order to eliminate the misclassification between suburban and vegetation areas.

Conclusion

In conclusion, this research found that most of the changes in land use and land cover in

Alachua County have taking a place in western and central portions of the County. Land use and









land cover change in Alachua County have taken two major trends, from vegetation to built

structures area and from built to vegetation area. In 1998, 1.07 % of the total area have

transformed from vegetation to built structures area. In 2003, the number of the total area that

have transformed from vegetation to built area was 4.97 %. In 1998, 2.23 % of the total area

transformed from built structures area to vegetation. Most of these changes in land use and land

cover occurred in the city of Gainesville area and in the western portion of Alachua County.

A traditional classification (supervised classification) and an advanced classification

(Knowledge based classification) methods were used in this research in order to determine

changes in land use and land cover in Alachua County. The study area likewise other areas that

have a large amount of vegetation cover in suburban development, which is causing enormous

confusion between suburban and vegetation. As a result, Knowledge based classification method

could not solve the spectral confusions between land use and land cover classes (suburban and

vegetation), even though the accuracy assessments of the knowledge based classification were

acceptable for remote sensing research. Better methods need to be developed to monitor

suburban developments over time in Alachua County such as using GIS layers for suburban

developments could help to improve the ability of differentiating suburban from vegetation.









Table 2-1. Land use and land cover descriptions for classification scheme.
Name Class Type Description
Water Land cover Lands inundated with water such as lakes, rivers, and
wetland.

Forest Land cover Lands cover with different species of trees and bushes
with > 25% canopy closure.

Pasture Land use Pasture composed of agricultural land, pasture,
grassland, and others.

Urban Land use Urban composed of 70-100% built materials such as
airport and shopping mall.

Suburban Land use Suburban composed of 40-70% built materials which
are mixed with vegetation such as Residential area.









Table 2-2. Final decision based rules used in the final classification procedure.
Class type Rule criteria Values Description


Water
Tasseled Cap 3
Rule. 1 Supervised
Texture 1
Low Pass 3
Forest
Band 6
Rule.2.A Supervised
Texture 1

Band 6
Rule.2.B Band 6
Supervised
Texture 1

Band 6
Rule.2.C Supervised
Texture 1

Tasseled Capl
Rule.2.D Texture 1
Low Pass 4
Pasture
Supervised
Rule.3.A Texture 1
Low pass 3
Tasseled Capl
Supervised
Rule.3.B Texture 1
Low pass 3


=3.5
2
<= 9.2566665
<= 43.5

<= 107.5
=2
<= 9.2566665

> 107.5
<= 108.5
=2
<= 9.2566665

> 108.5
=2
<= 9.2566665

<= 114.5
> 9.2566665
<= 17.5

2
<= 9.2566665
> 43.5
> 114.5
S4
> 9.2566665
> 50


Classify as Water all water pixels that = 3.5 in the tasseled cap (band 1), 2 in
the supervised images, <= 9.2566665 in the texture (band 1), and <= 43.5 in the
low pass (band 3).


Classify as Forest all forest pixels that <= 107.5 in the thermal band, = 2 in the
supervised images, and <= 9.2566665 in the texture (band 1).


Classify as Forest all forest pixels that > 107.5 <= 108.5 in the thermal band, =
2 in the supervised images, and <= 9.2566665 in the texture (band 1).


Classify as Forest all forest pixels that >108.5 in the thermal band,
supervised images, and <= 9.2566665 in the texture (band 1).


2 in the


Classify as Forest all forest pixels that <= 114.5 in the tasseled cap (band 1),
and > 9.2566665 in the texture (band 1), and <=17.5 in the low pass (band 4).


Classify as Pasture all pasture pixels that 2 in the supervised images, <=
9.2566665 in the texture (band 1), and > 43.5 in the low pass (band 3).

Classify as Pasture all pasture pixels that >114.5 in the tasseled cap (band 1),
4 in the supervised images, > 9.2566665 in the texture (band 1), and > 50 in the
low pass (band 3).









Table 2-2 Continued.
Class type Rule criteria Values Description


Tasseled Cap3
Rule.3.C Supervised
Texture 1
Low pass 3
Urban
Tasseled Capl
Rule.4.A Supervised
NDVI
Texture 1
Tasseled Capl
Rule. 4.B Supervised
NDVI
Texture 1


S Rule .4.C Tasseled Capl
Tasseled Capl
Supervised
NDVI
Texture 1
Suburban
Tasseled Cap2
Rule.5.A Tasseled Capl
NDVI
Texture 1
Low pass 4
Tasseled Cap2
Rule.5.B Tasseled Capl
Texture 1
Low pass 4


<= 3.5
/ 2
<= 9.2566665
<= 43.5


=4
<=


Classify as Pasture all forest pixels that <= 3.5 in the tasseled cap (band 3), 2
in the supervised images, <= 9.2566665 in the texture (band 1), and > 43.5 in
the low pass (band 3).


> 114.5 Classify as Urban all urban pixels that >114.5 in the tasseled cap (band 1), = 4
in the supervised images, <= 0.1286489 in the NDVI image, and > 9.2566665
0.1286489 in the texture (band 1).


> 9.2566665
> 136
=4
> 0.1286489
> 9.2566665

> 114.5
< 136
=4
> 0.1286489
> 9.2566665

> -24
<= 114.5
<= 0.3839777
< 9.2566665
> 17.5
<= -24
<= 114.5
> 9.2566665
> 17.5


Classify as Urban all urban pixels that >136 in the tasseled cap (band 1), = 4 in
the supervised images, > 0.1286489 in the NDVI image, and > 9.2566665 in the
texture (band 1).


Classify as Urban all urban pixels that >145 in the tasseled cap (band 1), <136
in the tasseled cap (band 1), = 4 in the supervised images, > 0.1286489 in the
NDVI image, and > 9.2566665 in the texture (band 1).



Classify as Suburban, all suburban mixed with vegetation pixels that >-24 in the
tasseled cap (band 2), < =114.5 in the tasseled cap (band 1), <= 0.3839777 in
the NDVI image, < 9.2566665 in the texture (band 1), and > 17.5in the low pass
(band 4).

Classify as Suburban, all suburban mixed with vegetation pixels that <= -24 in
the tasseled cap (band 2), < =114.5 in the tasseled cap (band 1), > 9.2566665 in
the texture (band 1), and > 17.5in the low pass (band 4).









Table 2-2 Continued.
Class type Rule criteria Values Description
Tasseled Capl > 114.5 Classify as Suburban, all suburban mixed with vegetation pixels that > 114.5 in
Rule.5.C Supervised / 4 the tasseled cap (band 1), J 4 in the supervised images, > 9.2566665 in the
Texture 1 > 9.2566665 texture (band 1), and <= 50 in the low pass (band 3).
Low pass 3 <= 50
Tasseled Cap2 > -24 Classify as Sub-urban, all suburban mixed with vegetation pixels that >-24 in
Rule.5.D Tasseled Capl <= 114.5 the tasseled cap (band 2), < =114.5 in the tasseled cap (band 1), > 0.3839777 in
NDVI > 0.3839777 the NDVI image, > 9.2566665 in the texture (band 1), and > 17.5in the low pass
Texture 1 > 9.2566665 (band 4).
Low pass 4 > 17.5









Table 2-3. Supervised classification 1998 accuracy assessment; overall accuracy, user's
accuracy, producer's accuracy, and KAPPA statistic. Ref. = Reference, Suburb=
Suburban Pro.acc = Producers accuracy, Clas.Tot = Classified total, and Users.Acc=
Users accuracy.
Supervised classification 1998
1998 Water Forest Pasture Urban Suburban Clas.Tot Users.Acc
Water 24 0 0 0 0 24 1.00
Forest 1 113 1 14 32 161 0.70
Pasture 1 2 29 6 18 56 0.518
Urban 0 1 0 1 4 6 0.167
Suburb 0 2 28 76 62 168 0.369
Ref.total. 27 120 60 101 143 451
Pro.acc 0.888 0.941 0.483 0.99 0.433
Overall Kappa Statistics = 0.3559, Overall Classification Accuracy = 50.78%


Table 2-4. Rule-based classification 2003 accuracy assessment; overall accuracy, user's
accuracy, producer's accuracy, and KAPPA statistic. Ref = Reference, Suburb=
Suburban Pro.acc = Producers accuracy, Clas.Tot = Classified total, and Users.Acc=
Users accuracy.
Rule-based classification 2003
2003 Water Forest Pasture Urban Suburban Clas.Tot Users.Acc
Water 17 0 0 0 0 17 1.00
Forest 0 96 0 0 0 96 1.00
Pasture 0 9 22 0 0 31 0.709
Urban 0 0 1 85 14 100 0.850
Suburb 7 11 4 21 165 208 0.793
Ref.total. 24 116 27 106 179 452
Pro.acc 0.701 0.828 0.815 0.802 0.921
Overall Kappa Statistics = 0.7898, Overall Classification Accuracy = 85.18%









Table 2-5. Rule-based classification 1998 accuracy assessment; overall accuracy, user's
accuracy, producer's accuracy, and KAPPA statistic. Ref. = Reference, Suburb=
Suburban Pro.acc = Producers accuracy, Clas.Tot = Classified total, and Users.Acc=
Users accuracy.
Rule-based classification 1998
1998 Water Forest Pasture Urban Suburban Clas.Tot Users.Acc
Water 25 0 3 0 0 28 1.00
Forest 1 117 0 0 3 121 0.955
Pasture 0 3 49 0 1 53 0.754
Urban 0 0 0 85 6 91 0.601
Suburb 1 0 8 16 132 157 0.836
Ref.total. 27 120 60 101 143 451
Pro.acc 0.926 0.975 0.817 0.842 0.923
Overall Kappa Statistics = 0.8735, Overall Classification Accuracy = 90.47%


Table 2-6. Rule-based classification 1998 accuracy assessment; overall accuracy, user's
accuracy, producer's accuracy, and KAPPA statistic. Ref = Reference, Suburb=
Suburban Pro.acc = Producers accuracy, Clas.Tot = Classified total, and Users.Acc=
Users accuracy.
Rule-based classification 1993
1993 Water Forest Pasture Urban Suburban Clas.Tot Users.Acc
Water 22 0 0 0 0 22 1.00
Forest 1 84 1 0 2 88 0.955
Pasture 0 1 46 3 11 61 0.754
Urban 0 0 7 98 58 163 0.601
Suburb 3 1 6 1 56 76 0.836
Ref.total. 26 86 60 102 128 402
Pro.acc 0.846 0.977 0.767 0.960 0.438
Overall Kappa Statistics = 0.6903, Overall Classification Accuracy = 76.12%









Table 2-7. Change trajectories of land use and land cover classes of interest (1993-1998-2003).
Sub= suburban and Veg= vegetation.
Number of Percentage Area in square
Class type classified pixels of area meters
Water-Water-Water 156698 6.06 141028200
Veg-Veg-Veg 990243 38.27 891218700
Sub-Sub-Sub 608897 23.53 548007300
Urban-Urban-Urban 30705 1.19 27634500
Veg-Veg-Urban 59253 2.29 53327700
Veg-Veg-Sub 453099 17.51 407789100
Veg-Sub-Veg 27208 1.05 24487200
Veg-Sub-Sub 48850 1.89 43965000
Urban-Urban-Sub 43647 1.69 39282300
Sub-Sub-Urban 123541 4.77 111186900
Total 2587296 100 2328566400


Table 2-8. Supervised classification 1993 accuracy assessment; overall accuracy, user's
accuracy, producer's accuracy, and KAPPA statistic. Ref. = Reference, Suburb=
Suburban, Pro.acc = Producers accuracy, Clas.Tot = Classified total, and Users.Acc:
Users accuracy.
Supervised classification 1993
1993 Water Vegetation Built up Clas.Tot Users.Acc
Water 30 1 0 31 0.967
Vegetation 0 186 6 193 0.969
Built up 0 0 174 174 1.00
Ref total. 30 187 180 390
Pro.acc 1.00 0.995 0.967
Overall Kappa Statistics = 0.9689, Overall Classification Accuracy = 98.24%









Table 2-9. Supervised classification 1998 accuracy assessment; overall accuracy, user's
accuracy, producer's accuracy, and KAPPA statistic. Ref. = Reference, Suburb=
Suburban Pro.acc = Producers accuracy, Clas.Tot = Classified total, and Users.Acc=
Users accuracy.
Supervised classification 1998
1998 Water Vegetation Built up Clas.Tot Users.Acc
Water 33 1 0 34 0.971
Vegetation 0 212 2 214 0.991
Built up 0 2 99 101 0.980
Ref total. 33 215 101 349
Pro.acc 1.00 0.986 0.980
Overall Kappa Statistics = 0.9729, Overall Classification Accuracy =
98.57%


Table 2-10. Supervised classification 2003 accuracy assessment; overall accuracy, user's
accuracy, producer's accuracy, and KAPPA statistic. Ref = Reference, Suburb=
Suburban Pro.acc = Producers accuracy, Clas.Tot = Classified total, and Users.Acc=
Users accuracy.
Supervised classification 2003
2003 Water Vegetation Built up Clas.Tot Users.Acc
Water 30 1 0 31 0.968
Vegetation 0 186 6 192 0.969
Built up 0 0 174 174 1.00
Ref total. 30 187 174 397
Pro.acc 1.00 0. 995 0. 967
Overall Kappa Statistics = 0.9689, Overall Classification Accuracy = 98.24%


Table 2-11. Change trajectories of land use and land cover classes of interest for supervised
classification images (1993-1998-2003). Built= Built up and Veg= vegetation
Number of classified Percentage of Area in square
Class type pixels area meters
Water-WaterWater 150092 5.54 135082800
Veg-Veg-Veg 2267954 83.64 2041158600
Built-Built-Built 52990 1.95 47691000
Veg-Veg-Built 134703 4.97 121232700
Veg-Built-Built 29009 1.07 26108100
Built-Veg-Veg 60528 2.23 54475200











Alachua County,State of Florida


Legend
Major Road
I Alachua
I Archer
Gainesville
Hawthorne
.___ High Spreings
La Crosse


Micanopy
Newberry
Waldo
Alachua Boundary


Figure 2-1. Map of study area.


0 3.5 7


* Kilometers
28


II I


- -


Figure 2-1. Map of study area.










Gainesville Monthly Rainfall From Feb-1992 to Mar-
1993
14


10

.4 -1993
6
Image
1992-1993
4
Average
S2
0



Months


Figure 2-2. Gainesville monthly rainfall from February 1992 to March 1993.


Gainesville Monthly Rainfall From Jan-1997 to Feb-
1998


1998


Image


.A -x I\1

-4mjV N
_iI\#


-1997-1997-1998
Average


1 41 (\ C, 9 c, -\ 41 4 4 M on
N3' p'4^ 0'''Months
Months


Figure 2-3. Gainesville monthly rainfall from January 1997 to February 1998.


14










0
S8
. S
' 6
4
*I 2
0


-


-
-











Gainesville Monthly Rainfall From Feb-2002 to Mar-
2003
12




1998
S 6 image

S4 2002-2003
2 OV -Average
2



cpi c 92 c ,9'A c, ,' 1< / Rc, ,o r ^


Months


Figure 2-4. Gainesville monthly rainfall from February 2002 to March 2003.











































Legend
o Suburban TS

Pasture TS

o Forest TS

Built up TS
Water TS


Figure 2-5. Aerial photos and training samples in central part of study area.


53


N


K1 2 3 4 nete
0051 2 3 4


























Tasseled Cap Low Pass Supervised Edge Thermal Texture (5x5)
(5x5) Classification Detector Band (band6)





Figure 2-6. Spectral and non-spectral data set analysis layers that created for rule-based
classification, with DEM= Digital Elevation Model; Dist= distance; NDVI=
Normalized Difference Vegetation Index.


TASCAP_6 TASCAP_
-18
-17
-18
-15
-17
-17
-14
-17
-13
-17
-15
-16
-14
-16
-16
-15
-13
-15
-17
-15


Figure 2-7. Sample set of forest extracted training
data layers.


TASCAP_4 TASCAP_3 TASCAP_2 TASCAP_1
29 -37 -9 178
31 -3 15 167
33 9 25 153
31 -7 1 143
33 7 18 152
29 -25 -2 178
33 7 8 136
31 -6 14 178
33 6 6 127
32 8 19 147
34 14 53 185
32 4 10 150
34 8 3 129
33 8 17 140
33 5 12 142
33 4 16 142
33 8 4 132
34 7 12 135
32 5 21 147
33 0 3 134

samples with their associated values from all


Type BAND_6
































Tasseled Cap


Low Pass (5x5)


Supervised
Classification


Edge Detector
(5x5)


Thermal Band
(band6)


Texture (5x5)


-----J


Figure 2-8. Process of creating models that have acceptable accuracy in COMPUMINEN's software, with DEM = Digital Elevation
Model; Dist = distance; NDVI= Normalized Difference Vegetation Index.













Method Number of rules Accuracy Total AUC Precision Recall AUC 1 2 3 4 5
Tree fold 11 73.585 0.913 1 0.500 0.143 0.913 1 0 6 0 0
1 2 0.857 0.643 0.935 1 18 8 0 1
3 0.412 1.000 0.958 0 0 14 0 0
4 0.885 1.000 0.979 0 0 0 23 0
5 0.957 0.647 0.833 0 3 6 3 22
Tree fold 14 73.585 0.922 1 1.000 0.429 0.899 3 0 4 0 0
2 2 0.952 0.714 0.946 0 20 8 0 0
3 0.343 0.857 0.858 0 0 12 0 2
4 1.000 0.913 0.987 0 0 0 21 2
5 0.846 0.647 0.890 0 1 11 0 22
Tree fold 14 70.755 0.910 1 1.000 0.571 0.935 4 3 0 0 0
3 2 0.500 0.964 0.937 0 27 0 0 1
3 1.000 0.143 0.842 0 11 2 0 1
4 0.950 0.826 0.921 0 3 0 19 1
5 0.885 0.676 0.902 0 10 0 1 23
Tree fold 11 77.358 0.927 1 1.000 0.429 0.939 3 0 4 0 0
4 2 1.000 0.714 0.965 0 20 8 0 0
3 0.467 1.000 0.944 0 0 14 0 0
4 0.759 0.957 0.973 0 0 0 22 1
5 0.958 0.676 0.854 0 0 4 7 23
Tree fold 15 71.429 0.916 1 1.000 0.167 0.861 1 0 5 0 0
5 2 0.792 0.655 0.919 0 19 9 0 1
3 0.371 1.000 0.908 0 0 13 0 0
4 0.958 0.958 0.963 0 0 0 23 1
5 0.905 0.576 0.894 0 5 8 1 19
Tree fold 15 80.952 0.953 1 1.000 0.500 0.944 3 3 0 0 0
6 2 0.614 0.964 0.955 0 27 1 0 0
3 0.857 0.429 0.943 0 8 6 0 0
4 0.955 0.875 0.936 0 2 0 21 1
5 0.966 0.848 0.969 0 4 0 1 28
Tree fold 12 76.190 0.928 1 1.000 0.500 0.947 3 0 3 0 0
7 2 0.895 0.607 0.901 0 17 9 0 2
3 0.448 0.929 0.889 0 0 13 0 1
4 0.955 0.875 0.971 0 0 0 21 3
5 0.813 0.788 0.934 0 2 4 1 26
Tree fold 13 78.095 0.963 1 n/a 0.000 0.914 0 0 6 0 0
8 2 0.955 0.750 0.971 0 21 7 0 0
3 0.433 0.929 0.927 0 0 13 0 1
4 0.913 0.875 0.986 0 0 1 21 2
5 0.900 0.818 0.965 0 1 3 2 27
Tree fold 15 81.905 0.964 1 1.000 0.500 0.929 3 1 2 0 0
9 2 0.923 0.857 0.979 0 24 4 0 0
3 0.500 1.000 0.956 0 0 14 0 0
4 0.913 0.875 0.964 0 0 2 21 1
5 0.960 0.727 0.961 0 1 6 2 24
Tree fold 13 80.000 0.960 1 1.000 0.333 0.923 2 0 3 0 1
10 2 1.000 0.679 0.951 0 19 6 0 3
3 0.500 0.929 0.949 0 0 13 1 0
4 0.955 0.875 0.996 0 0 0 21 3
5 0.806 0.879 0.953 0 0 4 0 29


Figure 2-9.Tree models with their number of rules and accuracy.


Correct class


Predicted class


















# 2


> 107.5


43.5


108.5


Figure 2-10. Sample of tree model nodes and branches that used to create classification rules.








STascap-1 >114.51

SSupervised == 4
= New Rule 1
? NDVI <=0.1286491

? Teture-1 > 9.25667

? Tascap-1 > 136

? Supervised == 4
/ =a New Rule 2
Built-up ? NDVI > 0.128649

STexture-1 >9.25667

\ Tascap-1 >114.51

? Tascap-1 <= 136

\ New Rule 3 Supervised == 4

? NDVI > 0.128649

? Texture-1 > 9.25667

Figure 2-11. Set of logic statements that transferred to Knowledge engineering classification
tool.

















Standard
Classification


























Accuracy Assessment
Algorithms s




-led Cap Supervised NDVI Low Pass (5x5) Texture (5x5) Thermal Band j
Classification thand(6)

Y.._.._.._.._.._.._..._.._.._...._.._.._.

SGIS Layers / Training
_.._.._..-"< .Samples

Compumine <----------

Data models

Models accuracy

Rules creation


Knowledge-Based
Classification


Accuracy Assessment



Final Classification


Figure 2-12. Flowchart of knowledge based classification methodology.












N


s















Legend
m ]Boundary
S| No data
Water
1, euelation

S| Buit Up
Suburban


0 2.5 5 10 15 20

Figure 2-13. First supervised classification approach of remotely sensed image for Alachua
County, Florida in 1998.













1200000


1000000

800000 -

600000

400000


200000

0

Water
E Veg-Sub-Veg
Built-Built-Veg
Built-Sub-Sub
Sub-Built-Built
SSi h-VPP-Sllh


* Veg-Veg-Veg
* \eg-Built-Built
* Built-Built-Built
M Built-Veg-Sub
* Sub-Built-Sub
.llh-VPP-Rli lt


Class Type
S*V g -Vt Built
* Veg-Built-Sub
* Built-Built-Sub
* Bult-Vg-'BuiltI
Sub-Sub-Veg
Slih-VPP-VPi


* Veg-Veg-Sub
* Veg-Sub-Built
* Built-Sub-Veg
SBuilt-Veg.Veg
SSub-Sub-Built


SVeg-Built-Veg
SVeg-Sub-Sub
Built-Sub-Built
*Sub-euilr-Veg
Sub-Sub-Sub


Figure 2-14. Changes Trajectories chart classes of interest (1993-1998-2003). Sub= suburban
and Veg= vegetation













N

W E

S








Legend
--] Boundary
Water
Vegetation since 1993
urban since 1993
urban in 2003
urban since 1993 ~ I
Suburban since 1998 -
Suburban in 2003
From Suburban to urban in 200,'
I From urban to Suburban in 21100
Vegetation 93 suburban 9'. v~t1 elallonl 03
U ." 18 .1


Figure 2-15. Knoweldge based classification changes trajectories map classes of interest (1993-
1998-2003). Sub= suburban and Veg= vegetation












N E
W+E
S














Legend ,
7[ ]Boundary
I Iurban
Suburban __
Vegetation
-Id ^^Water I
0 375 75 -15 :'25 30

Figure 2-16. Map of stable land cover classes from 1993 to 2003.














N

W+E

S














Legend
[i_] Boundia <
I Away from Built toward Sub 1998
= Away from Built toward Sub 2003
Away from Sub toward Built 2003
SAway from Sub toward Built 1998
Away from Sub toward Veg 2003
I Away from Veg toward Built 2003
SToward Built 2003
SToward Sub 1998
Toward Veg 2003


i "
L_




1' -1L


0 3.75 7.5


,




...t ..
tome


15 22.5


Figure 2-17. Map of land use and land cover changes from 1993 to 2003.






































Figure 2-18. Sample of suburban area in Alachua County.






























Legend



LIL
1 r idfBIH&f








J:4 j



Figure 2-19. Changes Trajectories map classes of interest for supervised images (1993-1998-
2003). Veg =Vegetation and Built = Built up.









CHAPTER 3
CONCLUSION

Determining what types of land use and land cover change have occurred in Alachua

County from 1993 to 2003 and what type of classification method produces the best results were

the main objectives of this research. So, three different classification methods were used to detect

the changes that occurred in the study area. These three classification methods were performed

for four classes; water, vegetation, urban, and suburban.

The initial classification technique employed was a simple unsupervised classification.

This was performed in 10, 20, and 30 classes for the study area in order to determine the

classification scheme. Based on the number of cluster classes that revealed from the

unsupervised classification fives land use and land cover classes created and to increase the

classification accuracy some of these clustered classes regrouped.

Second, a supervised classification methodology was used but this produced unacceptable

levels of accuracy, due specifically to confusion across certain classes in the landscape. As such,

he results of this type of classification were not acceptable for academic research because the

accuracy was low. To increase the accuracy of the classification, knowledge based classification

was performed for all images. This method increased the accuracy of all images, however there

was 17.51 % of the total area transformed from vegetated area to suburban. This large area of

land transformation indicated that there was misclassification between vegetation and suburban

areas. This misclassification relates to the confusion between these two classes because many

suburban developments include a large amount of vegetative cover and so spectrally the two are

not dissimilar. Hence, a final methodology was employed in which some classes were merged to

decrease the confusion between suburban and vegetation and this produced three major classes

built, water, vegetation.









Alachua County, like other regions that have a large amount of vegetation cover within the

suburban developments area, causes potentially significant problems of misclassification

between suburban and vegetation classes, for many developed regions in the U.S. Even using a

more advanced image classification technique, such as the knowledge based classifier, the

confusion between suburban and vegetation classes was not eliminated, and ultimately a

traditional classification for water, built, and vegetation classes only was performed on all

images. The accuracies of this classification were acceptable greatly improved and it provides a

better understanding of overall change trends. For future work better methods need to be

developed to monitor suburban developments and their changes over time, such as using a GIS

layer for suburban development in order to improve the ability of differentiating suburban from

vegetation. Moreover, using finer spatial resolution satellite images would help in differentiating

between suburban and vegetation.

This research found that western and central portions of Alachua County experienced most

of the changes in land use and land cover. The changes in land use and land cover have taken

two major trends, from built to vegetation area and from vegetation to built structures. 1.07 % of

the total area that have transformed from vegetation to built structures area in 1998. 4.97 % of

the total area that have transformed from vegetation to built structures area in 2003. The total

area that have transformed from built structures area to vegetation in 1998 was 2.23 %. The

western portion of Alachua County and Gainesville as the largest and most populated area has

experienced most of these changes in land use and land cover. The outputs of this research could

be used by the county planner as reference for future estimation and to understand land use and

land cover change trends in Alachua County.









LIST OF REFERENCES


BINFORD, M. W., GHOLZ, H. L., STARR, G., and MARTIN, T. A., 2006, Regional carbon
dynamics in the southeastern U.S. coastal plain: Balancing land cover type, timber
harvesting, fire, and environmental variation (DOI 10.1029/2005JD006820). Journal of
geophysical research., 111, D24S92.

BROCKERHOFF, M. P., 2000, An Urbanizing World. Population Bulletin, 55, 1.

CLARK, C., 1967, Population growth and land use (London; Melbourne [etc.; New York:
Macmillan; St. Martin's P.).

