You are on page 1of 7

Paper

▌FORUM

Monitoring Urban Growth In Alexandria-


Egypt Using Satellite Images

Loutfy Kamal Abdou Azaz


Postgraduate Researcher, APL

▬▬▬▬▬
Keywords
Alexandria, urban growth, satellite images


indicate that the use of this type of high with good management, prediction and
Abstract resolution satellite data are very useful planning.
for mapping land use and change
The aim of this paper is to find out the detection analysis. The study shows the Objectives
size and trend of urban growth in need for land use classification for
Alexandria-Egypt using satellite Egypt and the Middle East