CITY of GAINESVILLE., 2008, Gainesville Facts. A available online at:
http://www.cityofgainesville.org/about/ (accessed April 2008).

DANIELS, A., 2006, Incorporating domain knowledge and spatial relationships into land cover
classifications: a rule-based approach. International Journal of Remote Sensing, 27, 2949.

DIGITAL MAPPING SYSTEMS., 2006.Glossary. Available online at:
http://www.digimap.gg/glossary (accessed March 2008).

DOHRENWEND, R. E., 1978, The climate of Alachua County, Florida (Gainesville, Fla.:
Agricultural Experiment Stations, Institute of Food and Agricultural Sciences, University
of Florida).

FOLLY, A., BRONSVELD, M. C., and CLAVAUX, M., 1996, A knowledge-based approach
for C-factor mapping in Spain using Landsat TM and GIS. International Journal of Remote
Sensing, 17, 2401-2415.

FU-JEN CHIEN, T.-Y. C., 2000, The Study of Knowledge-Based Database Assist for Urban
Land Use Classification (100 WENHWA RD., TAICHING, TAIWAN: Feng Chia
University ).

GALBARD, A., and HAUB, C., World population beyond six billion. Population Bulletin, 54, 3.

GONG, P., and HOWARTH, P. J., 1992, Frequency-Based Contextual Classification and Grey-
Level Vector Reduction for Land-Use Identification. PHOTOGRAMMETRIC
ENGINEERING AND REMOTE SENSING, 58, 423.

GONG, P., and HOWARTH, P. J., 1990, An assessment of some factors intluencing
multispeetral land-cover classification.: Photogrammetric Engineering & Remoie
Sen.sing.), pp. pp. 597 603.

GREEN, G. M., C.M. SCHWEIK, AND J.C. RANDOLPH. 2005, "Retrieving Land-Cover
Change Information from Landsat Satellite Images by Minimizing Other Sources of
Reflectance Variability". In Seeing the forest and the trees : human-environment
interactions in forest ecosystems, edited by E. F. M. E. Ostrom (Cambridge, Mass.: Mit
Press).









HAACK, B. N., SOLOMON, E. K., BECHDOL, M. A., and HEROLD, N. D., 2002, Radar and
Optical Data Comparison/Integration for Urban Delineation: A Case Study.
Photogrammetric engineering and remote sensing., 68, 1289.

HAILD, M., 1997, Land Use-Cover Change Detection Using Knowledge based approaches:
Remote Sensing and GIS (50300 Jalan Tun Razak, Kuala Lumpur Malaysia Kalaysian
Center for Remote Sensing (MACRES)).

HAN, F., 2002, An examination of historic land use changes in Alachua County, Florida : a
technological approach.

HELP, A. D., 2005, ArcEditor (Redlands, CA: Environmental Systems Research, Institute
(ESRI)).

HEROLD, M., GOLDSTEIN, N. C., and CLARKE, K. C., 2003, The spatiotemporal form of
urban growth: measurement, analysis and modeling. Remote Sensing of Environment, 86,
286.

HOUGHTON, R. A., The worldwide extent of land-use change. Bioscience, 44, 305.

HSU, S.-Y., 1978 Texture-tone analysis for automated land-use mapping Photogrammetric
Engineering and Remote Sensing.).

INTERIOR, U. S. D. O. T., 2007, USGS Isis Workshop: Introduction to Filters: Astrogeology
Research Program ).

JANSSEN, L. L. F., and MIDDELKOOP, H., 1992, Knowledge-based crop classification of a
Landsat Thematic Mapper image. International Journal of Remote Sensing, 13, 2827.

JENSEN, J. R., 1996, Introductory digital image processing : a remote sensing perspective
(Upper Saddle River, N.J.: Prentice Hall).

LU, D., MAUSEL, P., BATISTELLA, M., and MORAN, E., 2005, Land-cover binary change
detection methods for use in the moist tropical region of the Amazon: a comparative study.
International Journal of Remote Sensing, 26, 101.

LU, D., MAUSEL, P., BRONDIZIO, E., and MORAN, E., 2004, Change detection techniques.
International Journal of Remote Sensing, 25, 2365.

LU, D., and WENG, Q., 2006, Use of impervious surface in urban land-use classification.
Remote sensing of environment., 102, 146.

MACRI PELLIZZERI, T., 2003, Classification of polarimetric SAR images of suburban areas
using joint annealed segmentation and "H/A/a" polarimetric decomposition. ISPRS
JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 58, 55-70.

MAKTAV, D., ERBEK, F. S., and JURGENS, C., 2005, Remote sensing of urban areas.
International Journal of Remote Sensing, 26, 655.









MARCEAU. D.J.. HOWARTH, P. J., DTIBOIS, J.M.M. AND GRATTON, D.J, 1990,
Evaluation of the grey-level co-occurrence matrix-method for land cover classiHcation
using SPOT imagery. IEEE Transactions on Geoscience and Remoie Sensing., 28, 513-
519.

MERTENS, B., and LAMBIN, E. F., 2000, Land-Cover-Change Trajectories in Southern
Cameroon. Annals of the Association of American Geographers, 90, 467.

MESEV, V., 1998, The Use of Census Data in Urban Image Classification.
PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING, 64, 431-438.

MEYER, W. B., and TURNER, B. L., 1994, Changes in land use and land cover : a global
perspective (Cambridge [England]; New York, NY, USA: Cambridge University Press).

MEYER, W. B., and TURNER II, B. L., 1992, Human Population Growth and Global Land-
Use/Cover Change. Annual Review of Ecology & Systematics, 23, 39.

ONSI, H. M., 2003, Designing a rule-based classifier using syntactical approach. International
Journal of Remote Sensing, 24, 637-647.

PETIT, C., SCUDDER, T., and LAMBIN, E., 2001, Quantifying processes of land-cover change
by remote sensing: resettlement and rapid land-cover changes in south-eastern Zambia.
International Journal of Remote Sensing, 22, 3435.

PHILLIPS, S., 1986, Factors Influencing Agricultural Land Use Change: A Study of Alachua
County, Florida.: University of Florida., Gainesville, Florida.

PUISSANT, A., HIRSCH, J., and WEBER, C., 2005, The utility of texture analysis to improve
per-pixel classification for high to very high spatial resolution imagery. International
Journal of Remote Sensing, 26, 733.

SOUTHWORTH, J., 2004, An assessment of Landsat TM band 6 thermal data for analysing land
cover in tropical dry forest regions. International Journal of Remote Sensing, 25, 689.

SOUTHWORTH, J., NAGENDRA, H., CARLSON, L. A., and TUCKER, C., 2004, Assessing
the impact of Celaque National Park on forest fragmentation in western Honduras. Applied
Geography, 24, 303.

SOUTHWORTH, J., NAGENDRA, H., and TUCKER, C., 2002, Fragmentation of a Landscape:
incorporating landscape metrics into satellite analyses of land-cover change.
LANDSCAPE RESEARCH, 27, 253-270.

U.S. CENSUS BUREAU., 2002, State and County QuickFacts. Available online at:
http://quickfacts.census.gov/qfd/states/12/12001.html (accessed March 2008).

U.S. GEOLOGICAL SURVEY., 2002, Public Land Survey System of the United States (Reston,
VA). Available online at: http://erg.usgs.gov/isb/pubs/gis_poster/index.html#what
(accessed March 2008).









TURNER, B. L., and INTERNATIONAL GEOSPHERE-BIOSPHERE PROGRAM "GLOBAL,
C., 1995, Land-use and land-cover change : science/research plan (Stockholm:
International Geosphere-Biosphere Programme).

TURNER, B. L. I. A. W. B. M., 1994, Land-Use and Land-Cover Change: An Overview. In
Changes in Land Use and Land Cover: A Global Perspective., edited by W. B. B. L. T. I.
MEYER (Cambridge, New York, Melbourne.: University of Cambridge).

TURNER II, B. L., and MEYER, W. B., Land use and land cover in global environmental
change: considerations for study. International Social Science Journal, 43, 669.

XIAOJUN, Y., and ZHI, L., 2005, Using satellite imagery and GIS for land-use and land-cover
change mapping in an estuarine watershed. International Journal of Remote Sensing, 26,
5275.

YANG, X., and LO, C. P., 2002, Using a time series of satellite imagery to detect land use and
land cover changes in the Atlanta, Georgia metropolitan area. International Journal of
Remote Sensing, 23, 1775.

YUNHAO, C., PEIJUN, S., XIAOBING, L., JIN, C., and JING, L., 2006, A combined approach
for estimating vegetation cover in urban/suburban environments from remotely sensed
data. Computers & Geosciences, 32, 1lp.

ZHANG, J., and FOODY, G. M., 1998, A fuzzy classification of sub-urban land cover from
remotely sensed imagery. International Journal of Remote Sensing, 19, 2721-2738.

ZHOU, B. A. K., K.M., 2006, Neighborhood Impacts on Land Use Change: A Multinomial
Logit Model of Spatial Relationships. The Annals of Regional Science.









BIOGRAPHICAL SKETCH

Muhammad Almatar was born in Kuwait. In 1997, He graduated from Salah Shiehab High

School, in his home country, majoring in science. He received his undergraduate degree as an

honor student in geography from Kuwait University in 2003. He worked as a high school teacher

in Salah Shiehab High School for approximately a year. In 2004, Kuwait University offered him

a scholarship to obtain his master's and PhD degrees in geographic information systems.

Muhammad intends to attend the University of Florida to obtain his PhD in geography.





PAGE 1

1 USING REMOTE SENSING AND GIS TO STUDY LAND-USE AND LAND-COVER CHANGE IN ALACHUA COUNTY, FLORIDA FROM 1993 TO 2003 By MUHAMMAD ALMATAR A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2008

PAGE 2

2 2008 Muhammad Almatar

PAGE 3

3 To my parents, wife, and children

PAGE 4

4 ACKNOWLEDGMENTS I acknowledge and thank Dr. Jane Southworth, for her guidance and supervision of this research. I thank Dr. Binford for providing m e with the research satellite images. I thank my research committee members and Dr. Waylen. I acknowledge my good colleague, Steven Forrest, for being supportive. I thank my parents and my wife for their support and understanding.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4LIST OF TABLES................................................................................................................. ..........7ABSTRACT.....................................................................................................................................9CHAPTER 1 INTRODUCTION..................................................................................................................10Background.............................................................................................................................10Study Area..............................................................................................................................11Use of Remote Sensing and GIS............................................................................................ 112 UTILIZING REMOTE SENSING AND GI S T O STUDY LAND-USE AND LANDCOVER CHANGE IN ALACHUA COUNT Y, FLORIDA FROM 1993 TO 2003.............. 13Introduction................................................................................................................... ..........13Background......................................................................................................................13Land Use and Land Cover Change..................................................................................14Urbanization and Land Use/ Land Cover Change........................................................... 14Use of Remote Sensing and GIS..................................................................................... 16Methodology...........................................................................................................................18Study Area.......................................................................................................................18Image Pre-Processing...................................................................................................... 21Classification Scheme.....................................................................................................21Training Samples and Fieldwork.....................................................................................22Image Classification........................................................................................................ 22Normalized Difference Vegetation Index................................................................ 23Texture analysis........................................................................................................ 24Low pass filter..........................................................................................................24Tasseled cap transformation.....................................................................................25Thermal band............................................................................................................ 26Non-spectral data layers........................................................................................... 26Rule-based classifier................................................................................................ 27Change Detection Analysis.............................................................................................29Results.....................................................................................................................................30Accuracy Assessment...................................................................................................... 30Change Trajectories.........................................................................................................31Discussion...............................................................................................................................34Summary of Finding........................................................................................................ 34Implication of Study........................................................................................................ 35Conclusion.......................................................................................................................40

PAGE 6

6 3 CONCLUSION..................................................................................................................... ..67LIST OF REFERENCES...............................................................................................................69BIOGRAPHICAL SKETCH.........................................................................................................73

PAGE 7

7 LIST OF TABLES Table page 2-1 Land use and land cover descriptions for classification scheme. ......................................42 2-2 Final decision based rules used in the final classification procedure. ...............................43 2-3 Supervised classification 1998 accuracy assessment......................................................... 46 2-4 Rule-based classification 2003 accuracy assessment........................................................ 46 2-5 Rule-based classification 1998 accuracy assessment........................................................ 47 2-6 Rule-based classification 1998 accuracy assessment........................................................ 47 2-7 Change trajectories of land use and land cover classes of interest (1993-1998-2003). ..... 48 2-8 Supervised classification 1993 accuracy assessment......................................................... 48 2-9 Supervised classification 1998 accuracy assessment......................................................... 49 2-10 Supervised classification 2003 accuracy assessment......................................................... 49 2-11 Change trajectories of land use and land cover classes of interest for supervised classification im ages (1993-1998-2003)............................................................................ 49

PAGE 8

8 LIST OF FIGURES Figure page 2-1 Map of study area...............................................................................................................502-2 Gainesville monthly rainfall fr om February 1992 to March 1993.....................................512-3 Gainesville monthly rainfall fr om January 1997 to February 1998...................................512-4 Gainesville monthly rainfall fr om February 2002 to March 2003.....................................522-5 Aerial photos and training sample s in central part of study area.......................................532-6 Spectral and non-spectral da ta set analysis layers th at created for rule-based classification......................................................................................................................542-7 Sample set of forest extracted training samples with their associated values from all data layers ................................................................................................................... ......542-8 Process of creating models that ha ve acceptable accuracy in COMPUMINENs software..............................................................................................................................552-9.Tree models with their number of rules and accuracy............................................................. 562-10 Sample of tree model nodes and branches th at used to create classification rules............ 572-11 Set of logic statements that transferred to Knowledge engineering classification tool.....582-12 Flowchart of knowledge based classification methodology.............................................. 592-13 First supervised classification approach of remotely sensed image for Alachua County, Florida in 1998.....................................................................................................602-14 Changes Trajectories chart cla sses of interest (1993-1998-2003)..................................... 612-15 Knoweldge based classification changes trajectories map classes of interest (19931998-2003). ................................................................................................................... ....622-16 Map of stable land cover classes from 1993 to 2003.........................................................632-17 Map of land use and land cover changes from 1993 to 2003............................................ 642-18 Sample of suburban area in Alachua County.....................................................................652-19 Changes Trajectories map classes of in terest for supervised images (1993-19982003). ........................................................................................................................ ........66

PAGE 9

9 Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science USING REMOTE SENSING AND GIS TO STUDY LAND-USE AND LAND-COVER CHANGE IN ALACHUA COUNTY, FLORIDA FROM 1993 TO 2003 By Muhammad Almatar August 2008 Chair: Jane Shouthworth Major: Geography This research investigates land use and la nd cover change in Al achua County, FL from 1993 to 2003. Determining the best classification me thod to classify satellite images for the study area was the second objective of this resear ch. Different classification methods were used in the research in order to improve the cl assification accuracy. Supe rvised classification (Gaussian Maximum Likelihood) was the first appr oach to classify three Landsat images. The accuracy assessment of this classification wa s not acceptable for remote sensing research. Knowledge based classification was used to impr ove the accuracy in a ll images. The accuracy increased in all images, however there was misc lassification in suburban class due to the mixed texture of this area. A third classification was performed after removing suburban class, the results of this classification were acceptable for academic research. The tr ends of changes in land use and land cover are taking a place in west and central portions of Alachua County. This research found that 4.97% of the total area tr ansformed from vegetation to built areas by 2003, 1.07% of the total area transf ormed from vegetation land cover to built area by 1998, and 2.23 % of the total area have converted from built structure to vegetation area in 1998.

PAGE 10

10 CHAPTER 1 INTRODUCTION Background Am ong social sciences disciplines Geography is the only science that examines the relations between society and the natural envi ronment (Peet, 1998). Geography as known is the study of the earths surf ace and how natural force and living organisms reshape this surface. Analyzing the environmental system from both physical and human perspectives, over time, allows us to investigate how humans modify th e earths surface and how humans change their surrounding environment to obtain th eir essential needs such as f ood and shelter. According to Turner and Meyer (1994) the earth has experien ced a great deal of change, including species extinctions, and changes such as expansion of ci ties and loss of forests, although the world has never experienced such rapid land changes as it is today. The inter action between humans and their surrounding environm ent is resulting in changes on land cover patterns and these changes could be monitored to recognize human actions that drive them to enable us to better understand, model, manage and predict further environmenta l change (Turner and Meyer, 1994). One way to monitor these changes on the earths surface is to study land use and land cover changes of a specific geographic area over a period time. Land use and land cover change is the appro ach of monitoring ch anges on the earths surface that occurred as a func tion of the interaction between humans and their surrounding environment. Land use refers to the use of the la nd that reflects human behavior. For example, a forested land cover could include multiple land uses It could be used as a resource for logging or as a park for recreation purposes. Land cover chan ge, may be related to changes in land use, or could simply be a function of changes in the natural system e.g., wetland area changing to bare land as a function of changes in the climate syst em (Turner and Meyer, 1994). Land use and land

PAGE 11

11 cover change plays an important role in recent research in order to unde rstand the environmental and social consequences of change and to identif ying the temporal and spa tial extent of changes on the earths surface. Today, Remote Sensing and Geographic Information System (GIS) have been used widely to study and analyze land use and land cover. This study examines the changes that have occurred in land use and land cover in Alachua County from 1993 to 2003. Study Area Alachua County is located in the North Centra l part of Florida. Alachua County includes num ber of cities Alachua, Archer, Hawthorne High Springs, LaCrosse, Micanopy, Gainesville, Newberry, and Waldo across 960 square miles (Ph illips, 1986). Alachua Countys climate differs from other parts of Northern Florida, due to its unique latitudinal location, the variety of surface features, and distance from the sea. Gainesville is this regions largest city and most populated, and is home of the largest public Un iversity in the state of Florida (The University of Florida). Use of Remote Sensing and GIS According to Yang et al. (2002) land use and land cover change could be monitored by using rem otely sensed date to study spatial and temporal features of these changes. Remote sensing is the science of gathering informa tion about the earths su rface without touching it (Digital Mapping Systems, 2006). Analyzing this information over spatial and temporal scales could be an excellent tool to detect changes that occurred on the earths surface in order to provide a better understanding of how humans change their surrounding environment. According to U.S. Geological Survey, GIS is a computer system and software that used to manipulate, amass, organize and present geographically referen ced data based on their location. It is basically a database system that used to display ge ographically referenced data as map images. Remote Sensing techniques have been used widely in research to classify satellite images according to the spectral signature of the land co vers. These land cover types will be presented as

PAGE 12

12 categories and each category include s a group of pixels that have similar values. Changes in land use and land cover that occurred in the earths surface could be obtained from the classified satellite images. Our study used Remote Sensing and GIS to examine changes in land use and land cover in Alachua County from 1993 to 2003. Three Landsat images were obtained from two different satellites ( Landsat TM and Landsat ETM), and used to monitor changes in land use and land cover. These three satellites images classifi ed by three different techniques, unsupervised classification, supervised classification and knowle dge based classification in order to improve the classification of remotely sensed images and to detect the trend of land use and land cover change for Alachua County, Florida from 1993 to 2003. RQ1: What type of land use and land cover change patterns have occurred in Alachua County from 1993 to 2003? RQ2: What type of classification method produces the best results? Why?

PAGE 13

13 CHAPTER 2 UTILIZING REMOTE SENSING AND GIS T O STUDY LAND-USE AND LAND-COVER CHANGE IN ALACHUA COUNTY, FLORIDA FROM 1993 TO 2003. Introduction Background Land transform ations have been going on si nce the beginning of time. The earth has experienced the spread and extinc tion of species, loss of forests, drainage of wetlands, expansion of cropland and major developments of cities, but the world has never suffered such rapid land transformations as it is today (Turner and Meyer, 1994). Acco rding to many scholars, recent environmental and land changes are attributed to human activities. Th e human impact on the earth has reached its peak, is unprecedented in its scale and is significantly influencing the resources needed in sustaining the biosphere (Tur ner and Meyer, 1994). It is therefore important to recognize these human actions and the forces that drive them to understand, model, manage and predict further environmental change (Tur ner and Meyer, 1994). One way to accomplish this is by studying the land use and land cover change s of a specific geographic area over a period time. Today, modern technologies in Remote Sensi ng and Geographic Information System (GIS) have made it easier to study and analyze land us e and land cover. This research examines the changes in land use and land cover in Alachua County from 1993 to 2003. Alachua County is located in the north-central part of the state of Florida. Gainesville is this regions most populated and largest city, and is home of the largest public University in Florida: The University of Florida. One the of the projects objectiv es is to see what kind of la nd use and land cover changes Alachua County experienced in the last decade. Urban planners can use this information to plan for the future and potentially overcome any proj ected mishaps. More importantly, this project

PAGE 14

14 aims to showcase the need in using GIS and remo te sensing technologies to determine, analyze, and predict land use and land cover change. The ear th is changing ever so rapidly, and land use and land cover research is essent ial to providing solutions to overcome such changes (Maktav et al., 2005). Land Use and Land Cover Change Land use and land cover change is the appro ach of understanding ch anges on the earths surface that are reached by the in teraction between humans and their environment. Land use refers to the use of the land associated with hum an behavior. A forested land cover, for example, can have multiple land uses. It could be used for recreation purposes, as a park or even a plantation forest. Land cover, on the other hand, refers to changes in land cover, not related to land use. For example, forest land changing to bare land (Turner and Meyer, 1994). Land cover change can cause either convers ion or modification. Conversion refers to changing from one class to another, like from forest to cropland. Modification involves a change within a land cover class, like a change in compos ition of a forest. However, while land cover changes usually follow those of land use, the opposite is not always true: land cover change could occur even if the land is unaltered by humans (Turner and Meyer, 1994). According to Turner and Meyer (1994), la nd use and land cover are connected by the proximate sources of change, which are th e human actions that influence the physical environment directly. These proximate sources re flect social, economical, political, and cultural aspects in humans that shape and drive the ch anges in land use (Turner and Meyer, 1994). Urbanization and Land Use/ Land Cover Change The increase of urbanization gl oba lly is certainly one of the human driving forces that directly affects land use and land cover. This si gnificant increase in ur ban populations and urban areas is continuously affecting natural and human environments at all geographic scales and

PAGE 15

15 locations (Herold et al., 2003). The worlds population is projected to increase from 6.1 billion in 2000 to 7.8 billion in 2025 with the urban share estimated to reach 58 percent. By 2025, the majority of the worlds populati on will be living in cities or ur ban areas (Gelbarad et al 1999). For developing countries this trend can lead to problems like poverty, health issuesincrease in infant mortality and decrease in life expectancyand natural di sasters (Brockerhoff, 2000). For the more developed world, which this study is more focused upon, challenges are arising less from population growth and more from the ch anges in population composition and distribution (Brockerhoff, 2000). In the United States, unaffordable housing, in adequate educational opportunities and lack of public safety in inner cities have all aided increasing povert y and homelessness rates. Also, the increase of legal and illegal immigration in the United States could be viewed as a great opportunity for diversity and cultu ral interaction, but ethnic immigr ants living in urban cities create ethnic clusters that usually cause intolerance and racial tensions between different immigrant groups. These same challenges, in addition to low govern ment investments in the infrastructure of inner cities have motivated the more affluent and middle-class population to relocate to the suburbs. This movement of people and jobs away from the central cities toward suburbs of major metropolitan areas is called decentral ization, and it is considered as one of the most salient trends of urbanization in the United St ates since the 1960s (Brockerhoff, 2000). The increasing growth of suburbs, however, results in what is called urban sprawl, which is an outward growth of subdivisions of land segregated by specific use, including low-density commercial and residential settlements.

PAGE 16

16 According to Brockerhoff (2000), the future of Americas urban and spatial trends will be determined by international flow s of capital, information, labor, technology and decisions made by individuals and government agen cies. But the future needs to be beneficial for the future generation, and urban planners ne ed to use the resources and m echanisms available to analyze past and current land use and land cover change predict future imp lications, and provide solutions to prevent those potential outcomes In other words, effective schemes by urban planners require up-to-date and complete information regarding past la nd use changes and their impacts. Certainly, the increasing intere st in research on land use and land cover stems from a growing concern over globally occurring environmental changes (Houghton, 1994). As mentioned above, land use and land cover change is the approach of understanding changes on the earths surface that is reached by interpre ting the interactions between humans and their environment. These changes could affect enviro nmental aspects like climate change, hydrologic processes and extermination of pl ant and animal species. However, systematic analyses of the change trends are needed to further comp rehend and understand the drivers behind these transformations (Tur ner el al., 1995). Land use and land cover change plays an importa nt role in choosing the correct research approaches that will lead to identifying the tem poral and spatial extent of changes on the earths surface. In order to understand the environmental and social consequences of change, researchers have to observe and monitor between various patt erns of change. To do so, researchers have to utilize the technologies in GIS and remote sensing. Use of Remote Sensing and GIS Re motely sensed date can be used to study sp atial and temporal features of land use and land cover change (Yang et al. 2002). Remote se nsing is the science of obtaining information

PAGE 17

17 about the earths surface, usually fr om aircraft or satellite, using instruments that may use visible light, infra-red or radar to acquire data (Dig ital Mapping Systems, 2006). Use of such data over spatial and temporal scales can detect changes occurring on the earth a nd can provide a better understanding of how the land is being influe nced by human activity. Having the ability to observe and collect data for large areas within a short period of time, re mote sensing can be a useful tool for GIS. According to U.S. Geological Survey, GIS is a computer system and software whose purpose is to organize, amass, manipulate and present geographi cally referenced data based on their location. It is basically a database system that manages spatially referenced data which can be graphically displayed as map images. There are several Remote Sensing and GIS tec hniques that have been employed in research to classify satellite images based on the spectral signature of land covers Image classification is the process of grouping pixels that have similar values into cla sses or categories, which signify certain types of land use and land cover. Classifi ed satellite images provide information of land use and land cover classes, which generally used to identify changes in the earths surface. Parametric techniques are commonly used to classify remotely sensed data that are distributed normally by using map algebra, which clusters like-pixel values in the image statistically. Maximum likelihood and unsupervised classification are widely used as parametric techniques. For non-normally distributed remotely sensed data nonparametric techniques are implemented to classify the imagery such as fuzzy and neares t-neighbor classification. One problem relating to such optimal statistical clusteri ng of data is that it may not correspond to very meaningful information classes when using unsupervised methods. On the other hand, spectral signatures

PAGE 18

18 used in supervised classification often overlap an d make effective discrimination based solely on spectral reflectance characteristics very difficult (Daniels, 2006). The level of difficulty in achieving an accurate land cover classification relates to the definition of the land cover classes used. For ex ample, urban and suburba n areas are not similar to other land cover types, which are more easily classified such as water and forest, because of its heterogeneous nature. Urban and suburban as land use/cover classes usually include other types of land cover in addition to a built signature such as grass, water, soil, and trees besides the built materials. The mixed spect ral signatures of this land cove r result in a need for advanced classification techniques to differe ntiate between them. In this pa per, GIS data sets, implemented with remotely sensed data in a rule based cl assification technique, enga ge both spectral and non spectral information to try to improve classification accuracy of remotely sensed images for Alachua County, FL and detect the changes that occurred in land use and land cover across the region, from 1993 to 2003. Study questions addressed are; RQ1: What type of land use and land cover change patterns have occurred in Alachua County from 1993 to 2003? RQ2: What type of classification me thod produces the best results? Why? Methodology Study Area Located in the North Central part of Florida, Alachua County covers about 960 square m iles (Figure 2-1). Alachua, Archer, Hawt horne, High Springs, LaCrosse, Micanopy, Gainesville, Newberry, and Waldo are all part of Alachua County (Phillips, 1986). Due to its unique latitudinal location, di stance from the sea and the variety of surface features, Alachua Countys climate differs from other parts of Northern Florida. The county has one warm rainy season and a cool dry season. The rainy season begins in the middle of May and

PAGE 19

19 ends in September. The average annual rainfall for Gainesville, which is the largest city in Alachua, is 35 inches per year, with most of the rain falling between June and July (Dohrenwend, 1978, City of Gainesville, 2008). Th e average high temperature in winter is 18.3 C and the average temperature in the summer is 31.6 C (City of Gainesville, 2008). Monthly rainfall from 1992 to 2003 was provide d by Gainesville Airport Weather Station (Figure 2-1). Figure 2-2 shows the monthly rainfall for Gainesville, Alachua County from February 1992 to March 1993 and the average rainfall for 1992-2003 fro comparison. The figure indicates that the rainfall in February 1993, one month before obtaining the Landsat TM 1993 image, was above average by 1.75 inches. This difference in the rainfall made the study area slightly wetter than normal. Figure 2-3 demonstr ates the monthly precipitation for Gainesville from January 1997 to February 1998 and the mo nthly rainfall from February 1992 to March 2003. In January 1998 the rainfall one month be fore the Landsat TM 1998 was obtained, was more than the average by 8.56 inches. Since Al achua County experienced the highest amount of rainfall in the beginning of 1998, the area of stud y was wetter than any time else across the study period. Figure 2-4 shows the mont hly precipitation for Gainesville from February 2002 to March 2003 and the average monthly rainfall from Febr uary 1992 to March 2003.The figure indicates that the rainfall in February 2003, one month before obtaini ng the Landsat ETM 2003 image, was below the average by 2.98 inches. The rainfa ll in February 2003 made the study area drier. These differences in precipitation may affect the amount of area in th e Water land cover class which can be easily monitored, but should not impact the other classes, du e to them being such broad descriptions of land cover e.g., an area would not convert from any land cover class due to precipitation difference given here, accept for a tem porary increase or decrease in a water class,

PAGE 20

20 due to flooding or drought, and really only 1998 mi ght show significant ch anges. If there are potential impacts, it wi ll be discussed in the results section. In the last decade, Alachua County experi enced major population growth, mainly due to the expansion of University of Florida and the Gainesville vicin ity. According to the population Census Counts, the population in Alachua C ounty increased from 181, 600 in 1990 to 217,955 in 2000, an equivalent of about a 20 percent increas e. In 2006, the population was estimated to be 227, 120 (U.S. Census Bureau). Also, the recent success of the University of Floridas basketball and football teamsNational Basketballs ch ampions in 2005 and 2006, and the 2006 football championshas given the city of Gainesville media publicity and in essence, increased its popularity. These athletic successes caused mo re high-school graduates to apply for the University of Florida. The successes also brough t in more money to the city, which has helped in its growth. With this growing population, the land use in Alachua County has certainly experienced significant changes. Most of the land use change as implied above, has occurred in Gainesville. The growth of the University of Florida produced changes in Alachua county and Gainesville. The downtown area in Gainesville has become a government and professiona l center, and retail stores migrated west to Interstate 75. As a result of thes e changes, population growth has traveled west, and government o fficials have trying to maintain a balance between the areas natural environment and this growt h. (Gainesvilletoday.com, 2008) Because of these rapid changes and growth in population occurring in Alachua County, it was inferred that studying and anal yzing the land use and land cove r change of this region would provide valuable information for urban planners to potentially overcome any upcoming crisis and to successfully plan for the future.

PAGE 21

21 Image Pre-Processing Two Landsat TM im ages were obtained fo r Path 17, Row 39 in March 1993, February 1998, and one Landsat ETM image in March 2003. Since the most recent Landsat ETM image that was available was 2003 (pre Landsat 7 problems) this ten year period was determined as the study period. The imagery was obtained during the dry season, when cloud free images are more available. However, the March 1993 image had so me small areas of clouds that were then masked out of all images by using subset anal ysis. The masked areas are assigned as no data within the study site. In order to reference the images sixty distributed gro und control points (GCP), where gathered with GPS receivers in region. UTM coordinate system (zone 17 N), NAD 83 datum was used to register all images and field checked for accuracy (Binford et al 2006). The root mean square error (RMS) for all imag es ranged from 6.8 m to 8.9 m and the average was 8.0 m. CIPEC atmospheric cal ibration techniques (Green et al., 2005) were performed on all images to minimize the atmospheric variation such as solar elevation and curve, and atmospheric effects. Classification Scheme To determ ine the classification scheme unsupe rvised classification was performed in 10, 20, and 30 classes for the study area. Based on the clustered cla sses of unsupervised classification, five land use and land cover classes were desi gnated and to increase the classification accuracy some land cover classes grouped under their main types (Table 2-1). Land use classes related to suburban, urban, pasture. Land cover classes were water and forest. This proposed classification scheme was based on the main aim of the paper which is land use and land cover changes in Alachua County from 1993 to 2003, much of which is urban and suburban development, and loss of more natural vegetation in the process.

PAGE 22

22 Field work was conducted for the study area to experiment with the efficiency of the classification scheme. Water as land cover, consists of an area that is inundated by water such as rivers, lakes, and wetland. Forest is area cont aining different species of trees and bushes with 25% canopy closure. Pasture as la nd use, composed of agricultural land, pasture, grassland, and others. Urban, composed of 70-100% built materi als such as the airport and shopping malls. Suburban, composed of 40-70% built materials which are mixed with vegetation such as Residential. Training Samples and Fieldwork Training samples for the study ar ea were collected with Trimble GPS fr om the field. Trimble GPS is one of well known high accurate GPS, which has < 1m spatial accuracy. Field work was conducted from May to August 2007. Around 300 distributed training samples were gathered from across the study area (Figure 2-5). These training samples were used with aerial photos, from Florida Geographic Data Library, to determine land use land cover types. The interpretation of aerial photos increased the number of training samples from 300 to over 1000. These training samples were used to implemen t the classification and then to assess the classification via an accuracy assessment. Mo reover, around 800 photos were taken during the field work for land use and land cover types in the study area. Image Classification In the first part of this research trad itional classification m ethodology supervised classification using the Gaussi an Maximum Likelihood was perf ormed on all image dates in order to compare to decision based classifi cation method. The average accuracy assessment (overall accuracy) of these images was not more than 55%, which is unacceptable for remote sensing classification. These classe s are hard to differentiate fr om some spectrally similar land

PAGE 23

23 cover classes. As result, it was essential to implement an advanced classification technique which was rule-based in nature. Building on this initial classi fication methodology a rule-based classification scheme was then created for the study region. For this analysis additional data layers had to be created. A number of spectral analysis layers were created in the ERDAS Imagine processing software for use in the classification process. These are the initial classification; an Normalized Difference Vegetation Index (NDVI), Tasseled Cap Transfor mation (TCT), Texture Analysis (TA), low pass filter, and edge detector analysis. Moreover, a set of non-spectral analysis layers were created in ArcGIS for use in the classificati on process. These are a Digital Elevation Model (DEM) at 30 m spatial resolu tion which was obtained from Fl orida Geographic Data Library (FGDL), Distance to home, Distance to roads, and distance to school (Figure 2-6). These distances were measured by using ArcGIS Spatia l Analyst tool; Euclidean Distance. This tool measures the distance between each pattern and it surrounding pattern and it used to relate land covers to this distance, for example, it measures the distance between all major roads in Alachua County. Normalized Difference Vegetation Index The Nor malized Difference Vegetation Index (N DVI) is a radiometric measure of remotely sensed imagery that designates activity and relative abundance of h ealthy vegetation, which includes percentage of greens cover, leaf area index, green biomass, chlorophyll content, and absorbed photosynthetic activ e radiation (APAR) (Jenson 2000). NDVI has been widely employed in remote sensing research to indicat e vegetation classes from other land use and land cover classes. NDVI analysis was performed base d on the calculation of red and near infrared (NIR) bands: NDVI = NIR-RED/NIR+RED

PAGE 24

24 NDVI calculation values vary from -1.0 to 1.0. The lower value signifies no vegetation, while the higher value signifies green and healthy vegetation. In order to differentiate vegetation classes from non-vegetation classes in this study, NDVI layers were created for each image date. Texture analysis Texture analysis techniques are frequently us ed to indica te the spatial information of diverse object classes into a cl assification. Many resear chers have revealed that structural and spatial information can lead to momentous improvements in the accuracy assessment of built up areas classification (Hsu 1978, Gong and Howarth 1990, Marceau et al 1990, and Puissant et al 2005). Texture analysis outputs may be used as a method to classify an image or as additional data layer that could be used in the classification. Texture an alysis incorporates information based on the surrounding pixels, so it is involved with spectrally based data. This analysis method has proven to be especially useful for urban and suburban area differentiation (Puissant et al 2007). Texture analysis has several optimal windows sizes such as 3 x 3, 5 x 5, and 13 x 13. The optimal window size determines the amount of spa tial information. If the window size is very small, inadequate spatial information is extrac ted to characterize a spec ific land cover, while large windows may overlap two types of land cove r resulting in erroneous spatial information. For that reason, image resolution and the content within the image are used to identify window size for texture analysis (Puissant et al 2007). In this study, textur e analysis was performed on all images using 5 x 5 windows based on resoluti on and the content w ithin the images. Low pass filter The low pass filter was used to atten uate noise and boundary errors, and to enhance features in the remote sensing imagery, by pr oducing new brightness valu es after blocking the high spatial frequency detailed, this filter introdu ces useful spatial frequency information that

PAGE 25

25 could be extracted from the im age (Jenson 2000). Low pass filter has different size of windows, such as 5 x 5 and 7 x 7. The smaller size windows are usually used to distinguish small features from their surrounding area. On the other hand, la rge size windows are used to eliminate small features in order to recognize large features. These windows are used to generate the average of pixel values in the box for each pixel value in the image (USGS 2007). Yang and Liu (2005) used a 3 x 3 filter to decrease the difficulty of the classification process. They performed 3 x 3 filter to reduce classes boundaries error due to the occurrences of intra-pixel spectral mixing and signal noise. Since this study used remotely sensed data that we re obtained from Landsat TM and ETM with a spatial resolution of 30 x 30 m, a 5 x 5 filter was used in all images. Using a 5 x 5 filter improved the ability of detecting small feat ures within the study area and to reduce noise and boundary errors. Tasseled cap transformation Tasseled Cap Transformation (TCT) is a ty pe of principle com ponents analysis. It transforms image bands into brightness (to indicate urban and soil), greenness (to indicate vegetation), and moisture or wetness. Accord ing to Jensen (2000) the brightness band in Tasseled Cap Transformation is used to identify urbanization areas which are particularly evident in this band. The greenness band is an impor tant source which provides information about vegetation. Moisture status of the wetland information presents in the wetness band. TCT could be helpful for use anywhere to disaggregate th e quantity of soil brightness, vegetation, and moisture content in independent pixels in satellite imagery (Jense n 2000). Since this study attempted to detect changes that occurred in land use such as urba n and land cover such as forest, using this kind of analysis would be helpful to improve the classification accuracy. Tasseled Cap Transformation analysis was performed on all im ages for use in suburban area discrimination.

PAGE 26

26 Thermal band The therm al band was extracted from all images for use in the classification process in order to improve the accuracy. According to Jensen (2000), this band calculates the quantity of infrared energy that is released from the earth s surface and it is practical for locating geothermal activity. Using the thermal band as layer data coul d be helpful to improve classification accuracy by indicating the temperature of each land use a nd land cover classes. Lu and Weng (2006) in their study of impervious surface in urban land-u se classification in Indianapolis/Marion County, Indiana, USA, used thermal band as input layer to classify urban land use classes, in order to decrease the number of mixed pixels. Built structures have many different thermal properties than vegetation area, so using th ermal band can be used to discriminate across the classes. As a result, including the thermal band in the classifica tion as a data layer may help to differentiate between suburban and vegetation areas. Non-spectral data layers Other, non-spectral data layers were also incorporated. A Digital Elevation Model (DEM) at 30 m spatial resolution was obtained from Fl orida Geographic Data Library (FGDL) and used as a data layer for the rule-based classification process. Using a DEM could be useful to identify the association between land use and land cover cla sses with the elevation of the landscape. This data layer could be used to diffe rentiate certain land use types fr om others. Moreover, distance to home, distance to roads, and distan ce to school data layers were also created for the rule-based classification. The input data fo r their creation (homes, public sc hools, and roads) were obtained from the Department of Growth Manageme nt in Alachua County as GIS layers. According to the Department of Growth Management in Alachua County (2008), homes layer is a dataset that contains locations and information pertai ning of homes, public school layer contain all of the Alachua county public school, and roads layer is contain all of the Alachua

PAGE 27

27 county roads. Distance to home layer was crea ted from a GIS raster layer which highlights homes in Alachua County, by using ArcGIS Spatial Analyst tool; Euclidean Distance. This tool measures distance from the center of a source cel l to the center of surro unding cells, the outputs of this tool is a raster of floating-point distance values (Arc GIS Desktop help). In other words, this tool measure the distance between each pattern from the surrounding one. For example, it measures the distance between Gainesville hi gh school and Waldo high school, this distance could be useful to indicate urban or suburban classes that exits in this range. Using this tool for homes, major roads, and school could be practical to indicate the influence of these land use patterns within the study region (Southworth et al. 2002), in order to im prove the classification accuracy. Rule-based classifier The study employed GIS and data m ining software to generate classifica tion rules that will be used for knowledge based classification. ArcMap and COMPUMINEs Rule Discovery Systems (RDS) were used to create sets of rule s to classify each land use and land cover class based on Boolean logic (if/then). All spectral and non-spectral data layers that were created in ERDAS Imagine and ArcGIS were imported into ArcMap software. Training samples for each land use and land cover class were also imported into ArcMap as GIS layers. The data layers, which included spectral and non-spectral informa tion, were associated w ith training samples as corresponding values for all variables. To do so, the extraction tool in ArcMap Spatial analyst tool box was used to parse out each training sample with their associated values from all data layers. Figure 2-7 shows a sample of the extracted training samples with their values. This table was imported into COMPUMINENs software to determine the appropriate model that will be used to create a set of rules for rule-based classification.

PAGE 28

28 COMPUMINENs Rule Discovery System (RDS) is a data-mining software that classifies information based on RDS criteria. This software is employed to create a tree model that will be used in rule crea tion. Using the n-fold cross validation method, ten tree models were generated for all spectral and non-spectral data layers. Each model built a set of rules from a number of variables and each model of these te n tree models has an assigned accuracy. This process was performed many times in order to ob tain the highest degree of accuracy. Figure 2-8 shows the process of creating models that have acceptable accuracy in COMPUMINENs software. Figure 2-9 shows ten tree models with thei r rules number and accuracy, based on accuracy one of them will be used to create the classifi cation rules. Accuracy of model number nine was the highest, 81.905(%). Since improving final classi fication accuracy assessment is one of the study aims, model number nine was used to create th e classification rules. This model has fifteen rules that were developed fr om a set of spectral variables; The Normalized Difference Vegetation Index (NDVI), Tassele d Cap Transformation (TCT), Texture Analysis (TA), and the low pass filter. The results of model number nine were presen ted as tree diagrams which determined the rules that will be used for rule -based classification. These diag rams included nodes and branches for each class (Figure 2-10). Each node introduces detail about one particular class with their values. For example, water node indicates that any pixel = 3.5 in the tasseled cap (band 1), 2 in the supervised images, <= 9.2566665 in the textur e (band 1), and <= 43.5 in the low pass (band 3) will classify as water. Tr anslating model node and branch into rules, which will be transformed to language for ERDAS Imagine, wa s performed for all modes and branches. Table 2-2 shows fifteen rules that will be used in the decision based classification.

PAGE 29

29 Multiple rules were created for different land use land cover classes from the optimal tree model. These rules were transfer red to the Knowledge engineeri ng classification tool in ERDAS Imagine processing software. Transferring rules to the Knowledge engineering classification tool is an essential step to translate final rules into hypotheses with their value to classify all images. Figure 2-11 shows a set of logical statements that were used to transfer the fifteen rules to Knowledge Engineering Classification tool as h ypotheses. After transfe rring all rules to ERDAS Imagine format the rule-based classi fication was performed on all images. Change Detection Analysis Change detection techniques have been used widely in land use and land cover changes research. Change detection analysis is a procedur e of detecting changes that occur in particular areas during a period of tim e. In order to determine the changes in land use and land cover that occurred in Alachua County from 1993 to 2003, cha nge trajectories were created. This technique requires the acquisition of remote ly sensed images for the same area over two or more time periods (Southworth et al 2002). In this study three images were classified by performing decision based classification, and th en incorporated into a change analysis. Based on these three images, change trajectory was implemented to indicate changes occurring across each image date, by incorporating these changes into a single change image. With three images and five land use and land cover classes, 125 possible cha nge trajectories cla sses are feasible, and interpretation can get confus ing (Southworth et al 2002). These all trajectories were not realistically occurring in the study area so expert opinion was also used to find the likely and acceptabl e options. Since the popul ation growth and the cities expansion in Alachua County increased last ten years, the la nd cover conversion is generally taking a shape of land transformation from vegetation toward urban and suburban areas. The 125 possible change trajectories were grouped to minimize confusion because

PAGE 30

30 showing all trajectories visually is very difficult especially ma ny of them are spatially limited and meaningless. For example, urban does not c onvert to water in area that demands more urbanization area. To eliminate such confusion some of change trajectories grouped to highlights the dominant and meaningful conversion. Results Accuracy Assessment Accuracy as sessment is considered one of mo st important elements of the classification process because without it the classification would be worthless. The Kappa coefficient assessment method was performed for all images. Th is method is defined by Lu et al (2005) as a calculation of overall sta tistical agreement of a matrix. Aerial photos were take on December 1994 were used to maximize the number of point for the accuracy assessment in order to distribute the training samples trough the study area w ithin ten years peri od. Training samples collected from fieldwork and aerial photo interpretation were randomly divided into two sets. The first set contained 70% of to tal training samples, which were used to create the initial classifications. The other set contained 30% of to tal training samples, whic h were held back and then used for accuracy assessment and validation. The initial, traditional supe rvised classification which was performed for the study area, resulted in unacceptable accuracy levels (Table 2-3 and Figure 2-13) and so knowledge based classification was used to attempt to improve upon these initial, inferi or results. Performing knowledge based classification increased the accur acy average from 50% to 80 % (Tables 2-4, 25, and 2-6). The accuracy assessment measures that were used in this study included overall accuracy, users accuracy, producers accuracy, and the KAPPA statistic. In the 2003 image the overall Kappa stat istic was 0.7898 and overall classification accuracy was 85.18%, which means the overall ac curacy improved by 35%. In the 1998 image

PAGE 31

31 the overall Kappa statistic was 0.8735 and overa ll classification accuracy was 90.47%, which means the overall accuracy improved by 40%. In the 1993 image the overall Kappa statistic was 0.6903 and overall classification accuracy was 76.12%, which means the overall accuracy improved by 25%. The study established that knowledge based classification improved the accuracy of the classification significantly. Change Trajectories Since indicating the expansion of urban and su burban areas was one of m ain goals of the study, pasture and forest were combined to vege tation. This produced a cl assification with four land use and land cover classes; water, vegetati on, urban, and suburban. In change trajectories detection for four land use land cover classes (1993-1998-2003), 64 possible trajectories may take place as land use and land cover changes. Th is study used change trajectories detection to identify land use and land cover changes that tr ansformed from one to another type. In order to minimize trajectories for the study area, only change s that present more than one percent of the study area will be discussed (Table 2-7, Figures 2-14 and 2-15). Water land cover was experienced some potential change s due to the increase of the monthly precipitation in 1998 and decrease in 2003. These changes in water land cove r were less than one pe rcent of the study area, as a result, these changes will not be discussed. Change trajectory tables i ndicate 38.27% of the study area occupied by vegetation area (Veg-Veg-Veg) across th e three time periods (1993-1998 -2003). While, suburban area (SubSub-Sub) that remains without changes since 1993 to 2003 occupied 23.53% of the study area, urban area occupied just 1.19% within same periods of time. According to table 7, 6.06% of the study area was covered by water class (Water-W ater-Water) from 1993 to 2003 (Figures 2-16). In 1998, 2.29% of the total area was transfor med from vegetation (Veg-Veg-Urban) to urban area. The trajectory determined that 1.89 % of the total area was transformed from

PAGE 32

32 vegetation (Veg-Sub-Sub) to suburban in 1993. Meanwhile, 1.69% of the total area transformed from urban (Urban-Urban-Sub) to suburban in 1998, 4.77% of the total area transformed from suburban (Sub-Sub-Urban) to urban area in 1998. 1.05% of the study area occupied by vegetation area was transformed to suburban area in 1998 (Veg-Sub-Veg) then went back to vegetation in 2003. These changes in land use a nd land cover in Alachua County (Figure 2-17) indicates the area of land use and land c over changes across the study area from 1993 to 2003. Even though knowledge based classification improved the accuracy assessment in all images, 17.51% of the total area transformed fr om vegetation to suburban in 2003 (Veg-VegSub) and 1.05% from suburban to vegetati on (Veg-Sub-Veg) in 1998, which lead to misclassification between vegetatio n and suburban. While there was su rely an increase in this suburban class related to the process of development, 17% seems excessive. Han (2002) demonstrated in his research an examination of historic land use ch anges in Alachua County, Florida: a technological approach that there are decrease in overa ll residential and increase in government property from 1992 to 2003, so from th ese results the high amount of vegetation converted for suburban are not acceptable for th e study area. In addition, the confusion between these classes relates to the fact that many suburba n developments in Gainesville, do indeed have a large amount of vegetative cove r (Figure 2-18), so the spectral signatures of these classes are quite close. A vegetated area with some bare so il, may be very similar to heavily vegetated suburban area with some built structures (built and soil are both low NDVI and greenness, high in reflectance and brightness, and so similar in these respects). To eliminate this potential misclassification a third classifi cation was undertaken, a supervis ed classification with only; water, vegetation, and built as cl asses. Built includes urban and s uburban structures (buildings, roads, and paving), which are not mixed with vege tation, so this can be distinguished from this

PAGE 33

33 confusion class of suburban here in Gainesville that is we break out a suburban class into built verses vegetated, when possible Note that this problem is not just particular to Gainesville, but it is especially problematic here, due to th e almost complete canopy closure in many older residential neighborhoods making them appear as forest from above ( although there fully vegetated, full canopy suburban developments, will st ill appear an vegetated as the satellite sees the trees). New supervised classifications were performed on all images for these new classes; water, vegetation, and built areas. Water as land cover, consists of an area that is inundated by water such as rivers, lakes, and wetland. Vegetation is area containing different species of trees and bushes with 25% canopy closure. Built up class includes urban and suburban types, which are not mixed with vegetation. Accuracy assessm ent was performed for the new supervised classification images (Table 2-8, 2-9, and 210). Overall Classification Accuracy was 98.24% and Overall Kappa Statistics was 0.9689 in 1993 image. In 1998 the Overall Classification Accuracy was 98.57%, while Overall Kappa Statis tics was 0.972. Table 2-10 shows that Overall Classification Accuracy was 98.24% and Overa ll Kappa Statistics was 0.9689 in 2003 image. Change trajectories were created for the ne w simplified supervised classification images. Since these supervised classification images (1993-1998-2003) are for three land cover only 27 change trajectories are possible as land use a nd land cover changes. In order to minimize trajectories for the study area, only changes that present more than one percent of the study area will be discussed (Table 2-11 and Figure 2-19). The change trajectory indicates 83.64% of the study area occupied by vegetation area (Veg-V eg-Veg) across the three time periods (19931998-2003), although we know this includes many s uburban areas now classified under this option and with > 25% canopy closur e. This is unavoidable however due to their appearing as

PAGE 34

34 forest from above. A GIS layer of suburban deve lopments would really be the only option to truly differentiate there areas as this research has clearly shown they are not differentiable spectrally, and this is important in highlighti ng the need for adaptable land use and land cover classes, as illustrated in this research From 1993 to 2003 built up area (Built-Built-Built) is occupied 1.95% of the total st udy area. According to table 11, 5.54% of the study is area covered by water class (Water-Water-Water) from 1993 to 2003. In 1998, 4.97% of the total area was transformed from vegetation (Veg-Veg-Built) to urban area. The trajectory determined that 1.07% of the total area was transf ormed from vegetation (Veg-Bu ilt-Built) to built between 1993 and 1998. This new classification gives a much better accounting of the changes in built and urban areas, but we can no longer differentiate between vegetation and most of the suburban classes due to the canopy closure issues. In this resear ch multiple methods were utilized in order to understand the dominate changes in land use and land cover. A single me thod did not suffice. This has obvious implications for this region, and multiple methods must be used to fully evaluate the changes in land cover and land use he re in Alachua County, as well as in many other suburban regions across the U.S. Discussion Summary of Finding The objectives of this study were to detect changes in land use and land cover in Alachua County from 1993 to 2003 and to find the best cla ssification m ethod for the study area. Different classification methods were used (supervised and knowledge based classification) to determine changes in land use and land cover that occurred in the study area. The results of these two classification proved that Alachua County is faci ng land transformation in terms of cover. This transformation tends to relate to the increase of population across the study area and particularly

PAGE 35

35 in Gainesville city limits which is the most populated area in the county. Gainesville experienced major population growth in Alachua County due to the expansion of the University of Florida, Shands and the popularity of Florida as a location. The results of the classifications indicate th at 4.97% of the total area transformed from vegetation to built areas by 2003. Moreover, 1.07% of the total area transformed from vegetation land cover to built area by 1998. This transformation is mostly locate d in the central and western portions of Alachua County, wh ich supports trend of popula tion growth requiring more urbanization. Such changes in land use and land cover relate to the population increase, which stimulates land owners to transform vegeta tion areas to built areas often via suburban developments. In Gainesville, much of this expa nsion has occurred in the west of the region as Gainesville has expanded out a long Archer Rd, Newberry Rd, and 39th street. Much of the new developments include cleaning of previously vegetated land and the building of suburban developments. These new non-vegetated developm ent may will change when homeowners plant trees and eventually such neighborhoods may be si milar to the older subu rban developments in Gainesville, which appear fr om above as forest. Implication of Study Determ ining land use and land cover change in Alachua County from 1993 to 2003 for four classes; water, vegetation, urban, and s uburban, was the main objective of this study. As a result, two different classificati on methods were used to determine optimum results for the study area. Supervised classification was the first method that was performed for the study area remotely sensed images. The results of this me thod were not accurate enough for remote sensing research. In order to increase the accuracy of the cl assification, knowledge based classification was performed for all images. Th e results of the classification were accurate when compared with supervised classification outputs. Even though, knowledge based classification improved

PAGE 36

36 the accuracy assessment in all images, 17.51% of th e total area transformed from vegetation to suburban in 2003, which led to misclassification between vegetation and suburban. While there was surely an increase in this suburban class related to the process of development, 17% seems excessive. This results led to the fact that suburb an class is hard to diffe rentiate from vegetation class in area like Alachua County because the confusion between thes e classes relates to the fact that many suburban developments, do indeed have a large amount of vegetative cover. Even using advanced RS classificati on techniques suburban would not be fully distinguished, so an additional classification of water, built, vegetation was created to give better understanding of overall change trends and to remove the suburban class, which proved to be combination of built (urban) and vegetation, so create d problems of spectral seperbility across these classes ( urban, suburban, vegetation). Hence, in order to fully understand this region a simplified classification of land use land cover classes (built, vegetation, a nd water) had to be created with more detailed land cover classification scheme obtained via decision tree classifi ers. Only when we used both classification results together could we fully comprehend and interpret the changes in land use and land cover classes in Alachua County. Changes in land cover in Alachua County may relate to the county government policies. The city of Gainesville government properties such as services building in creased especially in the downtown area in order to provide more public services (Hang, 2002). The government policies demand more urbanization areas which re sults more changes in land use and land cover areas. Transportation played a role in the trends of land use and land cover in the study area. Zhou and Kockelman (2006) in their study of neighborhood impacts on land use change in the City of Austin, Texas, found that the use of different models proved that there was a relation between land use changes and the distance to the nearest highway. In Alachua County the

PAGE 37

37 commercial property uses increased in the we stern part and it extended along US 441. US 441 magnetized commercial activities which makes it act now as a commercial corridor (Hang, 2002). Supervised classification techniqu es have been widely used in remote sensing research to determine changes in land cover classes. This traditional technique would be a powerful classification technique for classes that have un ique spectral signatures, in other words, for classes that could be easily discriminated, e.g., water from urban. Some scholars have incorporate supervised classificat ion with different analysis tec hniques to study land use and land cover change. Southworth (2004) used supervis ed classification in the Yucatan, Mexico, to successfully differentiate forest growth based on the earths surface temperature, from Landsat band 6 data. In areas like Alachua County, it is hard to differentiate between suburban and vegetation due to the similarity of both spectral signatures, which makes using supervised classification to study three classes reasonable. Due to the difficulties of classifying suburban and urban areas, several researchers have created new techniques to eliminate the misclassification of mixed pixels in remotely sensed images. Zhang and Foody (1998) used a fuzzy classification approach to evaluate the clas sification for suburban land cover, in the city of Edinburgh, UK. They used two remotely sensed images (Landsat TM and SPOT HRV) for an area of 2 km2, located within the city of Edinburgh. The study area co ntained suburban structures such as residential and commercial and its rich with geographical diversity and appropriate for fuzzy classification. They found that the Kappa coefficient (an indica tion of classification accuracy) more than doubled when they applied the fuzzy technique (i.e., this allows a pixels to belong to more than one classification, so suburba n can be a member of both built and vegetated classes). Lu and Weng (2006) in their study of impervious surface in urban land-use

PAGE 38

38 classification in Indianapolis/M arion County, Indiana, USA, used land surface temperature data from Landsat ETM+ and population density to impr ove the classification of five urban land use classes. They extracted land surface temperature to get better impervious surface mapping, then combined this information with population density to improve the classifi cation of their land use classes (Lu and wand, 2006). The study found that the integration of surface temperature provided substantially improved impervious surface coverages, which was reflected in the classification accuracy results. Mesev (1998) in a study of census data in urban image classification, in the United Kingdom used populati on sensed data for supervised classification (linking of urban class from remotely sensed da ta with urban functiona l characteristics from population sensed) in order to improve the classi fication accuracy of ur ban classification. While, Pellizzeri (2002) used Landsat TM and RADARSAT images to decrease the misclassification for suburban area in their study in Northern Italy. Gluch (2002) merged TM and SPOT-P data to monitor urban growth of Salt Lake City in Uta h. Other classification methods have implemented in many researches to eliminate the misclassifica tion in urban and suburban areas. This validates the combined methodologies used in this analys is as appropriate, over traditional classification techniques, which many resear chers have found equally problematic based on the surface types found in urban and suburban regions (Gluch 2002, Lu and wand 2006, Mesev 1998, Pellizzeri 2002, Southworth 2004, Zhang and Foody, 1998). Knowledge based classification techniques have been used recently to determine land use and land cover changes especially for urban regions (Chien and Chou 2000, Haild 1997, Onsi 2003). This technique classifies remotely sensed data base d on spectral and non-spectral information. Some land cover classes have simila r spectral signatures; this technique could differentiate these classes from each other. Comparing knowledge based classification with

PAGE 39

39 traditional classification techniques, knowledge based techniques have given better results than other techniques. Daniels ( 2006) has proven that knowledge based techniques improved the accuracy of classification for tropical land cove rs relating to the results of supervised classification. Moreover, knowledge based techni ques implemented in several studies have addressed specifically changes that have o ccurred in urban areas (Chien and Chou, 2000). Janssen (1992) used knowledge based classifica tion techniques to improve crop classification accuracy. This technique was also used to optimize the C-factor mapping in Spain by Folly et al, (1996). Haild (1997) use knowledge based approaches to link local knowledge, field data and the spectral land cover classes to indicate changes in urban land use classes. Knowledge based classification techniques pe rformed in this study greatly improved the classification accuracy. Spectral an d non-spectral data were imported into a data-mining software in order to assign the best rules for knowledge based classification. Based on the output model of this software fourteen spectral rules were us ed for knowledge based classification. The accuracy assessments of the knowledge based classification were acceptable for remote sensing research; however, it didnt eliminate th e confusion between suburban and vegetation. As a result, knowledge based classification techniques could not solve the spectral co nfusions between land use and land cover classes. This study demonstrat es that the spectral signature of a suburban class (a combination of roads, building and vege tation) in Alachua County and elsewhere, is too close to the vegetation and urban class signatures (it is in fact a merged class of these two components), even when using advanced knowledge based classification techniques. For future research more fieldwork to gather explicit GIS layers of suburban boundaries would help clarify the classification process and would help diffe rentiate suburban from vegetation and urban. Using finer spatial resolution sate llite images would also be help ful for better classification, as

PAGE 40

40 this would break down the pixels from mixed c overs, to individual components, e.g. Quickbird, 0.4 m spatial resolution panchromatic data although this is very expensive. In summary, this study found that 4.97 % of to tal area of Alachua County converted from vegetation to built area in 2003. These changes in land use and land cover have taken place mostly in central and wester n portion of Alachua County. In 1998 1.07 % of the total area converted from vegetation to built area. Generall y, the changes located in the central and western part of the study area. Gainesvi lle which is the largest city in Alachua County has the most changes that occurred in the count y. Newberry, which is located in th e west of Gainesville, is the second city that has changes after Gainesville. Changes in these cities ar e reflecting the increase in population and the expansion of University of Florida in Gainesville Transportation affects the land cover conversion esp ecially around US 441 highway. These results could be used by the county pl anner as reference for future estimation. However suburban misclassification could be controlled by developing expertise and methodologies in order to monitor changes in suburban area in Alachua County. Knowledge based classification have been proven in severa l research studies (Chi en and Chou 2000, Daniels 2006, Janssen 1992, Haild 1997, Onsi 2003 ) as an advanced technique for land use and land cover changes, although it was not fully successful in this stu dy, which reveals that no single method is appropriate for all cover types. This study demonstrated changes in land use and land cover in Alachua County and these changes have to be monitored for better planning. Suburban area in Alachua County, like elsewh ere, needs advanced analysis techniques and field work in order to eliminate the misclassificati on between suburban a nd vegetation areas. Conclusion In conclusion, this research found that m ost of the changes in land use and land cover in Alachua County have taking a pla ce in western and central portions of the County. Land use and

PAGE 41

41 land cover change in Alachua County have take n two major trends, from vegetation to built structures area and from built to vegetati on area. In 1998, 1.07 % of the total area have transformed from vegetation to built structures area. In 2003, the number of the total area that have transformed from vegetation to built area was 4.97 %. In 1998, 2.23 % of the total area transformed from built structures area to vegeta tion. Most of these changes in land use and land cover occurred in the city of Gainesville area and in the western portion of Alachua County. A traditional classification (supervised clas sification) and an advanced classification (Knowledge based classification) methods were us ed in this research in order to determine changes in land use and land cover in Alachua County. The study area likew ise other areas that have a large amount of vegetati on cover in suburban development, which is causing enormous confusion between suburban and vegetation. As a result, Knowledge based classification method could not solve the spectral conf usions between land use and la nd cover classes (suburban and vegetation), even though the accuracy assessments of the knowledge based classification were acceptable for remote sensing research. Better methods need to be developed to monitor suburban developments over time in Alachua C ounty such as using GIS layers for suburban developments could help to improve the abilit y of differentiating suburban from vegetation.

PAGE 42

42 Table 2-1. Land use and land cover desc riptions for classification scheme. Name Class Type Description Water Land cover Lands inundated with wa ter such as lakes, rivers, and wetland. Forest Land cover Lands cover with different species of trees and bushes with 25% canopy closure. Pasture Land use Pasture composed of agricultural land, pasture, grassland, and others. Urban Land use Urban composed of 70-100% built materials such as airport and shopping mall. Suburban Land use Suburban composed of 4070% built materials which are mixed with vegetation such as Residential area.

PAGE 43

43Table 2-2. Final decision based rules used in the final classification procedure. Class type Rule criteria Values Description Water Rule.1 Tasseled Cap 3 Supervised Texture 1 Low Pass 3 = 3.5 2 <= 9.2566665 <= 43.5 Classify as Water all water pixels th at = 3.5 in the tasseled cap (band 1), 2 in the supervised images, <= 9.2566665 in the texture (band 1), and <= 43.5 in the low pass (band 3). Forest Rule.2.A Band 6 Supervised Texture 1 <= 107.5 = 2 <= 9.2566665 Classify as Forest all fo rest pixels that <= 107.5 in the thermal band, = 2 in the supervised images, and <= 9.2566665 in the texture (band 1). Rule.2.B Band 6 Band 6 Supervised Texture 1 > 107.5 <= 108.5 = 2 <= 9.2566665 Classify as Forest all fo rest pixels that > 107.5 <= 108.5 in the thermal band, = 2 in the supervised images, and <= 9.2566665 in the texture (band 1). Rule.2.C Band 6 Supervised Texture 1 > 108.5 = 2 <= 9.2566665 Classify as Forest all fo rest pixels that >108.5 in the thermal band, = 2 in the supervised images, and <= 9.2566665 in the texture (band 1). Rule.2.D Tasseled Cap1 Texture 1 Low Pass 4 <= 114.5 > 9.2566665 <= 17.5 Classify as Forest all fo rest pixels that <= 114.5 in the tasseled cap (band 1), and > 9.2566665 in the texture (band 1), and <=17.5 in the low pass (band 4). Pasture Rule.3.A Supervised Texture 1 Low pass 3 2 <= 9.2566665 > 43.5 Classify as Pasture a ll pasture pixels that 2 in the supervised images, <= 9.2566665 in the texture (band 1), and > 43.5 in the low pass (band 3). Rule.3.B Tasseled Cap1 Supervised Texture 1 Low pass 3 > 114.5 4 > 9.2566665 > 50 Classify as Pasture all pa sture pixels that >114.5 in the tasseled cap (band 1), 4 in the supervised images, > 9.2566665 in the texture (band 1) and > 50 in the low pass (band 3).

PAGE 44

44Table 2-2 Continued. Class type Rule criteria Values Description Rule.3.C Tasseled Cap3 Supervised Texture 1 Low pass 3 <= 3.5 2 <= 9.2566665 <= 43.5 Classify as Pasture all forest pixels that <= 3.5 in the tasseled cap (band 3), 2 in the supervised images, <= 9.2566665 in the texture (band 1), and > 43.5 in the low pass (band 3). Urban Rule.4.A Tasseled Cap1 Supervised NDVI Texture 1 114.5 = 4 <= 0.1286489 > 9.2566665 Classify as Urban all urban pixels that >114.5 in the tasseled cap (band 1), = 4 in the supervised images, <= 0.1286489 in the NDVI image, and > 9.2566665 in the texture (band 1). Rule. 4.B Tasseled Cap1 Supervised NDVI Texture 1 > 136 = 4 > 0.1286489 > 9.2566665 Classify as Urban all urban pixels that >136 in the ta sseled cap (band 1), = 4 in the supervised images, > 0.1286489 in the NDVI image, and > 9.2566665 in the texture (band 1). Rule .4.C Tasseled Cap1 Tasseled Cap1 Supervised NDVI Texture 1 114.5 < 136 = 4 > 0.1286489 > 9.2566665 Classify as Urban all urban pixels that >145 in the tasseled cap (band 1), <136 in the tasseled cap (band 1), = 4 in the supervised images, > 0.1286489 in the NDVI image, and > 9.2566665 in the texture (band 1). Suburban Rule.5.A Tasseled Cap2 Tasseled Cap1 NDVI Texture 1 Low pass 4 -24 <= 114.5 <= 0.3839777 < 9.2566665 > 17.5 Classify as Suburban, all suburban mixed w ith vegetation pixels that >-24 in the tasseled cap (band 2), < =114.5 in th e tasseled cap (band 1), <= 0.3839777 in the NDVI image, < 9.2566665 in the textur e (band 1), and > 17.5in the low pass (band 4). Rule.5.B Tasseled Cap2 Tasseled Cap1 Texture 1 Low pass 4 <= -24 <= 114.5 > 9.2566665 > 17.5 Classify as Suburban, all suburban mixed with vegetation pixels that <= -24 in the tasseled cap (band 2), < =114.5 in the tasseled cap (band 1), > 9.2566665 in the texture (band 1), and > 17.5in the low pass (band 4).

PAGE 45

45Table 2-2 Continued. Class type Rule criteria Values Description Rule.5.C Tasseled Cap1 Supervised Texture 1 Low pass 3 > 114.5 4 > 9.2566665 <= 50 Classify as Suburban, all suburban mixed with vegetation pixels that > 114.5 in the tasseled cap (band 1), 4 in the supervised images, > 9.2566665 in the texture (band 1), and <= 50 in the low pass (band 3). Rule.5.D Tasseled Cap2 Tasseled Cap1 NDVI Texture 1 Low pass 4 > -24 <= 114.5 > 0.3839777 > 9.2566665 > 17.5 Classify as Sub-urban, all suburban mixed with vegeta tion pixels that >-24 in the tasseled cap (band 2), < =114.5 in the tasseled cap (band 1), > 0.3839777 in the NDVI image, > 9.2566665 in the textur e (band 1), and > 17.5in the low pass (band 4).

PAGE 46

46 Table 2-3. Supervised classification 1998 accura cy assessment; overall accuracy, users accuracy, producers accuracy, and KAPPA statistic. Ref. = Reference, Suburb= Suburban Pro.acc = Producers accuracy, Clas.T ot = Classified total, and Users.Acc= Users accuracy. Supervised classification 1998 1998 Water Forest Pasture Urban Suburban Clas.Tot Users.Acc Water 24 0 0 0 0 24 1.00 Forest 1 113 1 14 32 161 0.70 Pasture 1 2 29 6 18 56 0.518 Urban 0 1 0 1 4 6 0.167 Suburb 0 2 28 76 62 168 0.369 Ref.total. 27 120 60 101 143 451 Pro.acc 0.888 0.941 0.483 0.99 0.433 Overall Kappa Statistics = 0.3559, Overal l Classification Accuracy = 50.78% Table 2-4. Rule-based classification 2003 accura cy assessment; overall accuracy, users accuracy, producers accuracy, and KAPPA statistic. Ref. = Reference, Suburb= Suburban Pro.acc = Producers accuracy, Clas.T ot = Classified total, and Users.Acc= Users accuracy. Rule-based classification 2003 2003 Water Forest Pasture Urban Suburban Clas.Tot Users.Acc Water 17 0 0 0 0 17 1.00 Forest 0 96 0 0 0 96 1.00 Pasture 0 9 22 0 0 31 0.709 Urban 0 0 1 85 14 100 0.850 Suburb 7 11 4 21 165 208 0.793 Ref.total. 24 116 27 106 179 452 Pro.acc 0.701 0.828 0.815 0.802 0.921 Overall Kappa Statistics = 0.7898, Overal l Classification Accuracy = 85.18%

PAGE 47

47 Table 2-5. Rule-based classification 1998 accura cy assessment; overall accuracy, users accuracy, producers accuracy, and KAPPA statistic. Ref. = Reference, Suburb= Suburban Pro.acc = Producers accuracy, Clas.T ot = Classified total, and Users.Acc= Users accuracy. Rule-based classification 1998 1998 Water Forest Pasture Urban Suburban Clas.Tot Users.Acc Water 25 0 3 0 0 28 1.00 Forest 1 117 0 0 3 121 0.955 Pasture 0 3 49 0 1 53 0.754 Urban 0 0 0 85 6 91 0.601 Suburb 1 0 8 16 132 157 0.836 Ref.total. 27 120 60 101 143 451 Pro.acc 0.926 0.975 0.817 0.842 0.923 Overall Kappa Statistics = 0.8735, Overal l Classification Accuracy = 90.47% Table 2-6. Rule-based classification 1998 accura cy assessment; overall accuracy, users accuracy, producers accuracy, and KAPPA statistic. Ref. = Reference, Suburb= Suburban Pro.acc = Producers accuracy, Clas.T ot = Classified total, and Users.Acc= Users accuracy. Rule-based classification 1993 1993 Water Forest Pasture Urban Suburban Clas.Tot Users.Acc Water 22 0 0 0 0 22 1.00 Forest 1 84 1 0 2 88 0.955 Pasture 0 1 46 3 11 61 0.754 Urban 0 0 7 98 58 163 0.601 Suburb 3 1 6 1 56 76 0.836 Ref.total. 26 86 60 102 128 402 Pro.acc 0.846 0.977 0.767 0.9600.438 Overall Kappa Statistics = 0.6903, Overal l Classification Accuracy = 76.12%

PAGE 48

48 Table 2-7. Change trajectories of land use a nd land cover classes of interest (1993-1998-2003). Sub= suburban and Veg= vegetation. Class type Number of classified pixels Percentage of area Area in square meters Water-Water-Water 156698 6.06 141028200 Veg-Veg-Veg 990243 38.27 891218700 Sub-Sub-Sub 608897 23.53 548007300 Urban-Urban-Urban 30705 1.19 27634500 Veg-Veg-Urban 59253 2.29 53327700 Veg-Veg-Sub 453099 17.51 407789100 Veg-Sub-Veg 27208 1.05 24487200 Veg-Sub-Sub 48850 1.89 43965000 Urban-Urban-Sub 43647 1.69 39282300 Sub-Sub-Urban 123541 4.77 111186900 Total 2587296 100 2328566400 Table 2-8. Supervised classification 1993 accura cy assessment; overall accuracy, users accuracy, producers accuracy, and KAPPA statistic. Ref. = Reference, Suburb= Suburban, Pro.acc = Producers accuracy, Clas.T ot = Classified total, and Users.Acc= Users accuracy. Supervised classification 1993 1993 Water Vegetation Built up Clas.Tot Users.Acc Water 30 1 0 31 0.967 Vegetation 0 186 6 193 0.969 Built up 0 0 174 174 1.00 Ref. total. 30 187 180 390 Pro.acc 1.00 0.995 0.967 Overall Kappa Statistics = 0.9689, Overal l Classification Accuracy = 98.24%

PAGE 49

49 Table 2-9. Supervised classification 1998 accura cy assessment; overall accuracy, users accuracy, producers accuracy, and KAPPA statistic. Ref. = Reference, Suburb= Suburban Pro.acc = Producers accuracy, Clas.T ot = Classified total, and Users.Acc= Users accuracy. Supervised classification 1998 1998 Water Vegetation Built up Clas.Tot Users.Acc Water 33 1 0 34 0.971 Vegetation 0 212 2 214 0.991 Built up 0 2 99 101 0.980 Ref. total. 33 215 101 349 Pro.acc 1.00 0.986 0.980 Overall Kappa Statistics = 0.9729, Ov erall Classification Accuracy = 98.57% Table 2-10. Supervised classification 2003 accur acy assessment; overall accuracy, users accuracy, producers accuracy, and KAPPA statistic. Ref. = Reference, Suburb= Suburban Pro.acc = Producers accuracy, Clas.T ot = Classified total, and Users.Acc= Users accuracy. Supervised classification 2003 2003 Water Vegetation Built up Clas.Tot Users.Acc Water 30 1 0 31 0.968 Vegetation 0 186 6 192 0.969 Built up 0 0 174 174 1.00 Ref. total. 30 187 174 397 Pro.acc 1.00 0. 995 0. 967 Overall Kappa Statistics = 0.9689, Over all Classification Accuracy = 98.24% Table 2-11. Change trajectories of land use and land cover classes of interest for supervised classification images (1993-1998-2003). Built= Built up and Veg= vegetation Class type Number of classified pixels Percentage of area Area in square meters Water-WaterWater 150092 5.54 135082800 Veg-Veg-Veg 2267954 83.64 2041158600 Built-Built-Built 52990 1.95 47691000 Veg-Veg-Built 134703 4.97 121232700 Veg-Built-Built 29009 1.07 26108100 Built-Veg-Veg 60528 2.23 54475200

PAGE 50

50 Figure 2-1. Map of study area.

PAGE 51

51 Figure 2-2. Gainesville monthly rainfa ll from February 1992 to March 1993. Figure 2-3. Gainesville monthly rainfa ll from January 1997 to February 1998.

PAGE 52

52 Figure 2-4. Gainesville monthly rainfa ll from February 2002 to March 2003.

PAGE 53

53 Figure 2-5. Aerial photos a nd training samples in cen tral part of study area.

PAGE 54

54 Figure 2-6. Spectral and non-spectra l data set analysis layers that created for rule-based classification, with DEM= Digital El evation Model; Dist= distance; NDVI= Normalized Difference Vegetation Index. Figure 2-7. Sample set of forest ex tracted training samples with their associated values from all data layers. NDVI Dist. To Home Dist. To School Dist. To Roads (DEM) Standard Classification GIS Layers (Non-spectral data) Low Pass (5x5) Tasseled Cap Edge Detector Thermal Band (band6) Texture (5x5) Supervised Classification

PAGE 55

55 Figure 2-8. Process of creating models that have acceptable accuracy in COMPUMINENs software, with DEM = Digital Elevation Model; Dist = distance; NDVI= Normalized Diff erence Vegetation Index. Compumine Data models Models accuracy NDVI Dist. To Home Dist. To School Dist. To Roads (DEM) Standard Classification Algorithms (Spectral data) GIS Layers (Non-spectral data) Low Pass (5x5) Tasseled Cap Edge Detector (5x5) Thermal Band (band6) Texture (5x5) Supervised Classification Training Samples

PAGE 56

56 Correct class Predicted class Method Number of rules Accuracy Total AUC Precision Recall AUC 1 2 3 4 5 Tree fold 1 11 73.585 0.913 1 0.500 0.143 0.913 1 0 6 0 0 2 0.857 0.643 0.935 1 18 8 0 1 3 0.412 1.000 0.958 0 0 14 0 0 4 0.885 1.000 0.979 0 0 0 23 0 5 0.957 0.647 0.833 0 3 6 3 22 Tree fold 2 14 73.585 0.922 1 1.000 0.429 0.899 3 0 4 0 0 2 0.952 0.714 0.946 0 20 8 0 0 3 0.343 0.857 0.858 0 0 12 0 2 4 1.000 0.913 0.987 0 0 0 21 2 5 0.846 0.647 0.890 0 1 11 0 22 Tree fold 3 14 70.755 0.910 1 1.000 0.571 0.935 4 3 0 0 0 2 0.500 0.964 0.937 0 27 0 0 1 3 1.000 0.143 0.842 0 11 2 0 1 4 0.950 0.826 0.921 0 3 0 19 1 5 0.885 0.676 0.902 0 10 0 1 23 Tree fold 4 11 77.358 0.927 1 1.000 0.429 0.939 3 0 4 0 0 2 1.000 0.714 0.965 0 20 8 0 0 3 0.467 1.000 0.944 0 0 14 0 0 4 0.759 0.957 0.973 0 0 0 22 1 5 0.958 0.676 0.854 0 0 4 7 23 Tree fold 5 15 71.429 0.916 1 1.000 0.167 0.861 1 0 5 0 0 2 0.792 0.655 0.919 0 19 9 0 1 3 0.371 1.000 0.908 0 0 13 0 0 4 0.958 0.958 0.963 0 0 0 23 1 5 0.905 0.576 0.894 0 5 8 1 19 Tree fold 6 15 80.952 0.953 1 1.000 0.500 0.944 3 3 0 0 0 2 0.614 0.964 0.955 0 27 1 0 0 3 0.857 0.429 0.943 0 8 6 0 0 4 0.955 0.875 0.936 0 2 0 21 1 5 0.966 0.848 0.969 0 4 0 1 28 Tree fold 7 12 76.190 0.928 1 1.000 0.500 0.947 3 0 3 0 0 2 0.895 0.607 0.901 0 17 9 0 2 3 0.448 0.929 0.889 0 0 13 0 1 4 0.955 0.875 0.971 0 0 0 21 3 5 0.813 0.788 0.934 0 2 4 1 26 Tree fold 8 13 78.095 0.963 1 n/a 0.000 0.914 0 0 6 0 0 2 0.955 0.750 0.971 0 21 7 0 0 3 0.433 0.929 0.927 0 0 13 0 1 4 0.913 0.875 0.986 0 0 1 21 2 5 0.900 0.818 0.965 0 1 3 2 27 Tree fold 9 15 81.905 0.964 1 1.000 0.500 0.929 3 1 2 0 0 2 0.923 0.857 0.979 0 24 4 0 0 3 0.500 1.000 0.956 0 0 14 0 0 4 0.913 0.875 0.964 0 0 2 21 1 5 0.960 0.727 0.961 0 1 6 2 24 Tree fold 10 13 80.000 0.960 1 1.000 0.333 0.923 2 0 3 0 1 2 1.000 0.679 0.951 0 19 6 0 3 3 0.500 0.929 0.949 0 0 13 1 0 4 0.955 0.875 0.996 0 0 0 21 3 5 0.806 0.879 0.953 0 0 4 0 29 Figure 2-9.Tree models with their number of rules and accuracy.

PAGE 57

57 Figure 2-10. Sample of tree model nodes and branches that used to create classification rules.

PAGE 58

58 Figure 2-11. Set of logic statements that tran sferred to Knowledge engineering classification tool.

PAGE 59

59 Figure 2-12. Flowchart of knowledge based classification methodology. Remote Sensing Images Standard Classification Al g orithms GIS Layers Training Samples Knowledge-Based Classification Final Classification Accuracy Assessment Compumine Data models Models accurac y Supervised Cl assification Low Pass (5x5) Tasseled Cap Thermal Band (band6) Texture (5x5) NDVI R u l es c r eat i o n

PAGE 60

60 Figure 2-13. First supervised cl assification approach of remote ly sensed image for Alachua County, Florida in 1998.

PAGE 61

61 Figure 2-14. Changes Trajectories chart classes of interest (1993-1998-2003). Sub= suburban and Veg= vegetation

PAGE 62

62 Figure 2-15. Knoweldge based cla ssification changes trajectories map classes of interest (19931998-2003). Sub= suburban and Veg= vegetation

PAGE 63

63 Figure 2-16. Map of stable la nd cover classes from 1993 to 2003.

PAGE 64

64 Figure 2-17. Map of land use and land cover changes from 1993 to 2003.

PAGE 65

65 Figure 2-18. Sample of suburban area in Alachua County.

PAGE 66

66 Figure 2-19. Changes Trajectories map classes of interest for supervised images (1993-19982003). Veg =Vegetation and Built = Built up.

PAGE 67

67 CHAPTER 3 CONCLUSION Determ ining what types of land use and land cover change have occurred in Alachua County from 1993 to 2003 and what type of classi fication method produces the best results were the main objectives of this resear ch. So, three different classificati on methods were used to detect the changes that occurred in th e study area. These three classification methods were performed for four classes; water, vegetation, urban, and suburban. The initial classification t echnique employed was a simple unsupervised classification. This was performed in 10, 20, and 30 classes for the study area in order to determine the classification scheme. Based on the number of cluster classes that revealed from the unsupervised classification fives land use and la nd cover classes created and to increase the classification accuracy some of these clustered classes regrouped. Second, a supervised classification methodology was used but this produced unacceptable levels of accuracy, due specifically to confusion across certain cla sses in the landscape. As such, he results of this type of classification were not acceptable for academic research because the accuracy was low. To increase the accuracy of the classification, knowledge based classification was performed for all images. This method increa sed the accuracy of a ll images, however there was 17.51 % of the total area transformed from vegetated area to suburban. This large area of land transformation indicated that there was mi sclassification between vegetation and suburban areas. This misclassification relates to the confusion between these tw o classes because many suburban developments include a large amount of vegetative cover and so spectrally the two are not dissimilar. Hence, a final methodology was empl oyed in which some classes were merged to decrease the confusion between suburban and ve getation and this produced three major classes built, water, vegetation.

PAGE 68

68 Alachua County, like other regions that have a large amount of vegeta tion cover within the suburban developments area, causes potential ly significant problems of misclassification between suburban and vegetation classes, for many developed regions in the U.S. Even using a more advanced image classification technique, such as the knowledge based classifier, the confusion between suburban and vegetation cl asses was not eliminated, and ultimately a traditional classification for water, built, and vegetation classes only was performed on all images. The accuracies of this classification were acceptable grea tly improved and it provides a better understanding of overall change trends. Fo r future work better methods need to be developed to monitor suburban developments and their changes over time, such as using a GIS layer for suburban development in order to impr ove the ability of differentiating suburban from vegetation. Moreover, using finer sp atial resolution satellite images would help in differentiating between suburban and vegetation. This research found that western and central portions of Alachua C ounty experienced most of the changes in land use and land cover. The changes in land use and land cover have taken two major trends, from built to vegetation area an d from vegetation to built structures. 1.07 % of the total area that have transformed from vege tation to built structures area in 1998. 4.97 % of the total area that have transformed from vege tation to built structures area in 2003. The total area that have transformed from built struct ures area to vegetation in 1998 was 2.23 %. The western portion of Alachua County and Gainesville as the largest and most populated area has experienced most of these changes in land use an d land cover. The outputs of this research could be used by the county planner as reference for fu ture estimation and to understand land use and land cover change trends in Alachua County.

PAGE 69

69 LIST OF REFERENCES BINFORD, M. W ., GHOLZ, H. L., STARR, G., and MARTIN, T. A., 2006, Regional carbon dynamics in the southeastern U.S. coasta l plain: Balancing land cover type, timber harvesting, fire, and environmental vari ation (DOI 10.1029/2005JD006820). Journal of geophysical research., 111, D24S92. BROCKERHOFF, M. P., 2000, An Urbani zing World. Population Bulletin, 55, 1. CLARK, C., 1967, Population growth and land use (London; Melbourne [etc.; New York: Macmillan; St. Martin's P.). CITY of GAINESVILLE., 2008, Gainesvi lle Facts. A vailable online at: http://www.cityofgainesville.org/about/ (accessed April 2008 ). DANIELS, A., 2006, Incorporating domain knowledge and spatial relationshi ps into land cover classifications: a rule-based approach. International Journa l of Remote Sensing, 27, 2949. DIGITAL MAPPING SYSTEMS., 2006.Glo ssary. A vailable online at: http://www.digimap.gg/glossary (accessed March 2008). DOHRENWEND, R. E., 1978, The clim ate of Alac hua County, Florida (Gainesville, Fla.: Agricultural Experiment Stations, Institute of Food and Agricultural Sciences, University of Florida). FOLLY, A., BRONSVELD, M. C., and CLAVA UX, M., 1996, A knowledge-based approach for C-factor mapping in Spain us ing Landsat TM and GIS. Inte rnational Journal of Remote Sensing, 17, 2401-2415. FU-JEN CHIEN, T.-Y. C., 2000, The Study of Knowledge-Based Database Assist for Urban Land Use Classification (100 WENHWA RD., TAICHING, TAIWAN: Feng Chia University ). GALBARD, A., and HAUB, C., Wo rld population beyond six billion. Population Bulletin, 54, 3. GONG, P., and HOWARTH, P. J ., 1992, Frequency-Based Contextu al Classification and GreyLevel Vector Reduction for Land-Us e Identification. PHOTOGRAMMETRIC ENGINEERING AND RE MOTE SENSING, 58, 423. GONG, P., and HOWARTH, P. J., 1990, An assessment of some factors intluencing multispeetral land-cover classification.: Photogrammetric Engineering & Remoie Sen.sing.), pp. pp. 597 603. GREEN, G. M., C.M. SCHWEIK, AND J.C. RANDOLPH. 2005, "Retrieving Land-Cover Change Information from Landsat Satellite Images by Minimizing Other Sources of Reflectance Variability". In Seeing the fo rest and the trees : human-environment interactions in forest ecosystems, edited by E. F. M. E. Ostrom (Cambridge, Mass.: Mit Press).

PAGE 70

70 HAACK, B. N., SOLOMON, E. K., BECHDOL, M. A., and HE ROLD, N. D., 2002, Radar and Optical Data Comparison/Integration for Urban Delineation: A Case Study. Photogrammetric engineeri ng and remote sensing., 68, 1289. HAILD, M., 1997, Land Use-Cover Change Detec tion Using Knowledge based approaches: Remote Sensing and GIS (50300 Jalan Tun R azak, Kuala Lumpur Malaysia Kalaysian Center for Remote Sensing (MACRES)). HAN, F., 2002, An examination of historic land use changes in Alachua County, Florida : a technological approach. HELP, A. D., 2005, ArcEditor (Redlands, CA: E nvironmental Systems Research, Institute (ESRI)). HEROLD, M., GOLDSTEIN, N. C. and CLARKE, K. C., 2003, The spatiotemporal form of urban growth: measurement, analysis and m odeling. Remote Sensing of Environment, 86, 286. HOUGHTON, R. A., The worldwide extent of land-use change. Bioscience, 44, 305. HSU, S.-Y., 1978 Texture-tone analysis for automated land-use mapping Photogrammetric Engineering and Remote Sensing.). INTERIOR, U. S. D. O. T., 2007, USGS Isis Workshop: Introduction to Filters: Astrogeology Research Program ). JANSSEN, L. L. F., and MIDDELKOOP, H., 1992, Knowledge-based crop classification of a Landsat Thematic Mapper image. Internat ional Journal of Remote Sensing, 13, 2827. JENSEN, J. R., 1996, Introductory digital image processing : a remote sensing perspective (Upper Saddle River, N.J.: Prentice Hall). LU, D., MAUSEL, P., BATISTELLA, M., and MORAN, E., 2005, Land-cover binary change detection methods for use in the moist tropi cal region of the Amaz on: a comparative study. International Journal of Remote Sensing, 26, 101. LU, D., MAUSEL, P., BRONDIZIO, E., and MOR AN, E., 2004, Change detection techniques. International Journal of Remote Sensing, 25, 2365. LU, D., and WENG, Q., 2006, Use of impervious surface in urban land-use classification. Remote sensing of environment., 102, 146. MACRI PELLIZZERI, T., 2003, Classi fication of polarimetric SAR images of suburban areas using joint annealed segmentation and "H/A/a" polarimetric decomposition. ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 58, 55-70. MAKTAV, D., ERBEK, F. S., and JURGENS, C., 2005, Remote sensing of urban areas. International Journal of Remote Sensing, 26, 655.

PAGE 71

71 MARCEAU. D.J.. HOWARTH, P. J., DTIBOIS, J.M.M. AND GRATTON, D.J, 1990, Evaluation of the grey-level co-occurrence matrix-method for land cover classiHcation using SPOT imagery. IEEE Transactions on Geoscience and Remoie Sensing., 28, 513519. MERTENS, B., and LAMBIN, E. F., 2000, LandCover-Change Trajectories in Southern Cameroon. Annals of the Association of American Geographers, 90, 467. MESEV, V., 1998, The Use of Census Data in Urban Image Classification. PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING, 64, 431-438. MEYER, W. B., and TURNER, B. L., 1994, Cha nges in land use and land cover : a global perspective (Cambridge [England]; New Yor k, NY, USA: Cambridge University Press). MEYER, W. B., and TURNER II, B. L., 1992, Human Population Growth and Global LandUse/Cover Change. Annual Review of Ecology & Systematics, 23, 39. ONSI, H. M., 2003, Designing a rule-based classifier using syntactical approach. International Journal of Remote Sensing, 24, 637-647. PETIT, C., SCUDDER, T., and LAMBIN, E., 2001, Quantifying processes of land-cover change by remote sensing: resettlement and rapid land-cover changes in south-eastern Zambia. International Journal of Remote Sensing, 22, 3435. PHILLIPS, S., 1986, Factors Influencing Agricu ltural Land Use Change : A Study of Alachua County, Florida.: University of Fl orida., Gainesville, Florida. PUISSANT, A., HIRSCH, J., and WEBER, C., 2005, Th e utility of texture analysis to improve per-pixel classification for high to very hi gh spatial resolution imagery. International Journal of Remote Sensing, 26, 733. SOUTHWORTH, J., 2004, An assessment of Landsat TM band 6 thermal data for analysing land cover in tropical dry forest regions. International Journal of Remote Sensing, 25, 689. SOUTHWORTH, J., NAGENDRA, H., CARLSON, L. A., and TUCKER, C., 2004, Assessing the impact of Celaque National Park on forest fragmentation in we stern Honduras. Applied Geography, 24, 303. SOUTHWORTH, J., NAGENDRA, H ., and TUCKER, C., 2002, Frag mentation of a Landscape: incorporating landscape metrics into sate llite analyses of land-cover change. LANDSCAPE RESEARCH, 27, 253-270. U.S. CENSUS BUREAU., 2002, State and Coun ty QuickFacts. A vailable online at: http://quickfacts.census .gov/qfd/states/12/12001.htm l (accessed March 2008). U.S. GEOLOGICAL SURVEY., 2002, P ublic Land Survey System of the United States (Reston, VA). A vailable online at: http://erg.usgs.gov/isb/pubs/gis_poster/index.htm l#what (accessed March 2008).

PAGE 72

72 TURNER, B. L., and INTERNATIONAL GEO SPHERE-BIOSPHERE PROGRAM "GLOBAL, C., 1995, Land-use and land-cover change : science/research plan (Stockholm: International Geosphere-Biosphere Programme). TURNER, B. L. I. A. W. B. M., 1994, LandUse and Land-Cover Change: An Overview. In Changes in Land Use and Land Cover: A Global Pe rspective., edited by W. B. B. L. T. I. MEYER (Cambridge, New York, Melbour ne.: University of Cambridge). TURNER II, B. L., and MEYER, W. B., Land use and land cover in global environmental change: considerations for study. International Social Science Journal, 43, 669. XIAOJUN, Y., and ZHI, L., 2005, Using satellite imagery and GIS for land-use and land-cover change mapping in an estuarine watershed. International Journal of Remote Sensing, 26, 5275. YANG, X., and LO, C. P., 2002, Using a time series of satellite im agery to detect land use and land cover changes in the Atlanta, Georgia me tropolitan area. Inte rnational Journal of Remote Sensing, 23, 1775. YUNHAO, C., PEIJUN, S., XIAOBI NG, L., JIN, C., and JING, L., 2006, A combined approach for estimating vegetation cover in urban/suburban environments from remotely sensed data. Computers & Geosciences, 32, 11p. ZHANG, J., and FOODY, G. M., 1998, A fuzzy classification of sub-urban land cover from remotely sensed imagery. Internationa l Journal of Remote Sensing, 19, 2721-2738. ZHOU, B. A. K., K.M., 2006, Neighborhood Imp acts on Land Use Change: A Multinomial Logit Model of Spatial Relationships. The Annals of Regional Science.

PAGE 73

73 BIOGRAPHICAL SKETCH Muhamm ad Almatar was born in Kuwait. In 1997, He graduated from Salah Shiehab High School, in his home country, majoring in science. He received his undergraduate degree as an honor student in geography from Kuwait University in 2003. He worked as a high school teacher in Salah Shiehab High School for approximately a year. In 2004, Kuwait University offered him a scholarship to obtain his masters and PhD degrees in geographic information systems. Muhammad intends to attend the University of Florida to obtain his PhD in geography.


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20101203_AAAAAA INGEST_TIME 2010-12-03T05:34:13Z PACKAGE UFE0022396_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 6889 DFID F20101203_AAAARK ORIGIN DEPOSITOR PATH almatar_m_Page_32thm.jpg GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
7c80f2eb6786ffa0bd4d0761f11f14bc
SHA-1
e16e57b6460f558d6987501486defdde1cf59d98
1051815 F20101203_AAAAHP almatar_m_Page_53.jp2
0fcef9ae222b834187711b40748fd0e2
51c8dd8f40629dae626372fdc2739be3299f5f11
6438 F20101203_AAAACS almatar_m_Page_04.QC.jpg
219b3cca17266fe745c61af2dc697562
e66822e103baaee6841ea9297206a5cad2530b0e
6150 F20101203_AAAAMN almatar_m_Page_60.pro
c54bebc1f5fdcdcbdc407cbfd07c8390
c618add7af2d2758a338bfcaa9f4d446bf633413
28123 F20101203_AAAARL almatar_m_Page_33.QC.jpg
b88f420a57d1fd59af8c7869e5ed6c59
809a9f17ed0f8e0ff47ff95e5a63eb5d108ce7e5
1024218 F20101203_AAAAHQ almatar_m_Page_54.jp2
b44ac118b5f18d4d4a9859e7a83c6654
48d3975fcfe1f2b1d47d53be332fe6ff18f06b03
91618 F20101203_AAAACT almatar_m_Page_11.jpg
aea45038c1591bfa8c62ab3e95969771
3a044cf4a6ad8cfe5ca33e9cdc1b38292ba682ca
17009 F20101203_AAAAMO almatar_m_Page_61.pro
ee2d0f047896cf4906ce91e5fbec94c0
80121cc65f609a9f686415bc2f638e7a613df7e6
6661 F20101203_AAAARM almatar_m_Page_33thm.jpg
a92dd9d9debec39cc64239d50de0a927
3328e2697807525329f88009f84f635b8ea6285a
409816 F20101203_AAAAHR almatar_m_Page_55.jp2
8a92f8b418f8b8947d37fc8ccfcf5dd2
a4f4a4fbbaf3cfeff1feee8874aaa802efd14c4c
94 F20101203_AAAACU almatar_m_Page_03.txt
e54d897a466c8167068f22f0b1da0225
5fc0cca844316105aa85a3bb44507c0d3efdaeab
12149 F20101203_AAAAMP almatar_m_Page_62.pro
07449927d69e78be5e33d6f3f9114ca9
a76f0320d4ba450615c9f47d191f0299dc2de190
27703 F20101203_AAAARN almatar_m_Page_34.QC.jpg
956eeb44ce9df9ab445fe555a11f44eb
9a63e524ed3c2012488dd5119327618ecda5eab3
860059 F20101203_AAAAHS almatar_m_Page_56.jp2
3e15faf8a42633d885b13387ba256ea4
61071d448d155e5391d3f63da4e2745e2cf4d677
6725 F20101203_AAAACV almatar_m_Page_31thm.jpg
92c620666d0481b3f36cecc302f82485
5ea9110145d12db10320ec4098a8ca5277f86dbd
6680 F20101203_AAAAMQ almatar_m_Page_63.pro
3297b899612c98f8cb705ef8c779ca0f
91643813ac8d92b538ed616e27f17ff46be682dd
6529 F20101203_AAAARO almatar_m_Page_34thm.jpg
015e8f2977c1524b844d0cde1140ff21
c579406d1b0e2198bce64e2ba2ecc06c51c213cf
271068 F20101203_AAAAHT almatar_m_Page_57.jp2
f6d68c58f903831de360a36240bd45d0
c2ed17ba599318ca43134ea0ec5da1adf9d8ee65
18522 F20101203_AAAACW almatar_m_Page_46.QC.jpg
0824438cfc5f85e4071d59d29440dbcc
389e9055be00a4fa14a2919b31b5dbb45bc93b57
13301 F20101203_AAAAMR almatar_m_Page_64.pro
98a89716a6a50ba1527114450f5711cd
df05e2a243c0a1eeac7d04b7ea648e13dcb5ebc0
28986 F20101203_AAAARP almatar_m_Page_35.QC.jpg
3ba2c1c3d10451feb478763e8ed4d139
dfa354092899cbb87754363ded3ad1da77e448e9
527343 F20101203_AAAAHU almatar_m_Page_58.jp2
8b862ef7a43a771616454269950103eb
16a240a871728332affcd7255c8c44d5b20ff480
25271604 F20101203_AAAACX almatar_m_Page_19.tif
78b988429ac0867dbd57b718f3420137
358c77237cd001592c4bd264bab947e599ce3832
1716 F20101203_AAAAMS almatar_m_Page_65.pro
cd70a17af01047bf424fb628c16b24f2
a167341ff172884e5a74d2846a2776157d9c5cbb
6833 F20101203_AAAARQ almatar_m_Page_35thm.jpg
0c05f7c79b4fcc90df1675c5f63828cc
ac14dc97902e1bbdef2480a7ee3f3813b6eb95a4
10684 F20101203_AAAACY almatar_m_Page_52.QC.jpg
83b7da0c392616f56427f78d3fe024b3
6e8407a5d2abd4ecea4ca7aa356cf6f717f93799
5084 F20101203_AAAAMT almatar_m_Page_66.pro
2d8b1de245bdfdc0f22cd92071f08ad9
3cf1fdd066a48ad4a2e549764e4fbc1e6ac58d83
405914 F20101203_AAAAHV almatar_m_Page_59.jp2
f5eddba343ab4608d341be8cf94aa7a1
7f9a6ae9d752ddb96eadaea7425b70e8514106f7
28536 F20101203_AAAARR almatar_m_Page_36.QC.jpg
d234de6c7cf11a92a0ea8e55918ba36d
560faaa47b2b832cb15f7bcea0e4b195ef533c04
F20101203_AAAACZ almatar_m_Page_14.tif
db8f1d3f5d2b8f68134b8dc57bbb9a90
95440a24d7521dc65f72106ab2a1f9da7b1372f2
1051965 F20101203_AAAAHW almatar_m_Page_60.jp2
fe7d8ec67613e09fa5c5ed3ebb912165
fa32ef6c73aa82b09e88426307d36ddbba126c48
6810 F20101203_AAAARS almatar_m_Page_36thm.jpg
b2c5a9f17c5d5834e94124029311d676
b90d3113278efe68fc9798659b41e1ff63815984
51687 F20101203_AAAAMU almatar_m_Page_67.pro
147f49eb7b400843c2cdd74022d43717
2b047d3671c139cef67ac15a977ab9a384a2445f
342390 F20101203_AAAAHX almatar_m_Page_61.jp2
c92bf9e7a7468ad0220ff491c6331a5c
bad134bc9dfa7143adeb35f54cc2d75c5fc3aea3
28034 F20101203_AAAART almatar_m_Page_37.QC.jpg
060f1cb662cdd38de6f60103a73ce0f9
f65f7da3a6d171a682c64831cf6a48662e0d9ae5
51673 F20101203_AAAAMV almatar_m_Page_68.pro
879ce266b5d268d807fcbbca04604420
b5f25cc2b5892f12d1b457ffc090ee099d330e4e
53256 F20101203_AAAAFA almatar_m_Page_44.jpg
31f0b162503283ffb2d1b1afff0576c2
86627afa5bbde2722d7530bce3f5785d35c1adf2
1051964 F20101203_AAAAHY almatar_m_Page_62.jp2
07b79a637cad1bb34b041535a15201de
17600fffa92917613576597ee19f088def5bb4c3
6753 F20101203_AAAARU almatar_m_Page_37thm.jpg
ad3cf8bf4a00a68c20a1af1f720e2369
68727c84e9a97abbb3be2a9155972965c34f9c3d
61010 F20101203_AAAAMW almatar_m_Page_69.pro
2f9aec93322a161775ce571a3f027c93
4cae7d4b716127a63bef90b1d9623929ded7a18a
23446 F20101203_AAAAFB almatar_m_Page_45.jpg
2d0fb1642e70d1aa1e26ad9d260a4487
4522caac3e4368e10c2003848dc5b95cc351c406
1051950 F20101203_AAAAHZ almatar_m_Page_63.jp2
9d59c840d5d1e88d125b32e140874aaa
032f78a05e467bca2a875185898f3c42190455fb
29137 F20101203_AAAARV almatar_m_Page_38.QC.jpg
579acac1a2fb2fa179664dd9ddfb4fb2
76c56059f2a8a653f97a70ed409bce8a368c9292
59985 F20101203_AAAAMX almatar_m_Page_70.pro
dd83801d8631ae12762dc7fd48b3644b
58fca5efc0efbd55d32c3142d44f30c943b4eb39
63240 F20101203_AAAAFC almatar_m_Page_46.jpg
c4aed1ea35a8f0212bf524d1ff142640
219ef9a9602b369df3c1ae2d870527efd1411369
6876 F20101203_AAAARW almatar_m_Page_38thm.jpg
1a10dc995b548f64a3c7bcb4bf2cf729
b7d2cbf35da7fc7f344ca2cd096f25f64aa66b1b
F20101203_AAAAKA almatar_m_Page_58.tif
623d1dd5c00a700e001d086fa3f6bde1
cfb4b2c9a4058f7b33d1186d0c8208279aa06e00
65959 F20101203_AAAAMY almatar_m_Page_71.pro
7ec5ff5a7eba521534adf589ec8bbc1f
c4efcdc6bf14886fa747d5020b8b4f8baa754437
63208 F20101203_AAAAFD almatar_m_Page_47.jpg
9da3380caa1f1b1b2d1759f4efee196d
4f38f92b7b0468e464454227e634b07570e263b1
29162 F20101203_AAAARX almatar_m_Page_39.QC.jpg
1ea747eb1138e5f60c1d90067acbb7f0
f9ee1ce7170cf9fc301bcf60914c41c4bcc876bf
2214 F20101203_AAAAAH almatar_m_Page_73thm.jpg
42d84667fade749837eea0cb3ca5cbc0
ed9785e7da7836e887d960c19d1ffd0f5dd718bb
25265604 F20101203_AAAAKB almatar_m_Page_59.tif
65d2587b31aeead3be7874a425cc0e36
bdf5d6e1c36e80a7d5abc4c5807871f9c7439c1e
40081 F20101203_AAAAMZ almatar_m_Page_72.pro
7817bf45300edc738ddbd07d286a6ccd
93d72a2a3c3d2c9523c893171ae7b329dee32a08
58516 F20101203_AAAAFE almatar_m_Page_48.jpg
41f99114888c3f30085791dfd1f3660e
ca4eeb7592386ee5120a7be0962a8a9b2d5ce077
6847 F20101203_AAAARY almatar_m_Page_39thm.jpg
63d662706bf1cb688874dd8325bef50c
a29f60b632452545bfef3c9fd39bc75dc9070b50
F20101203_AAAAKC almatar_m_Page_60.tif
a6360250d8c633aa548a49df6068ce16
06376cad2452e6a9b14203af346da9398db907c0
78116 F20101203_AAAAFF almatar_m_Page_49.jpg
20547ea838631a9d7890f8ced8f393ec
a592404c24fb7092bb78f3beaeb21116fcacd09f
F20101203_AAAAKD almatar_m_Page_61.tif
0a0cffdfbb38d98fba0ad9ce9d695ad0
d7fba23079a4611569d456e745ae6505b94292a3
55887 F20101203_AAAAFG almatar_m_Page_50.jpg
d6f391891d337f817fae8e1283b1503a
187c82d08af3d9cd7caf3d0444734b4fc44bd478
667 F20101203_AAAAPA almatar_m_Page_62.txt
e696ceca830d7929eb1c9458e4417bf1
4867e233c4320ce9b19aec7e2e4a519ad0d13f08
821 F20101203_AAAAAI almatar_m_Page_42.txt
b4aa45eea434a2e531bac6c47fdb5ea9
97caa2303d033f7c94ada6ae05da4745aaca519d
6840 F20101203_AAAARZ almatar_m_Page_40thm.jpg
eb1a081179e70012e0fcab8a05c9e8d2
9ea8d40f2624cfcc4c904e41c0a9ade09f336ab9
F20101203_AAAAKE almatar_m_Page_62.tif
1ad121a9ef53f5be68e50b9c63f36bd0
ca69a485bc1db5a1631e0b8edfef77b6cc8a4315
53332 F20101203_AAAAFH almatar_m_Page_51.jpg
70f81beac439324853d8f91466982a5e
faef2d17c32e86f5ae0fb61820ae06d69fdcb5e5
402 F20101203_AAAAPB almatar_m_Page_63.txt
a24dd2918076b6d4ff49db9be0b757fe
4874e64aa43a43b0c6fed43a1f9216e3f61c5d51
27818 F20101203_AAAAAJ almatar_m_Page_25.QC.jpg
95c85ba61fd34efa645dd60390a98956
0ca39ee67b8e45e8c07b21256ea8fe3efd82f9c1
F20101203_AAAAKF almatar_m_Page_65.tif
45549324e229e0417b2e198b61e9a9fc
9bbc4afa661b7708cf777536e8c14d79176028ec
29779 F20101203_AAAAFI almatar_m_Page_52.jpg
40902331c12400c047bbe8f9a71e5ba0
21c36a73cdca70309f5ebf091a029753e94cec79
759 F20101203_AAAAPC almatar_m_Page_64.txt
7e2d5bfac6c21c1db476cf31f1a58670
59db037bd3b96e70fe6203d949844642ec339f66
26603 F20101203_AAAAAK almatar_m_Page_59.jpg
2899d932437ab888540a970082ffc60b
94ec2fa375ff9e4b31860e19163b708ecb8fec1d
4604 F20101203_AAAAUA almatar_m_Page_72thm.jpg
56aa39b7ae97986c972d3e09bcf45a39
931337601d52c7da6a627a2c437f19079098b98a
F20101203_AAAAKG almatar_m_Page_66.tif
dc1188fecc46b776717ce7ff04ede622
38459021cc055902383bc72b75335c32e062e5a1
96937 F20101203_AAAAFJ almatar_m_Page_53.jpg
961daaf8148b7c60b0e9f2c0f8c58680
e5421fa19c82ee6f8baa8f81d9d9bd8a07d6f6c2
130 F20101203_AAAAPD almatar_m_Page_65.txt
2035d481867d8a190931166b274d931a
66880e34519547fd0674e89749b3645d59806ab9
5349 F20101203_AAAAAL almatar_m_Page_60thm.jpg
86e63097eec4fc9780ea043e5a6a1874
c2bfb4dc78c991770c2d71084562083f1be9f051
9111 F20101203_AAAAUB almatar_m_Page_73.QC.jpg
e90a28fe64f7f0b4fa1e55bb2c67b2cc
aaf08d884198e3f265acae44c85a6f943d295ece
F20101203_AAAAKH almatar_m_Page_67.tif
37529ecbde4eb70f395fc9558bd86d90
ca7e31aaf034831d4c47565c635ab58c5160c937
62342 F20101203_AAAAFK almatar_m_Page_54.jpg
c4b35f918b5e5e4f2bc4aa3ed527c4fc
a7658bb8c121abf42e5532f41ba5c5063e9f38cd
319 F20101203_AAAAPE almatar_m_Page_66.txt
13cabd7f4a7e74dbc8625874fc268019
09779e1edd0317011cb3500f5851918b01280278
52356 F20101203_AAAAAM almatar_m_Page_34.pro
cfafe64ae1c69aa2f7917e8b7ca01db0
314deb7a7f14b611c5b590c477ffb6d0988ec9e3
86540 F20101203_AAAAUC UFE0022396_00001.mets FULL
174a226426ae0c79be3033f6b6cd3518
7fef94929fa6563d2e328ab52d4079d17b15ea2f
F20101203_AAAAKI almatar_m_Page_68.tif
6fd7c8151f4133ae8c7c553c9b2f3e27
0088ac29d86aba5050c23817133c1b94bca617bb
73119 F20101203_AAAAFL almatar_m_Page_56.jpg
bff0c3207681af6db62603b64bf3b46b
05d8c678b35b811e5161f0aee385dccfe391c03d
2120 F20101203_AAAAPF almatar_m_Page_67.txt
784b0f7392f980bf79c8e8e3ccc20dab
0487d6b71c4a04941336e9710a3d7855b443ccfe
427 F20101203_AAAAAN almatar_m_Page_55.txt
d769d8a30f9e47de47e43e3b4f33bd31
8175fb960cfe8c881eec77d9e5d9d37709734159
F20101203_AAAAKJ almatar_m_Page_69.tif
2730e5758cc064145fa288ee877b7350
cccd35a9adf3ca5341d8348664385f900d7d239e
40758 F20101203_AAAAFM almatar_m_Page_58.jpg
ad2be49de16053c8c51b6a574376cd9b
b5a010ff414ac21adbe0804c2729bb0757d0a51e
2039 F20101203_AAAAPG almatar_m_Page_68.txt
6ae6a93ace58b9ba0699b217547ac2d9
1f1330ece9ffc7d4bd964e09cdb8cee8540728bc
20786 F20101203_AAAAAO almatar_m_Page_66.QC.jpg
cbb62494b8dfb054c5dfcb237e5f04a3
cc0c5ef9c111346c6cfda6b2ce8722285d92587f
F20101203_AAAAKK almatar_m_Page_70.tif
3fbe9e73eec7e95da8d1e68884fa26c6
9483cc819cf7fb050db09985d89c8ac20860cfde
2463 F20101203_AAAAPH almatar_m_Page_69.txt
5617a65ad835dd12b2ec89471d9b01e0
04e970828ca9e6bd885d5f5a22a6660e8261e9ff
F20101203_AAAAAP almatar_m_Page_72.tif
048d0a5cd6d4bf4bb095e0afdbbca12c
dafcb00d05737d8c5b5febb79767d122d0963e0d
F20101203_AAAAKL almatar_m_Page_71.tif
1a32b6bcc947d4467b3d971a8a0ba494
c9714296fbd2f078a6e8850a1bf6cee75cf4e3f0
60528 F20101203_AAAAFN almatar_m_Page_60.jpg
a139065edca290a7219597eadb7d7668
aa0da770805be84724446efc1642f4540ce45592
2656 F20101203_AAAAPI almatar_m_Page_71.txt
344e9810bc15bd37d8230d8224c99072
01988906d708f449dd4746b2db57b5f24157158b
14498 F20101203_AAAAAQ almatar_m_Page_44.QC.jpg
f20b48600df7cf82af17f4786003204e
e0a93a34137725b38e061a1acc13ad0ea1036a84
F20101203_AAAAKM almatar_m_Page_73.tif
af772478915811ce72679601973853cc
be4e6acc53a0763be166a7516efe05219023373e
82734 F20101203_AAAAFO almatar_m_Page_62.jpg
7643b2979313ef8edfd40587136ef44e
b54c3f805e59380c612fe00a602e3bb7820ba8b1
1635 F20101203_AAAAPJ almatar_m_Page_72.txt
f34fa9af55916c23871088641dd43fd5
15e8ae3219433e8960c86714dbd090780ddf2f9b
5087 F20101203_AAAAAR almatar_m_Page_54thm.jpg
1574b687de03b7aeeaa25a28044888e1
462aac2223deba2d29e096f5b3affa08d5baac53
8900 F20101203_AAAAKN almatar_m_Page_01.pro
f09ad424c0f85e3b48b636a2f31b0dd5
0f06548f34b7a55ce71ef3e0cc9d249fe0eedfd2
73505 F20101203_AAAAFP almatar_m_Page_63.jpg
409f96c32ce496ead10b0247014f7ffc
e331ef4f646d0a54b46af7ccb0613a86507ef7d3
626 F20101203_AAAAPK almatar_m_Page_73.txt
0d2c38a8fcdb790fb825a485ff083ced
ee4573274d879719fa05ca0d3bc2ed3501301b04
2793 F20101203_AAAAAS almatar_m_Page_08.txt
c5083626c91583cd006b75a22256b8cb
f667e8f00f2326f23b7333393abefe515af7586e
886 F20101203_AAAAKO almatar_m_Page_02.pro
a5c6b37d1f7d72dc3428ba494df129b0
fd0c2608efe8716469c55649c47794b95bfd6690
62177 F20101203_AAAAFQ almatar_m_Page_65.jpg
b21174f195a259f3afeb232511313a67
94d83938acab4322c4e7636670305b3afd2582b0
1658 F20101203_AAAAPL almatar_m_Page_01thm.jpg
9fd4d70daaf85fa111366615882bd2b2
e55e803291e859663813cbd05fbc52bbc9671097
82745 F20101203_AAAAAT almatar_m_Page_18.jpg
3400386e73aa2bae4723272c52d57bbd
b1a3ac7bd4e4dca6a32cbce3fc4ae25bce18d43d
1144 F20101203_AAAAKP almatar_m_Page_03.pro
599df7c48ecaa6cff5b9fc8c470f0141
5c0bf7f369e6f0ea611800213ad114cf21c027d7
73918 F20101203_AAAAFR almatar_m_Page_66.jpg
065e845c76f779df281251cc51aec444
7a01ee77b571bcfd6e0843b1088e3cfc1c8ca993
7169 F20101203_AAAAPM almatar_m_Page_01.QC.jpg
53130d0a835a7a8e1c2fa7bde7192810
15542eca09cd8c5f2c709d0a1bac17c58f186f01
F20101203_AAAAAU almatar_m_Page_24.tif
70037f8e3dcbf26b8f5d6e00ee421c30
0ea4ab94a9df5c76ca31d7a2d334a9a6dec24ed1
95858 F20101203_AAAAKQ almatar_m_Page_05.pro
a2e8f13af637a04c224584ce560b35a8
b15db5c00366326225e3da51d1bde3b7f8bcaf06
85125 F20101203_AAAAFS almatar_m_Page_68.jpg
7bb659fd7c9340c378eb3d3825ca0173
b7b07af8d8184772669507635d0f824055f20387
455 F20101203_AAAAPN almatar_m_Page_02thm.jpg
2a009d83a272794948b0ce2390c2083e
56810b9bd6bd7c5ccd170f56e9e254a158f794e5
28980 F20101203_AAAAAV almatar_m_Page_32.QC.jpg
c063002332012c6bc2537b077a218b9b
8ce19bde918e0c37ac2657647851ba74f1360e86
36289 F20101203_AAAAKR almatar_m_Page_07.pro
0d99d95f685e832ff2eb7a6444f93275
b4099c31a1f12b0012f22b8b066fc8c29a817466
102651 F20101203_AAAAFT almatar_m_Page_70.jpg
3c1d4634d77d852098f933414220e941
422f4fe5e57ea06da7cae949cd352931968dee51
501 F20101203_AAAAPO almatar_m_Page_03thm.jpg
b73d82892124243fcf126a0fb47c1ced
89b1d13440e957b0bcd6a4c1e321e143b11845be
12097 F20101203_AAAAAW almatar_m_Page_64.QC.jpg
1291fdb3779b21125bbd4406a8101330
dc9ea8de18802df9d199eef976d51182c9f5f906
123807 F20101203_AAAAFU almatar_m_Page_71.jpg
59bed8f425a487c8108143e3162e434c
f951959e355ee51ed887be53a3cd238be8e5e3eb
19079 F20101203_AAAAPP almatar_m_Page_05.QC.jpg
282d17e79b8a7c12acd2ef18f55566c2
c2483252e7c4c5c196af5c7fdcc1f74077eaf7ad
10163 F20101203_AAAAAX almatar_m_Page_04.pro
1a9981e95206ebd5b16c4c60ed30f54a
1abbf5d345f660f0a831dfe531dc1f9d6a3e5284
72835 F20101203_AAAAFV almatar_m_Page_72.jpg
02337f30573b126554e5fa6514a99bc1
66e8c5f0ddcfc6959c58614059c896839f332671
4769 F20101203_AAAAPQ almatar_m_Page_05thm.jpg
fba556c10706d7e3634510e66194178b
17dbc523ac37fc602ef924c83a9660cca2b208ad
1051960 F20101203_AAAAAY almatar_m_Page_66.jp2
22d957615c45d9c71c7a5a8290e8e5b6
8860929aa0fea82ef62e7ac8f4f89a76681dcb42
69534 F20101203_AAAAKS almatar_m_Page_08.pro
c18b66850f623443426a4fd3ea6feef5
5632e86eef1b79e1306e3c97f3325bbf8d8a3dd2
29376 F20101203_AAAAFW almatar_m_Page_73.jpg
2153a1edfa8ae66cc08fc2ec3735c31e
4f1584a149a756e1c567cd571d297ab1fdf60767
3376 F20101203_AAAAPR almatar_m_Page_06.QC.jpg
a11e0c229d1193febbc398371188967b
1c15bd740fb0f413c05d701772fdfaeb201d5b83
29319 F20101203_AAAAAZ almatar_m_Page_40.QC.jpg
d22c41f590e2464ca064f793fa15b018
c61caf7d950769509a3ee0732e02fccdc045f347
43901 F20101203_AAAAKT almatar_m_Page_09.pro
ab59871ee5361ee87a53c34a205ba99c
45bc55234628ee7b29adeb9027e075c47e4ac6d7
268319 F20101203_AAAAFX almatar_m_Page_01.jp2
be9a9b4cf4cf635653a1c3f6ad21a565
023030e25ed9c7b610cae6a01a9247e39ef02fa2
1168 F20101203_AAAAPS almatar_m_Page_06thm.jpg
73a1008c1d733e359305f1beb8284ef5
dba847a86c516553954200350a9fc85413af8546
51953 F20101203_AAAAKU almatar_m_Page_10.pro
69bda94c398c826f87e4cd56e61a8c1b
a28ba3fb4d53818f5290bc3141cdf251d29a9334
14547 F20101203_AAAAPT almatar_m_Page_07.QC.jpg
8d444803a9cfca771038d3283a95a183
963af6a9962d6f674a66fbd471b4debe8fcc939f
F20101203_AAAADA almatar_m_Page_38.tif
b874e30b55e41285003bbe86db55f38d
977fa92ff22c32b8c09be031c29c6adffa729b8b
28099 F20101203_AAAAFY almatar_m_Page_02.jp2
c2f7ef443b93d6bea4b4e70179b5d01b
e28f2b7f02602b8e6145c5fe9f6594bf42d6eb45
55792 F20101203_AAAAKV almatar_m_Page_11.pro
59060cdb06acf640df0e3f8c0f355b4d
a3191b8d288634bc8751ba3e797b9ebffe7b2274
3578 F20101203_AAAAPU almatar_m_Page_07thm.jpg
678fc106f499dd31568fb5000407da28
723b936dcb2d04de07ec6fc9760a03182e49cac0
F20101203_AAAADB almatar_m_Page_71.jp2
4a8dda61d095a43d99473bb403c576f1
45a00653c1f14691c06a30daa95180a570708a79
29886 F20101203_AAAAFZ almatar_m_Page_03.jp2
39f477c5eeeeb9bad7a320d40802f9b6
54d4e7350ad3f0a1d3ef77e2e356e0f6e937c2f7
26415 F20101203_AAAAKW almatar_m_Page_12.pro
f2371850b98a52f6c1dc074bb5a47452
1ccffffba6934d1abeda2afed98490e000c7a9c7
6024 F20101203_AAAAPV almatar_m_Page_08thm.jpg
b878d75d77cc1e7a8eea096e321da133
7a754a36842590106d1f4d7aefef9b4fe08f42c9
F20101203_AAAADC almatar_m_Page_33.tif
2d40e54d000319f461fc50541ca32b9f
4b826dd847321d18517f0511075c7c7c2e4c8222
49316 F20101203_AAAAKX almatar_m_Page_13.pro
eb70ba1ad2b26c8adadeaee3f4c1234c
ca726e3372b7338c6a5900795f6d8e9799e9baf0
5481 F20101203_AAAAPW almatar_m_Page_09thm.jpg
14f08e67e87c8f101e91e2035bfb4e8d
e3558707c4bcc4bbc8d8fafcfe8cd83ecc9c66b3
1087 F20101203_AAAADD almatar_m_Page_12.txt
2cbc1811052a81da1ec3dbbb8b7a0e66
55ae3c20d8eb56983ed205d7e332aaef34b08e20
F20101203_AAAAIA almatar_m_Page_64.jp2
bc365b3086f425358d4bd4fb1d600fe5
fd0241c01e88e4b96c12d14eb6989d854ba63d0e
53533 F20101203_AAAAKY almatar_m_Page_14.pro
b761ac897c4bf626b03a9f04abb5e39f
baeb3452994ef9e8503ecd2ffd0844679fb20616
F20101203_AAAADE almatar_m_Page_01.tif
f180396f3e29b9a9e862341d1966ed1b
ebf96fc92dd94d9f3d48e4f4d578188447f3714a
F20101203_AAAAIB almatar_m_Page_65.jp2
0e20b232096f01540d33f051dfbfe61b
a72a8dc15047e2f962e3ed1ae378b86c167a5fca
49238 F20101203_AAAAKZ almatar_m_Page_15.pro
c9be2252559e9e92baf799ef45a8ea60
542bcdbe1477ea15c6ca431b0f9f688f59c05f6d
27269 F20101203_AAAAPX almatar_m_Page_10.QC.jpg
6bad861370cb4bc1e2907d81a5fc12d4
bdc76df8bc738382767f3d8323800de1ae289157
1051905 F20101203_AAAAIC almatar_m_Page_67.jp2
c8a54f47d67271468c7b24e6843e49c4
bbafa5f48164b4ca29ed4fad89376ba72a3189ce
1051884 F20101203_AAAADF almatar_m_Page_70.jp2
c25897affadd7734bffa2faace080271
2e9de56481a9a32b1d6ed2339a087fe890e3837d
6352 F20101203_AAAAPY almatar_m_Page_10thm.jpg
aa0cc09b4f8824381da390491e40afc6
4ca6a33492d6173308cef8bd4e9907975d76d4e2
14832 F20101203_AAAANA almatar_m_Page_73.pro
a2253ee3695a7d0a4c06a229743d5b45
27bb15fce8c117232c743da30cc9f78d2456599e
1051892 F20101203_AAAAID almatar_m_Page_68.jp2
2280a36d8495ec1eec257df8d39b5e7c
9a3e781ca345fe081900084687c516fce744fa79
2151 F20101203_AAAADG almatar_m_Page_33.txt
12e8ff46ce4d58ee9919d3e0b8237adb
3791d78cd71eeb4229ca42968b9dc09e7da7f457
28872 F20101203_AAAAPZ almatar_m_Page_11.QC.jpg
01fb656cf283a0b2ba3817ef1d1fe9e9
f3fa9fea71ad0c9f40df95078ccf422c9f1f8dc4
498 F20101203_AAAANB almatar_m_Page_01.txt
040c601b7b7af802166325c8533b0f38
fcf8830568ba6c3da45763063fc1373253eae700
F20101203_AAAAIE almatar_m_Page_69.jp2
2238312429378e86cc82fb2abe4988fd
6b3049ecddaa3e026642463b8e25324a3bef027a
25904 F20101203_AAAADH almatar_m_Page_08.QC.jpg
6d51466e1bce5d145365e9efa7c3b34f
62c01ef7129751b65d6c2a628f5cbd3d7e148e82
90 F20101203_AAAANC almatar_m_Page_02.txt
1b839dc5403764254af835582ba022f1
c6365d0ae74d68341b3a3e812adfac1ffda47ad4
917960 F20101203_AAAAIF almatar_m_Page_72.jp2
5625a616d49fa2f34a66b6931e128e41
f28c5642c06d0006a6f8e6794ac9047eb266ed5e
11290 F20101203_AAAADI almatar_m_Page_06.pro
c2efa7259c244312ba51180b1446cca8
a7bc73c6bfe157d3e812d571d7a5a69ccbd5c00a
20628 F20101203_AAAASA almatar_m_Page_41.QC.jpg
8d5f0c949f11d5140f9a3f7bbca350e0
40608af86b43c27d3cfeb307fbc0f330fbe8b631
445 F20101203_AAAAND almatar_m_Page_04.txt
7b96747d841660367d62ac63d43ffa18
b8ef3f62e7a3ad5e8bb3eaf7054e1aaefa3fa61c
358291 F20101203_AAAAIG almatar_m_Page_73.jp2
2064a0ce21d81deb18f8cbb6b4fd3f89
f6cd39a401643d6943cca3589460f765589447f0
F20101203_AAAADJ almatar_m_Page_44.tif
3c7527369fdb66187e2dbd6f83890b2b
75d2fdabf682748df3112cc365b5eafa7b9ec881
4864 F20101203_AAAASB almatar_m_Page_41thm.jpg
0699f01205d6b80d0e9ad1c84875afe9
3d7c601c63dd6f40dea07d90d43b38de1e45b10f
3999 F20101203_AAAANE almatar_m_Page_05.txt
4ff2447d9bd5ec7c543b3a30214e7027
92a291ddbb3de1eb73b607a13db2dc3b34c15c31
F20101203_AAAAIH almatar_m_Page_02.tif
2f160a9ad4895855b307d269b1463b71
45ba1c3b982ad88d7902ba7d75fed12396be79bc
1433 F20101203_AAAADK almatar_m_Page_03.QC.jpg
6a725163922601ff444b7ff70e364b20
47fece19ab2b4b957efcdbb8c811d6b64bf5f2df
9212 F20101203_AAAASC almatar_m_Page_42.QC.jpg
b233598986b9cb2072eb648a82203bfc
c8721800525ffdac1c2e06a901a5af4c528e029d
591 F20101203_AAAANF almatar_m_Page_06.txt
72e4fdd500b25094e25a7e8a0f30b2be
5eb13a48428e6892e5573e77f219ba8f36b8963f
F20101203_AAAAII almatar_m_Page_03.tif
3c60283051f9c485d7fa0ceb5f8c7704
57be4b446e48622e21a18247790806fb1e2e1471
2577 F20101203_AAAASD almatar_m_Page_42thm.jpg
0eea4f5307a9f294be6b738cac3430f9
540d56a97b0e5bd416038fd2f0fa3ee3efd252e1
1488 F20101203_AAAANG almatar_m_Page_07.txt
dff42a1616b5dfd5bb2abd0b0db7a867
f586450560c6633bfd6aaaa55abc3d5d87bdf1d6
F20101203_AAAAIJ almatar_m_Page_04.tif
b2432fa80b2445db46f77f651814346a
f799cb81c6fcc3ad58168358f6042a8811db551f
4930 F20101203_AAAADL almatar_m_Page_56thm.jpg
4df35e9287ac0d833f5304fa975a2413
f4662dd646559df4974028d032f84af7c54c0bab
14675 F20101203_AAAASE almatar_m_Page_43.QC.jpg
c45eebf8c78660c3163686f16423ef99
2f56eef1228b89c7f7a98321afee9e124b252559
1924 F20101203_AAAANH almatar_m_Page_09.txt
efeba421b017ab0bf17b40454a64029f
0f747e4aed95f2f0821e38b9aedf4caa55e33539
F20101203_AAAAIK almatar_m_Page_05.tif
aa20573a4f843d90daf0cf0cfdf486e6
3e07eec19cac2f54916cde4dfb9a249ae8aae642
111660 F20101203_AAAADM UFE0022396_00001.xml
2d48eb108f197b8acbf55643424b007d
46a8c4e769f36b15539cfac5463226cf4d5e3015
3848 F20101203_AAAASF almatar_m_Page_43thm.jpg
096cf43e52fea6ce407d3ab4ce753be1
2f8b600b9f7e961a16839a756e1cb48af3f993c4
2153 F20101203_AAAANI almatar_m_Page_10.txt
1bf66ac03068213d59a9281bd8654ba0
dc2fa7b4b39b2b7800a0359380b5205a490fc46a
F20101203_AAAAIL almatar_m_Page_06.tif
b826dabefaa766bc683ead70d9bda0ec
89f987074595f1a4ed7398fd85bfb20aab239d5e
3702 F20101203_AAAASG almatar_m_Page_44thm.jpg
9cdf49bba931fb91d6b14dcc517a3929
b16ec79ce2391c92b2fbdc680b9a4e753f061510
2261 F20101203_AAAANJ almatar_m_Page_11.txt
f3c1a849f220f71463b5bac89aed481e
070e3327b1ab339d281101830dd83ffc690f35e7
F20101203_AAAAIM almatar_m_Page_07.tif
2e32631446cc532454a5d476d286c75b
25e7b576b2fa2bf65a8842d8a2e2e9760449de91
6374 F20101203_AAAASH almatar_m_Page_45.QC.jpg
10bed4fda75e85c016997af0ac2a609d
cd72d5e91ff0a2319cd59315ffea829b54534565
2033 F20101203_AAAANK almatar_m_Page_13.txt
885abc6ba912c7e04a2ac711254ac6fc
4c309f19aff332d25cc0199a172c2d6803607d1e
F20101203_AAAAIN almatar_m_Page_08.tif
ab0f83f3e5200e2258beb4aada89275d
68c2d720a50d85ec12f386657de4d3a141a5d574
24218 F20101203_AAAADP almatar_m_Page_01.jpg
bd0f5e7dafc7771f8e2a048cb1da0b52
ebb538793726e0fa3720a94aa83c0fee5f5d7ab4
1696 F20101203_AAAASI almatar_m_Page_45thm.jpg
acb976252b8a8217b65f54f8eba920c5
03bff1b2d22d8dcedaed0416c9efb692f525c2cb
2107 F20101203_AAAANL almatar_m_Page_14.txt
efd82048af5e8a6fa7d8f4992ee144dd
2a32dd51188c4d346c599ee52a9e50c490cc9686
F20101203_AAAAIO almatar_m_Page_09.tif
b7854bf31fef1aa75c949025b520b2d9
947a5bde938142582954cfedb07aefd413b17bbe
4026 F20101203_AAAADQ almatar_m_Page_03.jpg
27f8e61559c3f04dc74868d7b7e28620
037c61baf57017a4be9fd9da53b2094203d91e92
4772 F20101203_AAAASJ almatar_m_Page_46thm.jpg
726892d6070ba88760e4f42a8b91c8ee
49dc7d64543cb91fc28d0ed8a6e7a3a8b2e44ea9
2075 F20101203_AAAANM almatar_m_Page_16.txt
4b431182e7adfabebf34c0180c833e2e
7d1e46d61885d1acc3bb1a022ce0fd8395093def
F20101203_AAAAIP almatar_m_Page_10.tif
1147e195ac69d105aa61638fe96e0bf9
f8f6c6d85f85dfb4647fce0859448f385e58db06
91722 F20101203_AAAADR almatar_m_Page_05.jpg
6008f73867cc7ee56c04247754decb45
6b606a3564e181714fe7a9a2fd62ff0f9f4ab531
18285 F20101203_AAAASK almatar_m_Page_47.QC.jpg
5892cd9d685e59de5da5161a79ce6992
1715e9b26a88f37850935c6f11b64a2e5b6c84ef
F20101203_AAAANN almatar_m_Page_17.txt
56f5cf6ee8a9b8a8ca374de7599f1f12
519d832142a93a8bf78e4474467b8ce167dd002f
12845 F20101203_AAAADS almatar_m_Page_06.jpg
7ae224e216c63911dfcccc204fec9930
d63478d0ce0e425e52a955d1a2b757fd82d7fe64
4707 F20101203_AAAASL almatar_m_Page_47thm.jpg
72ed635281c213340833e6d7a3aeaa48
36160d9e926a79707d330d9c56865f4fdb490c7e
F20101203_AAAAIQ almatar_m_Page_13.tif
faa0681f611f99e73183d9e1bf59dc73
3bd51c31c6ce2ea35b64d89d8916da498aee9fe6
51603 F20101203_AAAADT almatar_m_Page_07.jpg
c5658139deedb92f4bc4fcb32a64a684
346725e60b56bebedc76f2581d7d217b6a1fce98
2073 F20101203_AAAANO almatar_m_Page_18.txt
965eaf3e4d5b89052224fdee0195870b
ea1e67fd0f192d71b27558f4469e4e244e3353ab
16775 F20101203_AAAASM almatar_m_Page_48.QC.jpg
908077ecfe8e210f7408edc744da06b1
0a90e88f89a658cb265be60d5569d1bed8e8ea72
F20101203_AAAAIR almatar_m_Page_16.tif
1b786d0ee10b9961d4a878c35d278a22
2d38d5da4a8305fe38eddf96fcb4e1f129d6612e
96169 F20101203_AAAADU almatar_m_Page_08.jpg
bf4b9ed38cb25a8ce74c704fc89d00b6
fb2af36c3444034df6d2d085a1bf286f5443f863
2125 F20101203_AAAANP almatar_m_Page_19.txt
0f97e2d9a7d6d01d63af1863fbebee14
70e4308aa91e2bc3465307b4e001ed4edc0ba34f
21502 F20101203_AAAASN almatar_m_Page_49.QC.jpg
af79fd131131c04102de4379adce9d4f
00604f0cd5f9e31bc5e4ca8dac7a8457b8e7dbe7
F20101203_AAAAIS almatar_m_Page_17.tif
54657e69cc7ca04399e2f0aebf384874
b5fcc7c22f51947700c5d85f71db99e297278875
76448 F20101203_AAAADV almatar_m_Page_09.jpg
f563bd51eeeb240ac7090aca2b6947ae
4373ec3396ca0dc2f7d1b0d9de65232e682006c0
2043 F20101203_AAAANQ almatar_m_Page_20.txt
5db206b65b90709bc79c023c06beb168
266a5416a6533b094cd1b81823c6a05c491aade3
5346 F20101203_AAAASO almatar_m_Page_49thm.jpg
edb9f5b02f12450629af0021d4feb660
781960fca6f6cdcacdb358240abdf662ca7cc770
F20101203_AAAAIT almatar_m_Page_18.tif
a4465ebdab7cf4b9bfbefc7e1a611f2a
40e57e12a54f636cedacf9bd3e9aa00b66e55e2f
85993 F20101203_AAAADW almatar_m_Page_10.jpg
1f6a59208b045f34f6caa4c9a455ca91
2fdebbe019e819d23126dcff130c682eba257996
1948 F20101203_AAAANR almatar_m_Page_21.txt
6e06cf9d9aa9ee4ecd2d0c745f7af5bb
bc5f86b2e9cfd4bcd4d9f4fe25587db2d30f0e9c
21130 F20101203_AAAASP almatar_m_Page_50.QC.jpg
a0cf337b1b55d5cb7c6db387ecb475fc
c93990e328454be8c74e7e73bd561558b33210d3
F20101203_AAAAIU almatar_m_Page_20.tif
e95b8c28a2ff5f446678f39072bf135c
978ce8201c4eb2476a792d0c66254c2673c8b517
47904 F20101203_AAAADX almatar_m_Page_12.jpg
bf0cf63212f1c1c5124a00274d91f7ae
236be0de21fba4be1ec606d04489c5c01d109947
1965 F20101203_AAAANS almatar_m_Page_22.txt
7bd8cbd66c83e3ddb42a3f862c67af90
c73e4eaa3ee642a2f21533e044228b7117292dd8
6206 F20101203_AAAASQ almatar_m_Page_50thm.jpg
75b8892db47113ab0265d16741875978
681c9e8244bc2739de6be660a41ac133b1042620
F20101203_AAAAIV almatar_m_Page_21.tif
4e8912a372493741fe878df01d874e88
b48a89f445690cc929712bc8bfcda83ff81681cd
F20101203_AAAABA almatar_m_Page_40.tif
470556beae56a68db86c0fb49d0ba557
5216a1db218a1d4867605eba45fb1dc98f1a8c08
85225 F20101203_AAAADY almatar_m_Page_13.jpg
db96027f539f7cbd852151365563f426
9fa77d4b4e4f04e17a919246ab403dc40a854c7b
2002 F20101203_AAAANT almatar_m_Page_23.txt
fbe0a783f668e78b365e990fbc801d1a
27e1dc72ca7804e4796ea8162700e26ec21c82ed
19467 F20101203_AAAASR almatar_m_Page_51.QC.jpg
a9899795295e292f921d998af33b9990
3ee08430ed017530756a8c203f4ba1c584999d0d
F20101203_AAAAIW almatar_m_Page_22.tif
09febbc328cfaf2872225d61b0f1fda4
62c14b3a12a2345bc34bbe867b4093a68ba5a134
24319 F20101203_AAAABB almatar_m_Page_57.jpg
a0319b30b807a15962df90011c2bcafc
6d92c96f83d808328bf45f89ff7ceda9543b28b5
86472 F20101203_AAAADZ almatar_m_Page_14.jpg
2616b3c9c6e9499a0f41afa68e974f52
ec34898ff25c106d5bc4be4be96b40643d00628c
2118 F20101203_AAAANU almatar_m_Page_24.txt
1f30df5b8f406038ead00566c34ee95d
e83ec959da8dc715c828de4a81f8870ff1f9cd23
5339 F20101203_AAAASS almatar_m_Page_51thm.jpg
70092d97cc0e5a2e5ceb110196590edf
8d34b2a1772abde9ec838ef3dc69f6b58f219ab9
F20101203_AAAAIX almatar_m_Page_23.tif
c47d2bf8ab8c1478e6fccdd7de19ed25
f01d2c2c3b30241a8e362fa4f607a7532d195da4
54999 F20101203_AAAABC almatar_m_Page_26.pro
905b6f863e8baeeeea5c5a9ce0f8b5f3
fc0a5ca393ca59ff0f353f02ab694b9211311e9a
3078 F20101203_AAAAST almatar_m_Page_52thm.jpg
ce9d59df09ebba52632993f3c635a006
d7b074cc13397c74e145153b5cfdc80c28d2b503
244067 F20101203_AAAAGA almatar_m_Page_04.jp2
9c7fc7df7769014b2391aa8dc180377d
6b6aa41b5f04209cd33f7139818c0e73b6491ce1
F20101203_AAAAIY almatar_m_Page_25.tif
ccb099510e7b4fd460a91e68dd24e8a8
ee8f49a0fa7e69d44ac1e2ed2be6666a49a3e835
22695 F20101203_AAAABD almatar_m_Page_09.QC.jpg
fd2541ce607db2ceff3da80a0907d494
1b83416a6ecea5fbb3914630e35ee4df8299ad9f
2122 F20101203_AAAANV almatar_m_Page_25.txt
98ff8abc36969d80e3ab9e3402a373e5
7bfa11098ab1cb7eb2d35d7d6bd86ee1fbc834be
7117 F20101203_AAAASU almatar_m_Page_53thm.jpg
ced39877666e7563d3b57624fbd773fb
d78dfa5dafdeec0effa632fabe8583b841795db8
F20101203_AAAAIZ almatar_m_Page_26.tif
233912e21d66ff138cc222d98b151149
34211aa52584bd3d0a554c584a1355c8f3460088
38758 F20101203_AAAABE almatar_m_Page_41.pro
a60f59e2cb75aa8f56d38874162d5b35
3157c19a63977959440f402717ece07a6beeb289
2166 F20101203_AAAANW almatar_m_Page_26.txt
d0f9ed45cf715fdd1417af143f884f82
2338735a1698574bec889f5fe095a36fffbe7a1b
1051985 F20101203_AAAAGB almatar_m_Page_05.jp2
dcf8936a6e45dd874e2fe73d72f6ae73
ecfe23fe9ebba1370ac06f383021a25b5afa3013
18805 F20101203_AAAASV almatar_m_Page_54.QC.jpg
602f8e75dfff88d99808f9834717536c
f7ebf0363294bb9727a927fba89974e5987bf2ae
25947 F20101203_AAAABF almatar_m_Page_55.jpg
250ab104a6f7930576de41bbd87cf223
5542e3dd3dd5876828fd476d5e2f8426141ca07e
2055 F20101203_AAAANX almatar_m_Page_27.txt
49ebe49b2bd04902daa39abd5270297b
3d81a9ffd6c0ee4115114d293387ac58b4ae686e
229317 F20101203_AAAAGC almatar_m_Page_06.jp2
45cd6ec7ce8d147c0e1170c47206d0bc
31bed914d788d4ea058b1f898e9be629c2ff8af7
9043 F20101203_AAAASW almatar_m_Page_55.QC.jpg
4c6fdbdc8e9764516528d260e0857194
393f6f8dc87395f29d2829e3b5da05fd367db478
28895 F20101203_AAAABG almatar_m_Page_53.QC.jpg
530e5e478e969d7e134ac2852de3b650
c56045a9e6d884a4490a9e92fc36d983e2724432
51764 F20101203_AAAALA almatar_m_Page_17.pro
43b1d0f62d1f622da3007bee7754d098
98a4e0a860a9cd776fbf6400085c07de16a753d8
2217 F20101203_AAAANY almatar_m_Page_29.txt
8682826e92f8e51c74135a520ec35b06
129471e2817d8d3b251e38133e22aefffdba1a80
1051977 F20101203_AAAAGD almatar_m_Page_07.jp2
47e773a85499af38fb8d2205b01e18eb
654b4905aff55d75c1f4b0e8c4c529813793a593
2626 F20101203_AAAASX almatar_m_Page_55thm.jpg
cdf2f7a92c0c64c43f49603517abc7c0
80905a527754ebacd8557a3103a35cab6b501cc7
4074 F20101203_AAAABH almatar_m_Page_48thm.jpg
32973706c22af4711b80f0017e330711
90333e8e0ca685242cf820d4f7ff2e96a95f3fcc
51320 F20101203_AAAALB almatar_m_Page_18.pro
1fb8eba821c378bf11e7d8bcd19ec5ac
910b0a139262fa10dcb436b64164db87bacfbdda
2088 F20101203_AAAANZ almatar_m_Page_30.txt
8c01d0e1bcdaf3c6b120d413e17108b8
fdb32e8c1c18d7188fcd3b5c2389320adb841821
1051983 F20101203_AAAAGE almatar_m_Page_08.jp2
cf420dee47fe7fc39d21d15c5137df3d
89d3f8d2aee923c51fd96e85e935907cf6eac82e
19665 F20101203_AAAASY almatar_m_Page_56.QC.jpg
53e2c2c5dabf1aef1e9774d0bf612e34
07a08b465b5c552ef6febb0463a896f4965096d6
1051973 F20101203_AAAABI almatar_m_Page_14.jp2
efe63dedbae9807a2b2aa99ceaf444ee
fdcae1d8db113bcec088c3b55fb8b81f6a7a90f8
54078 F20101203_AAAALC almatar_m_Page_19.pro
128222bc586337affc01ce2f8b7095b0
64a7d2ab8f8c8f487668159f973375f057f98265
1006455 F20101203_AAAAGF almatar_m_Page_09.jp2
3fbaa57f9caa884a80a7a85b6addb4db
ea265c6de48afd40c73bebb2d5bacf1324afee24
7912 F20101203_AAAASZ almatar_m_Page_57.QC.jpg
bbde4874f90fdb22918722c66e197a5e
fc97ef3612b873550f5a9df82ff1cfe69bfbc9f3
6724 F20101203_AAAAQA almatar_m_Page_11thm.jpg
dc9baa94610dca3811a903f84b4967fe
aa3b922ddccf5f0f65235958efb7ac8e78840c7b
51724 F20101203_AAAALD almatar_m_Page_20.pro
5cfe765dc63fcafd8a67e6819c9aac52
754b3a6db3efd8de3cbba5d0be3c1d68fd853512
1051914 F20101203_AAAAGG almatar_m_Page_11.jp2
312ad622058ba9d17a8e5418893e1e47
6ba647a6d03ee42f9d157f97ac445f0d65ccb192
15088 F20101203_AAAAQB almatar_m_Page_12.QC.jpg
609eb1e86e9158904a830ae0d00841be
7d25a106f2c0a1c33ef3b03b78293bb480957a91
432717 F20101203_AAAABJ almatar_m_Page_45.jp2
4ed620469c8d3532a7621f344f76f8a2
8642a46322adef251b9fbc58d8adf9d52747789c
49451 F20101203_AAAALE almatar_m_Page_21.pro
1f683c82f51e120032255385ab3ba4d4
0c1986d124d48f73ea591b1d30c64dfe6afd56c3
609947 F20101203_AAAAGH almatar_m_Page_12.jp2
735aace13d55895f655b1f2b8c0922d5
2d751be0dcf5fc74421139c9a76602f6cb1687f1
3834 F20101203_AAAAQC almatar_m_Page_12thm.jpg
20be6f13d72f220afe1aa7ff2317a494
acc531479bfc367d79e3559c24798586a6a3b438
2046 F20101203_AAAABK almatar_m_Page_28.txt
c61bdc9e6bbdf18c54d56504fc661684
aaa54fe26dfa2f9f216a545906e556b2b1320883
49731 F20101203_AAAALF almatar_m_Page_22.pro
6218456f0a417ba2b390099a66ffaee3
9aafdecf1c9afba2134f454eb682b3b842d911b1
1051969 F20101203_AAAAGI almatar_m_Page_13.jp2
d389def7e509dc44af6b2451031ea4f0
a4db306a5abcfb88fe44fed9ca30a2b3c3215eba
26928 F20101203_AAAAQD almatar_m_Page_13.QC.jpg
a4e3402a94683e8a0acefbe4a6359528
6f86a35e0c5dc2e908ed2c6b868d688bc0e22184
F20101203_AAAABL almatar_m_Page_12.tif
344456d51a55fa51c93fb17e34abf5c1
0e82402435cc263037356772d84a3acbf6678ee8
49831 F20101203_AAAALG almatar_m_Page_23.pro
b42aa35827f8eae400a1c384af86b410
35ff6147210deba4a4b33995b8552c31407904f7
1051980 F20101203_AAAAGJ almatar_m_Page_15.jp2
ed6664ca160a08622e1574fcc24e43f4
bcfd15fc50187c741697013c73fcff97cadb0645
6487 F20101203_AAAAQE almatar_m_Page_13thm.jpg
69a516b3e4ce9637f0f41d480db5366c
6efd3b899c81f775b94e5aed78626869bfe510b8
52377 F20101203_AAAABM almatar_m_Page_27.pro
4d87dd36a9cba6fd61aa445a501bd0ae
0740ef4a905513641472ff75791381e139af588e
53551 F20101203_AAAALH almatar_m_Page_24.pro
7b0e0157cccdcd150391bcc2477b0fbb
8815d700671d5faf641349166f6688d7d96e1d12
F20101203_AAAAGK almatar_m_Page_16.jp2
4abafe46d36f678ae2fc2d6b61403c6e
85d7ade14ee00cf428e45c9bdb67bf8fe10c20df
28463 F20101203_AAAAQF almatar_m_Page_14.QC.jpg
335e1b27abf492be25dbb92d3bb5f63d
40ebdd4269cd2449fd66f6db5bef0bcdfcca2210
874933 F20101203_AAAABN almatar_m_Page_41.jp2
f74aa606c725eb487c32b52c58f258bc
741094d45d0ee33f8b1ffa8c7ff387d048a3f488
54180 F20101203_AAAALI almatar_m_Page_25.pro
40b252e1d28ac0335a4df7d7af949c67
52f6be6469d8677598e949eb70c057e470dc0fd6
1051932 F20101203_AAAAGL almatar_m_Page_17.jp2
b5ade8b38fe6a049a5fcb361e41aa09e
1a92084162005f0165ce1c5dd93015ba4e59a096
6747 F20101203_AAAAQG almatar_m_Page_14thm.jpg
f9fed6121212c44f9c0eefeca0f821d6
c62f43dccdd5d24bcba5e303977656fa49fd7cb9
2064 F20101203_AAAABO almatar_m_Page_54.txt
b680aec7d5dbdd55ee3aa4c615b8da86
4aeb8cd1909358032e5db2599cab21cf4410d691
51836 F20101203_AAAALJ almatar_m_Page_28.pro
eedfbf8d0defcf6dbbe21d8b6272ad23
49e4e674ae007d65c696c340ae2842112be41b74
1051966 F20101203_AAAAGM almatar_m_Page_18.jp2
bba4777759a70d48829d4420f6a41507
727d714ec5017a9a276c12319eaae7c33bf09abc
25664 F20101203_AAAAQH almatar_m_Page_15.QC.jpg
d86ba0733069c3d755c5cfdccfe363d4
9089d6ac206fd3508c0181f3782b267d99113bce
F20101203_AAAABP almatar_m_Page_64.tif
1d775cb5a45224efef88e375741a6dbd
687809c652dc8115e20ccc0eef3239c5811e81b5
56422 F20101203_AAAALK almatar_m_Page_29.pro
bed59aebd2c1439e0604ce5093bc335e
6340126f8907c76885a4a2d3117c80b14fb2176a
1051934 F20101203_AAAAGN almatar_m_Page_19.jp2
dda9d95360580f9b1ce4233ba23d0a5e
1d2ed1fd33d16673f4c78c97189101f7cd02d59e
27881 F20101203_AAAAQI almatar_m_Page_16.QC.jpg
2e0090cdf90569dcefee9c8fee8d8e93
4016590ec846706b4cbcbab0765a8be9e8dba2cc
21007 F20101203_AAAABQ almatar_m_Page_04.jpg
252954a678a0240d1298a7e6e561f7be
c84986ebc6933523d1031318d96bdc17e4008301
51901 F20101203_AAAALL almatar_m_Page_30.pro
a34a3981e57a1e7b9082ef24a9e7a689
6676e0f802b1043506a6ab7ac3643967aad599fd
6478 F20101203_AAAAQJ almatar_m_Page_16thm.jpg
966b1d30edf8cd450682e3a5807946f3
a1ccf64e4798e735ea01317f285f487f7bd7a3c2
1095 F20101203_AAAABR almatar_m_Page_02.QC.jpg
66bc36094944a138446804a927bcdb0d
85454227eb4689cd3f90692132e13d21f22aad0b
54761 F20101203_AAAALM almatar_m_Page_31.pro
b2a94813fdc9744e542dc2b78ed6d9b6
3e10cc25f6a6ac99b4f79bde6818efaa3a715707
1051922 F20101203_AAAAGO almatar_m_Page_20.jp2
a39193e605e57b3ec5fa409d455752be
e798df82f215a118b39c828202e71c90d850d9aa
26186 F20101203_AAAAQK almatar_m_Page_17.QC.jpg
c704950208bcfbb55cbfaa8e956d1514
47d47c25829cc37d977cb217f088985decb199e9
F20101203_AAAABS almatar_m_Page_10.jp2
6a24cbae00fc2af10761ff517e2869fb
b9ead092fbb97beeb2c127d46fddc2d74a46dd1c
56560 F20101203_AAAALN almatar_m_Page_32.pro
2b1f2cba7181297e55913918c44b3a83
dce0b183f635e59f5603cbb4e391a403149d3be1
F20101203_AAAAGP almatar_m_Page_22.jp2
0027644729845f5cc16aef3439bd205a
78c1da48170dc99aeafe39d717c1468b98a459f1
6329 F20101203_AAAAQL almatar_m_Page_17thm.jpg
1aafaaef6dc9d94652476f6ca14508aa
bba8c72add186e946351142cbebb03e2ab8fda77
85172 F20101203_AAAABT almatar_m_Page_67.jpg
a1993877bdfc5357f4d8efb223ef06a9
1082b9ef2c13918babde1a5423e11375a04297f9
54764 F20101203_AAAALO almatar_m_Page_33.pro
29f0aeae272394b287a650b62fa7e997
a5119d5d4d4fc78fb2090fd0885e68dbf4fd0c42
1051967 F20101203_AAAAGQ almatar_m_Page_23.jp2
97aab5df9f133cc10965de592688590a
61c7af166281ed9d17bfc60846248cc7cfcf91d4
26799 F20101203_AAAAQM almatar_m_Page_18.QC.jpg
cd5f85c0937b5093acc723815cf3cd74
821dff6a10bdf3597784fc2e6849d98740311d5f
1015319 F20101203_AAAABU almatar_m_Page_43.jp2
7da5c6f96c632e79293277d7f212c4d1
d583174b3c2125effd43c186e229806a0e319206
55331 F20101203_AAAALP almatar_m_Page_35.pro
5fe0470e1475d1ea36a170c138bd3876
4ed021ac84b5f57d695640111d2a2c17ca1fd54c
1051981 F20101203_AAAAGR almatar_m_Page_24.jp2
c6dee6a87f4da7e8e85f91a1a2fa0727
9c60323d05ff6eaffa37b6db5297bf4432a8a9ef
6312 F20101203_AAAAQN almatar_m_Page_18thm.jpg
5b7aef718f30c63d8fccc1e39d0ace0f
d14597f91880d450628c76e054b1d519f2ceeafc
30386 F20101203_AAAABV almatar_m_Page_42.jpg
052b5d14b7e947838fc9c2c138fafe66
6e301a2424a5cdfe7cb3272daf0eb95ed0cd1dc0
56482 F20101203_AAAALQ almatar_m_Page_36.pro
76f4441d18e9139064491e9483a1cec4
d8036945f4b742b1a6567c11de82eb7b8815e4ca
F20101203_AAAAGS almatar_m_Page_25.jp2
b3c7cd63aa835db5a18a350de1ce6dd7
1e574825c6746a9887da9ed0e3bda0e0162a9707
28088 F20101203_AAAAQO almatar_m_Page_19.QC.jpg
8870e846b3b5742b7a7d2b0af5d4ef68
957c90568b97a1c39cafd40a0cd3b9a7d30c42bf
5835 F20101203_AAAABW almatar_m_Page_15thm.jpg
134ef90449d162d370db19073e175ae4
03f3e0b345f57fa27f2c75858e0c1c44c4086356
54265 F20101203_AAAALR almatar_m_Page_37.pro
c08d9b6ab1dc84f1503e5e6a92be2212
4b6e1fb0775e1b42a95ed4cba6ed291bb091817e
1051933 F20101203_AAAAGT almatar_m_Page_26.jp2
6e6166041ab4a728b58ee95957121ba2
0f4e193ddea94a773fa098f213a43a0db90a79d3
6714 F20101203_AAAAQP almatar_m_Page_19thm.jpg
410c3e8d51e73b3f43d0e6cfd7a2e80d
5020761891bf9a6ea25bff0720cd2fd192ab0282
F20101203_AAAABX almatar_m_Page_15.tif
c1ba2d178737c4d564a4206a8d3a00fd
81fdcea2d32d7236a34453f7ee307e52c95e0e7e
55571 F20101203_AAAALS almatar_m_Page_38.pro
db446b2169774070c5f308916eb8d694
5760ad32b419cb9367519ddf855ac996cbdb4765
F20101203_AAAAGU almatar_m_Page_27.jp2
9669ede162aa489a5763ba45cfb81475
06152b76876ad05b57364e1e35c97487287c9e84
26926 F20101203_AAAAQQ almatar_m_Page_20.QC.jpg
6a2da3488313f5b8e872bca8fd8bbc84
dab50a90e2d0d03d8469f1d7e04b67ee259eed9c
26868 F20101203_AAAABY almatar_m_Page_30.QC.jpg
8578afe4fb6bebed924fd7bca5d9c66c
55096c471d42843b02b50a1ec1272e640119d7fd
1051920 F20101203_AAAAGV almatar_m_Page_28.jp2
c604653040eb446cb1494231747d9308
1191d42ed631bd1df4d5e3ba231c6fc17e39b8c6
25733 F20101203_AAAAQR almatar_m_Page_21.QC.jpg
56d3ebfee870b528be27e4a7a62bc53e
351b4cb5d33b891472d943ab3c810df33bb20ea9
43151 F20101203_AAAABZ almatar_m_Page_64.jpg
3e5fbde32674dfdf86e08d96b938809d
72e7140d2bf48c113469bf0b27382d4fd18a3384
56997 F20101203_AAAALT almatar_m_Page_39.pro
838e2964e24edec1fb05c0e435989e67
82279a19f320d82fffdd7b5537bbc87cecac70e9
F20101203_AAAAGW almatar_m_Page_29.jp2
b9ab75c164aef25efc6179c2c928a18c
7243e21e513823bac5e2912fc7dd275e239f6aa2
6233 F20101203_AAAAQS almatar_m_Page_21thm.jpg
99d7ede1f25acb4c0720cdd209c54cd7
4bb82acbff6f2e52f3655e4e812bdfaaaed4c5e2
55553 F20101203_AAAALU almatar_m_Page_40.pro
fda8bdec7ced1f839a47d5d899abfa26
62b127bbb91a2776d94b122e297bb13ae3e644b5
1051986 F20101203_AAAAGX almatar_m_Page_30.jp2
fe08cc07b61bdde18f09f23578d74de7
6ab2eec052ad5609db6005d2cb57dae80a9a2f3a
26387 F20101203_AAAAQT almatar_m_Page_22.QC.jpg
5b0c71edfac0b3fa0e52c68609cd9648
7f47eb0d9f113e083a61f038845a86cd694c90b4
16354 F20101203_AAAALV almatar_m_Page_42.pro
33ee5861897f28c30eae23cc3389e045
79362ee975e3d682eebfd4b9a714d478a0618be9
81839 F20101203_AAAAEA almatar_m_Page_15.jpg
882146b128b80074c15de54264499828
15d5dc027bd4482402f23172e64d28bfebc27a46
1051961 F20101203_AAAAGY almatar_m_Page_31.jp2
47b14b154a45b072dbb0278aa1678393
9fa0a8c24b6a2bdc2cfb1e4b7f85e6925d2d0229
6239 F20101203_AAAAQU almatar_m_Page_22thm.jpg
9d95ffa263083967e5c51c8d79ffe03e
3ceb31743fda24b52fde0efab44925e8a59c3bb2
48495 F20101203_AAAALW almatar_m_Page_43.pro
3091ac27bc6118a841a6b59ac3fa5ea9
fb7fb22d19eb8b7f32d26e7b5dff91cee2929846
87886 F20101203_AAAAEB almatar_m_Page_16.jpg
ca719e9217baa1e4deb68dbe4a5d218e
f7e1d37b5b11f9c4fa9409a23f819dd08bb224dc
F20101203_AAAAGZ almatar_m_Page_32.jp2
498789adc2a24d14b4536a7dad07dad1
21d2f67f1902a57251df784b405e6f590dc23180
25788 F20101203_AAAAQV almatar_m_Page_23.QC.jpg
8911b269f5d24fab4c0e8119178a2e26
c72fccd2122450f06a2ce53b7005acdf6b9f5319
51487 F20101203_AAAALX almatar_m_Page_44.pro
9d572564905f6b57ac8f545b4a29194e
d60e176968731840b8957267ea0254f5b09166b5
84835 F20101203_AAAAEC almatar_m_Page_17.jpg
90bdd0ad67b30638ce51a9fa193e22ee
affd60be65c482f668da1c0b83ae420efab86739
6303 F20101203_AAAAQW almatar_m_Page_23thm.jpg
639814e366ba6afeaaefaa464674c844
b3fd88ffa68eaef355d597b57d8cec87b07978d7
20635 F20101203_AAAALY almatar_m_Page_45.pro
adeb53a13d70e8f73cca7f9657c8ba0f
a400135f9eafaeded6fde4a7b6612133fdd34742
90927 F20101203_AAAAED almatar_m_Page_19.jpg
29a48b7348cd7754848ca7c7b104dda2
c0a48bc78928853ee3adf63d58d341a4d7493e37
F20101203_AAAAJA almatar_m_Page_27.tif
979b4bf3c4a645319f65c2cf226e8325
0b807f91e73fcbb33b4b68a8c2fdedccc607de8b
28236 F20101203_AAAAQX almatar_m_Page_24.QC.jpg
3de88c2bdefd655810d0601a99238dc1
83857df2f0264fb8978194b891d238e493fbe441
42760 F20101203_AAAALZ almatar_m_Page_46.pro
afb95304cfb0204b5bbe89a73ba5947c
374d8b91779a77627a0441f356389fd915f589a8
86863 F20101203_AAAAEE almatar_m_Page_20.jpg
091608ca821fa18b3608069aa2dee8ce
8171037a11a112ceda5e0ad2d09c8820f783933b
F20101203_AAAAJB almatar_m_Page_28.tif
26c41330be74b5d03bc122c684fbe43f
0158b803a111002717e7c34e8ec11749b2c56bbf
84439 F20101203_AAAAEF almatar_m_Page_21.jpg
d2c24b14e64a692a04e8b32bb34a76ce
07aa8a443124130b245a357bec716dce666b02d2
F20101203_AAAAJC almatar_m_Page_29.tif
46770cadee82c93a47e6b0c3b6579e8e
db7227e7f2f8e57682bd3c35be9a974bab138acc
6429 F20101203_AAAAQY almatar_m_Page_24thm.jpg
a48188c9a04d6da55cfc4882e653ce4e
41627c0eda3c2bc29d0cb7a6c88712638c0fef02
83130 F20101203_AAAAEG almatar_m_Page_22.jpg
52d547fbc0b16946a4febfed0f879cb4
77aa149832132e10b9bea3ab9c0ad3512b91bb74
F20101203_AAAAOA almatar_m_Page_31.txt
08b0f6acc94a9b93a0e832eab40dde59
13bd3795d5ad96517bad58b9719b1f32c223e76a
F20101203_AAAAJD almatar_m_Page_30.tif
b2fb1f29c0cb5fe6c0347c0a81858281
13bda869f0490cf9bb9bfb238ef0ce7daae97081
6709 F20101203_AAAAQZ almatar_m_Page_25thm.jpg
3804d1f80cacf8406f1d3fe9e118b919
1da53ae3afb92605df1ce9beffc0e3cf00816633
82909 F20101203_AAAAEH almatar_m_Page_23.jpg
d5daa192daca07d07e20198bf072817d
be19760836526e1d47dcdf3f4470653773f6e2a1
2212 F20101203_AAAAOB almatar_m_Page_32.txt
d5c60bbf9e4001dc56fb62368eb1c473
0c499770d72ba51a831725cb8d4498784fa49e92
F20101203_AAAAJE almatar_m_Page_31.tif
625d530bf40695beda76726ff8dbf15e
4681c1de7fa051101ef3d5c466ce055f11297e7e
85378 F20101203_AAAAEI almatar_m_Page_24.jpg
bc40565db58de32a15df4f5a9de8ddc9
b05011949a6a127ca9d7dbc9acd9c90e7780c268
2095 F20101203_AAAAOC almatar_m_Page_34.txt
eadc34d69924d834d43a2c9ec3682473
c8d1ef5cab84903cc6450484ff4e0da894e7cbdc
F20101203_AAAAJF almatar_m_Page_32.tif
965de4aa3d06999dab131e35cb73b544
d06f5511f39f48ee7bf278e75ca52c7a5a22291d
2613 F20101203_AAAATA almatar_m_Page_57thm.jpg
03a8413740ed500af05ad4d41dadd37b
9a59c1a9517b539a8bdfec53508fbfa4997165cf
89615 F20101203_AAAAEJ almatar_m_Page_25.jpg
8190e7ff235d7eaad16c4ce49433193d
9a7006c97e19c778c5420e9fd8af186f8f582793
2167 F20101203_AAAAOD almatar_m_Page_35.txt
d5847f2767a5b9d035acdaf8b4dd0a3e
5622f730b3516cec3b8478dec1700181747659ea
F20101203_AAAAJG almatar_m_Page_34.tif
4f93a9d9bb9dac60b6231994398194a7
1131d9d577a16ad14b1a410ff73fa5daccf6c69b
13861 F20101203_AAAATB almatar_m_Page_58.QC.jpg
bc932f888275770059501fada4cecbe2
a6204d9e8676e9b0057483dda93c1a1f7ae83e10
90458 F20101203_AAAAEK almatar_m_Page_26.jpg
05ba1de700e092e5ef65e8a73d446763
dd1318cf1fe6cc0fbd5e59fcf8be5ba58ffcdf32
2210 F20101203_AAAAOE almatar_m_Page_36.txt
92f8a1d711903ce7856414ce21b11a75
06dc6c51346ea03de6309dabd7c68f4796f4c7d3
F20101203_AAAAJH almatar_m_Page_35.tif
f422aff8d3cd0b5c3822ff1b1aa2a258
3204c94e94047b6c06912001815b6c574648a449
3921 F20101203_AAAATC almatar_m_Page_58thm.jpg
616e8d5cd3f0ea680a2490fe02ee8433
15aaed699014fa3bfed8cfc96f79b5e1ab6d79ff
87136 F20101203_AAAAEL almatar_m_Page_27.jpg
b64d9115b6db436c6a5f5360ca6ab35a
a262ab7524f7830394a96dcf488caa2a3035a0ee
F20101203_AAAAOF almatar_m_Page_37.txt
07c1ee27f294fdb46d0723c347217489
4211fc0f0d2a8469f0b6b46f112169d1cca78223
F20101203_AAAAJI almatar_m_Page_36.tif
c9c358be16f354fb8464019bcf28c2e0
1fed85cd9993bc30ac78be0b6d3d744f8a282383
8667 F20101203_AAAATD almatar_m_Page_59.QC.jpg
1157cb828724cd5f7d1196a145078af3
522f2ab7bd8caecec209048b22665073e1d0cf10
2176 F20101203_AAAAOG almatar_m_Page_38.txt
c9a5d042e1a68360d1c7bc6a5104e5a7
7a425f6a8e0d3285bbc86edf7b89e714c0cb9ee4
F20101203_AAAAJJ almatar_m_Page_37.tif
e295811245494e8f2d6d66e163559339
edf7f450e115c2dc6c1984dda0e9ade8e9cd39fd
2334 F20101203_AAAATE almatar_m_Page_59thm.jpg
4213116fab7c62b35f070eefdb3efb63
c7977c9aab929ce6ed9a6d58552d18e6930effd1
85603 F20101203_AAAAEM almatar_m_Page_28.jpg
988fa842abf762d08e6ef3dff5da5abe
c48c95bb18e52f07b348bad6095942f50513b41b
2229 F20101203_AAAAOH almatar_m_Page_39.txt
6c8b2e65a150ab1509cdde3f1077c13b
1c5d31d33e7f2721df61cc8b51a9084d7e0f1c94
F20101203_AAAAJK almatar_m_Page_39.tif
349e61e9890c56cbef08743b23cfb4ee
35d4314eb454d34447c70951813367f2ea29b764
18142 F20101203_AAAATF almatar_m_Page_60.QC.jpg
389988e518d856eddac712e5bc39bd85
4cfb307a62ee2c75e817b26d5d9d0d7790a78592
91757 F20101203_AAAAEN almatar_m_Page_29.jpg
bbbd8af38b5b65a2192bb54745446eed
aec774e7816ccde0c1cc0baa7e45b607f220904d
2189 F20101203_AAAAOI almatar_m_Page_40.txt
61def2c038699d2837d702c35af0e10d
0ceac3e9e6eaa348421fb63f27728fcb333e0c2f
F20101203_AAAAJL almatar_m_Page_41.tif
ac871c3c044f294d5bc20c99e20e7d2c
2f1540cb26315073626e5c03c350475bbb512ff1
10363 F20101203_AAAATG almatar_m_Page_61.QC.jpg
556034e33018001a259db85e0823cda2
345e0225b6bc89af96b472dd35ec097ba11bb9c2
85907 F20101203_AAAAEO almatar_m_Page_30.jpg
c0f282ea51bd01f0a25db3a09cc36dea
3c0ed77a0e747e8de3eb49a24fd302f5336b1f43
1539 F20101203_AAAAOJ almatar_m_Page_41.txt
60123cb454d778174ecd7f4bbb78de36
e021cdda22cbdbb6444cc9946959ca599b67c043
F20101203_AAAAJM almatar_m_Page_42.tif
4427ad9582ad11e6a13208fc76b02ae4
afb044ee58b2bd19c2cb60de3fe5777f701ecb08
2919 F20101203_AAAATH almatar_m_Page_61thm.jpg
081af0d1427f905964bb4d53638fec4b
9b9023f80f505c2b914b5ba002026946dc84cea1
89846 F20101203_AAAAEP almatar_m_Page_31.jpg
6e8c04be338533d245cb83f3b3a46922
9502d71cb6be65c7cea262d6e7bb8087f4665daa
2398 F20101203_AAAAOK almatar_m_Page_44.txt
4bfc08ac3e4364dc416cb277ac30d149
dbafcee9915f536d1692163664b362d5ff201b59
F20101203_AAAAJN almatar_m_Page_43.tif
7803d83c240a36da837ed5b893897c49
319b47f413740bd73101ae8bc1dba07baf264788
22995 F20101203_AAAATI almatar_m_Page_62.QC.jpg
9e52832573c5537cda9ca9c940c35e53
9579583e7b86052a99a150743e2aea3b42ea7820
90171 F20101203_AAAAEQ almatar_m_Page_33.jpg
e81497da3e0a890deab4ff2d6be281d2
3bb08e592cad1fcb889d4f0c9a6c6ad4368e18a9
976 F20101203_AAAAOL almatar_m_Page_45.txt
a1532a99739af9deb30910eec7c73ea8
4cffe4ce4427137cd165510641bf7bfbb9629e34
F20101203_AAAAJO almatar_m_Page_46.tif
9cc19e32a0a3ee80fded4e4bfabb14b4
ce1479fb7ac84175930ffa59ed58323c128f1bcf
20645 F20101203_AAAATJ almatar_m_Page_63.QC.jpg
0e81f06bedb5d175a86878e2753e1f91
e965c1ccbbb64a4748ec558b028092008c20497e
84897 F20101203_AAAAER almatar_m_Page_34.jpg
dcf0f32ce526aa511a6056dd6f6927f8
c0c025c0186e3efab608aa1bffc926a16c025089
2182 F20101203_AAAAOM almatar_m_Page_46.txt
9aa71207bccf714716217fa3106c6009
c0ef8cbfbc7523c22a8d26b48583bf3a254bafd9
F20101203_AAAAJP almatar_m_Page_47.tif
cd56fa42202eee1f7d98b08cfa4f1e78
d3cb3ebac9f3c582e0f0a9fbc88b910aebae9239
5653 F20101203_AAAATK almatar_m_Page_63thm.jpg
cff2e1ccb6648a717d5e60888fce3789
49f4905f36a3237d9da7a444396b22c499a19e89
91283 F20101203_AAAAES almatar_m_Page_35.jpg
d88aa8b0b787ae9fc07dc0b318a709f6
e626760d6df61323eb66addff0206054b04453d6
2162 F20101203_AAAAON almatar_m_Page_47.txt
c466237cd7232aeebd1f33bf21571002
d4e108dcbadb301852e6ce799b00528c4093e27b
F20101203_AAAAJQ almatar_m_Page_48.tif
1fbdb50615e18043645935a4800bd0ab
ec426bdd92df830e10cf9a1671693de9b6a2eb24
3355 F20101203_AAAATL almatar_m_Page_64thm.jpg
bb73c8789f2fb9772494f14bf6056556
d8e35b91e1551bd404fab60a41bab585723e1d87
91850 F20101203_AAAAET almatar_m_Page_36.jpg
b956ce18c9f4f2c717d43c001dcc2420
18c61f9032ac2379d33a5d93ec0183358e6fd237
1964 F20101203_AAAAOO almatar_m_Page_48.txt
f4415f1fccf4b178867e74fd4b82bbf6
39253bf4540f39b533b8455a0ba7b277f0376a9c
16800 F20101203_AAAATM almatar_m_Page_65.QC.jpg
25b6899bd7e2f10e9985cb217f917818
75610f67cbb096b836ce16034a3e33f0e5c843a6
89725 F20101203_AAAAEU almatar_m_Page_37.jpg
73ebbe7e920405760438f073616bcb57
b08453ea3bb2a578fbad71bdcb5bbc3c12755eb5
2533 F20101203_AAAAOP almatar_m_Page_49.txt
d898070ba0978ee686c155010307459e
1cf645a5e7ae9e21d8c817b66682f839116ec8fe
F20101203_AAAAJR almatar_m_Page_49.tif
85ebc756b8b7fc6c3d0f623de202c9fa
65c164df3d07ccdfb83ffdcd70176c63b80d737f
4508 F20101203_AAAATN almatar_m_Page_65thm.jpg
04d66f22499bce563c2c58933549f783
3204ffd0442cb54510d3575fa15c43dc68b410aa
92226 F20101203_AAAAEV almatar_m_Page_38.jpg
570760e0250b1cf59a4bc9e333122adc
4dec7de81bdfcb2ead9ff75d46e6547dec661631
358 F20101203_AAAAOQ almatar_m_Page_50.txt
0c8e60e5c832ab17cb797c79df2d01d4
e84575ccb7d9f7a76a63792055f0b5400f89dfc9
F20101203_AAAAJS almatar_m_Page_50.tif
b345733d671bf94e4a8ad8725c7c74dd
efa9a430229d3430222143dbbe4d27fa653b2142
5504 F20101203_AAAATO almatar_m_Page_66thm.jpg
52e178993014d972fd75631799a5af6c
9f05a7a452596ee982b5a820db9dc5d7de8b71c7
93630 F20101203_AAAAEW almatar_m_Page_39.jpg
94accff394ec6d5488355c4b7fbc8daf
3c879050d34c5dd9fbb18e6e153dae0684112f6a
1093 F20101203_AAAAOR almatar_m_Page_51.txt
453c899b11c31f4b3a55e39b183d7ecf
90e739a6bcdbbb7bc59c9d9304a46ede7934b23a
F20101203_AAAAJT almatar_m_Page_51.tif
600fcfa013c32809aa8322ff568f7b89
43c7d6bde8a36947327ae9b8482318cf7308dc0d
26376 F20101203_AAAATP almatar_m_Page_67.QC.jpg
5b022037502b4b3d54312960ef35de4a
a58cb34edd02a5cac15718d88d997b1838b32165
92092 F20101203_AAAAEX almatar_m_Page_40.jpg
9d860c01766922a0033c89fc68f958a2
c51b883e1c659fad5828ea2b808be5e2c252a416
570 F20101203_AAAAOS almatar_m_Page_52.txt
b7dff11c5459b967c0ac7a1083d07143
c18f4953f45217970b80a0bd2c67f581d1f03c97
F20101203_AAAAJU almatar_m_Page_52.tif
c42d9b63cbee3704a83540ed25eacf9e
d9654366ce9f78af093dd4594eeb8d7691081e9c
6295 F20101203_AAAATQ almatar_m_Page_67thm.jpg
ebcaadf1f1dc6fa2730eaeeee3075b5a
610def79fa97c0a262336827ce90acdd026d990c
64828 F20101203_AAAAEY almatar_m_Page_41.jpg
5a4a275e7b0a092d36d4ad319284bc8d
4d5468c3fbdc2b17fcde95bc3393e760379d894e
373 F20101203_AAAAOT almatar_m_Page_53.txt
174b00a9750f5aae8fe218dc80c6aaa0
e68191313b66e491bba6033a0e768119d0839864
F20101203_AAAAJV almatar_m_Page_53.tif
cf0e32284a2791a51b6e3b62fe533bc1
6a4df8f047df9cb7ddc410269e8be208fbd96e67
93767 F20101203_AAAACA almatar_m_Page_32.jpg
384ff01c399eac74b140ffc8bd5ac0ec
d277cf2b9dcfb9aa01ad86a2c9c3363cc5534b88
26708 F20101203_AAAATR almatar_m_Page_68.QC.jpg
01fc698338f7b89b3ae798d60753980a
31146d5e7a79afbebab6abf632a2cc53265eead1
F20101203_AAAAOU almatar_m_Page_56.txt
3c49fdc214ab1b7fe35ec76fc9130de4
e5782c384c20b1226e6edc5c6fde7e4a77f1ef96
F20101203_AAAAJW almatar_m_Page_54.tif
533546b83eeff879f14854efa67a0985
1f2d09a002ee348b3664f9b2fc1b49598cc8dcc0
33699 F20101203_AAAACB almatar_m_Page_61.jpg
07c1c98e900151b2462d0109d3a25dc1
07b3b80f1c3b0dd8e47bc93d709957711162e97b
52925 F20101203_AAAAEZ almatar_m_Page_43.jpg
1ad4c04b0eeec3c2fc55101e80ed17fa
6b14789738dbb51453d68209c87a5e7147cba898
6343 F20101203_AAAATS almatar_m_Page_68thm.jpg
560f1be2d9e5e4c3d6143bca95d8dfa7
5ed810ecc814382d3896b56b8fa609aa78e4452e
172 F20101203_AAAAOV almatar_m_Page_57.txt
ac283e7545f5bf7828ed55afea3f1b2f
90e92f6785ac97dcb33f88b41511b5477ddef6bc
F20101203_AAAAJX almatar_m_Page_55.tif
cf15cd5b9ff19865302db33d5b089f73
342c8c0f702e601d54840a5db16c34b64e8d8a04
2419 F20101203_AAAACC almatar_m_Page_70.txt
f1056701ba4f1d334df56b2784476e3e
660646cc4032306218637db695a9e4beb9a72191
29582 F20101203_AAAATT almatar_m_Page_69.QC.jpg
c2c7e3e0eece725082ba4b22c94bafda
10e723d05ae07f135b10fd2d48be11a1df362c77
1051942 F20101203_AAAAHA almatar_m_Page_34.jp2
98ede4e57af806adafe6020cef248e7c
474e509f5d38de06eb865a793d0f3fd5a160d885
F20101203_AAAAJY almatar_m_Page_56.tif
3524fd514dd2ab5227095b7399fb1ad4
f1fb72a353b1fc97f03cf1be8dfa9633455324da
1051974 F20101203_AAAACD almatar_m_Page_33.jp2
bd0de0956d8cc0b1b47f856f910c9f1a
9bf3a2ffcc931b5525d337b66a6512fd7893f8e0
7116 F20101203_AAAATU almatar_m_Page_69thm.jpg
cefb764a401f0cfb5ec76b2c36691092
e556f60414aa2dd9f12b8d5f958f8d5a008b20e6
1419 F20101203_AAAAOW almatar_m_Page_58.txt
76c0b0048f3b64e7ab2b46cbf6fe9a2d
515228c3594473ae3bb9bd37196c48497e47985a
F20101203_AAAAHB almatar_m_Page_35.jp2
b58a402733899889f699cd5b2b255fa1
caa55ff63c5d4c53b0bb1d6b6ac6aecdf89b8b07
F20101203_AAAAJZ almatar_m_Page_57.tif
802900cf2b5f28b869bd0521daea6c90
8b48cb0365f40ddeb65f19e15dc9ae20754b8155
3568 F20101203_AAAACE almatar_m_Page_02.jpg
cc211a8ecc2b33aee974fb35613d4f68
7c2337386e750586e4a0da8ac75515124cad6d6b
29068 F20101203_AAAATV almatar_m_Page_70.QC.jpg
688dd2234d3a0fee404fe0267c5d1589
75ed58306040942115c1b6ca2a2968e24a1944ca
1514 F20101203_AAAAOX almatar_m_Page_59.txt
e22644a2dc340286e7f28fa0c582ef93
eda989008d43b538f73f18e2689723503940f751
1051978 F20101203_AAAAHC almatar_m_Page_36.jp2
ac7a43d9ed9e808f9be407fe7535794e
2d5078176728b9c3cd9dcc24485ac95c46a1ffe6
F20101203_AAAACF almatar_m_Page_15.txt
75d481d0ccf04da9a1dee0b411568269
da151b98406d441f5c3c26eac36834a6a48768d9
6995 F20101203_AAAATW almatar_m_Page_70thm.jpg
748733bed3003d42f57a87ab42418735
7b841d5416ac0c32e11a7f0591c3c5bdb6c18757
367 F20101203_AAAAOY almatar_m_Page_60.txt
53e7454d21c34adfd7f011a190566315
5d178ff86f5f74687a75f71cb710650fcab81605
F20101203_AAAAHD almatar_m_Page_37.jp2
b9644476de1bf021420aca804a0893c9
420354ed2b7709c59be3fdf8c45a443b296ba66a
6052 F20101203_AAAACG almatar_m_Page_62thm.jpg
492e4b6b063a9264b7170f4bbbdcc4f3
0db97ceb1bab0ffe452e7bbcc807b674203c1429
42358 F20101203_AAAAMA almatar_m_Page_47.pro
eeb179180cc23747fc03660265e5f42b
c142e343f186548e2fa1ef4999b0f8f8ee548578
33063 F20101203_AAAATX almatar_m_Page_71.QC.jpg
30829ae1fa369f5e21665281cc84e95c
3d858450481db96233a8e67094ebcfb2e7348622
765 F20101203_AAAAOZ almatar_m_Page_61.txt
297b35152629dc6744962bee698fd7a2
66d9120190704bb8fd03f75f7cdab85c3fe744f1
1051930 F20101203_AAAAHE almatar_m_Page_38.jp2
69c209a9a1e18f3413a442058c2eacdd
679cd63bdccc7e0d255a275686688201ecb86539
F20101203_AAAACH almatar_m_Page_63.tif
e4ff8024603334323f54ccdb3b8517bc
09cbcdf508539275231877695d66c406dffbe999
36255 F20101203_AAAAMB almatar_m_Page_48.pro
db77f03e76639a986d72f5679a10223d
aada0f0bdb000c09825cd9880a131e3768235c26
7566 F20101203_AAAATY almatar_m_Page_71thm.jpg
5ea799221f40b529bf8f10dc7a6b23cf
0123dd9f1bafb5e7c7223a48b8778cd1adb591f2
F20101203_AAAAHF almatar_m_Page_39.jp2
fdff4901694220dfedf21db61e9379d2
17dad078fcd1ce9871ce446ad3428e156f43e364
6407 F20101203_AAAACI almatar_m_Page_20thm.jpg
56b921bd548b202a8858ea3bdefa20b6
f02e7c37632e5c53dd1d4bc5d9661d58678a6aaa
47873 F20101203_AAAAMC almatar_m_Page_49.pro
c9b8423b633ad9a6c43c1d6764c83923
24cac616af6bb4cf947334c68c02a01fe22c033d
19619 F20101203_AAAATZ almatar_m_Page_72.QC.jpg
0886f8ba18b9b3e1bd731335289fe4be
2f74da1c380ad327ee285f0bb7c8653812fb5b84
28534 F20101203_AAAARA almatar_m_Page_26.QC.jpg
d0d8eb5f5218975f80dd11c4fff15b6a
ecaa2177519b0fc6fd5877e682edc107c3111ade
368425 F20101203_AAAAHG almatar_m_Page_42.jp2
008b0964dc46751ebb899c8c5b11620d
9c5b89573e5a7ccf4ace3fc44aada36c426c42f8
F20101203_AAAACJ almatar_m_Page_45.tif
ff2c78f3ab3709bfd372936b7570a441
c1402c98d6780f7acd13666e4d0c08474ad54fc3
8696 F20101203_AAAAMD almatar_m_Page_50.pro
c605691e5d55fd17fa70056ede275700
56a9d040fb71a74e03aa3fa7df975176ef2705fd
6453 F20101203_AAAARB almatar_m_Page_26thm.jpg
0214f0b95b61722ded64c6c73f4019be
23b2b75c0e1485917c5b75a8414f0af0be78575f
1051947 F20101203_AAAAHH almatar_m_Page_44.jp2
82ecac1578d9383ccbe0b8b6b7ba22c0
4f9129a4a403671ee39eea27e86b08a387c47e11
14871 F20101203_AAAAME almatar_m_Page_51.pro
f1a4dd14c0cfe6ba4d3a17e9b5a859d6
292b094158d2a850c7aa6b4ab5ee16bd77096452
26802 F20101203_AAAARC almatar_m_Page_27.QC.jpg
b38b1f0622a004efc1b03245fbbc157f
108488aa6985bebee4795b9286555e7f3fd3af28
782650 F20101203_AAAAHI almatar_m_Page_46.jp2
fc03299c7af273d8c2b3871111c7d215
213131a18b8590166ce0d67754c2c97ad3cbc1bb
1051957 F20101203_AAAACK almatar_m_Page_21.jp2
51e3912bfb8487c5182d7e9d2391d819
97577907e994b5e3066467c3ef739a54a395fd4b
9217 F20101203_AAAAMF almatar_m_Page_52.pro
84f8046b8d78389ee540c4192513c58c
f745253c4f17e879567aa28d924e6d3590dde7fb
6591 F20101203_AAAARD almatar_m_Page_27thm.jpg
70689da551c085c9675a71a1c5bb746a
9f461ece530726ff742fc236e44b38435d886c03
782044 F20101203_AAAAHJ almatar_m_Page_47.jp2
4c820e4717d894647dc06391fac20263
b08ba1095ebc8c4325a32e651abbd1b719f70986
1582 F20101203_AAAACL almatar_m_Page_04thm.jpg
e8643a532ac99c4e2be17df8edfcdbc6
f9f9bc7526016afbbbf1ee3c5d55b8cca2949e59
5973 F20101203_AAAAMG almatar_m_Page_53.pro
0191d0340c7e877322b24b33b20a6bae
043e1260182047b94e13f44f3e5819b55518800d
26925 F20101203_AAAARE almatar_m_Page_28.QC.jpg
3ba7b0c57d2688659eb32f5da8a78d28
469c99700e43757799b1f1028e31b462930fbba1
740408 F20101203_AAAAHK almatar_m_Page_48.jp2
71b978e4db43e509c740eb20c0d2415c
0ec377ee5ba787384da478943a4e85c4c00443c7
102567 F20101203_AAAACM almatar_m_Page_69.jpg
5e4da9a22d4905506c7970dce45eca29
24f4b4c572df56661c8ec82afd77ec50726f705a
32771 F20101203_AAAAMH almatar_m_Page_54.pro
afb27f37e8dc6bbed3885cb4b30e0bbc
f359887ab6e9598fca85ce9ed8faba831bb8e80a
6467 F20101203_AAAARF almatar_m_Page_28thm.jpg
75b6407c09257acb3a196308346d9e9d
0fdd814c0fe24f6f772e15a7d10ef5b8b2aeeb61
977313 F20101203_AAAAHL almatar_m_Page_49.jp2
b5941c1b6a389c9b343e68104ce71b6c
eaa5b9dc7a6560e7278d94d8eb6756ee9cc5dc6f
F20101203_AAAACN almatar_m_Page_11.tif
52b988a01d68096c8d8bb82125e71a40
4f17c40a5baae9f46970e7e2932af6122d8d5551
8909 F20101203_AAAAMI almatar_m_Page_55.pro
ab8b3c0cf6fb809c4d102510b4cd2d31
cbfe43f3d251393dd084125b8eb7ac7eebac8cf2
29563 F20101203_AAAARG almatar_m_Page_29.QC.jpg
d8efb422e3e085d4469fd01ff61548d0
a1257f0540a9f2cca36c646ef210817498452a67
965124 F20101203_AAAAHM almatar_m_Page_50.jp2
d2e04255c0e5f9bda31c148e95ab1e9e
d4298457bb273aaa29d9f8d05df4e81c898395dc
52423 F20101203_AAAACO almatar_m_Page_16.pro
bdb0f47d0a6e313f4219f79ea0cd679f
a132c40791577dfe91ccbd0e68ed7bd1d48e3e67
62171 F20101203_AAAAMJ almatar_m_Page_56.pro
3b89804c30f05c0e8648f3d22aa6607a
93d68f32a5196d7b167215018565896a3b7ece9f
6594 F20101203_AAAARH almatar_m_Page_29thm.jpg
96aa1f9d3e4ef571fd79f2a140eb4fa9
de367f8aecd5dc910d7903eb30ef36a68e3bb6a8
668108 F20101203_AAAAHN almatar_m_Page_51.jp2
6719c3246fb53dcedae274434bdf5b68
91209bc2127c1acf25d14b24f3b976518f9fb3f7
1051943 F20101203_AAAACP almatar_m_Page_40.jp2
3ace452dbb412264426725bff6b6a26b
949b75ae5118b8ecc43536459d7013ffd8e22e45
3657 F20101203_AAAAMK almatar_m_Page_57.pro
bc8f95e04d08167faefc7f54450bca52
32e7a16fe98aa7fe472c331f71aa645200756d2d
6439 F20101203_AAAARI almatar_m_Page_30thm.jpg
1b1e94683a349920242947458d222f6e
c842281f6971fac594149f3c6c4f41b25f0afbf2
345408 F20101203_AAAAHO almatar_m_Page_52.jp2
62b37c4302306006108d894577d834e4
5c1ee51b878b47477418439879a2d3b8a597dadf
1461134 F20101203_AAAACQ almatar_m.pdf
781770eed09b331cbc57ccc937a6c5c3
90cb5298285713497e937dc03ef9a29f356a6278
12559 F20101203_AAAAML almatar_m_Page_58.pro
e7d5764c106733c6d9d9db6bb4940306
905e0b41bcd55362c966f5dea813db172e7d6d25
28527 F20101203_AAAARJ almatar_m_Page_31.QC.jpg
660d9591605c0546b6b89a5551725660
acb2e54ebdc975928478a42dce29743503530424
2201 F20101203_AAAACR almatar_m_Page_43.txt
24564cd209b26c3731951f19aedcc2c5
9ddb79dfb9a31d45d4601b92b39f5527ba2a5351
15423 F20101203_AAAAMM almatar_m_Page_59.pro
8bedae4b1ec97f377a587d6a50db614f
748d71c1e8b19ed2b6d1bec3262433b87c870afa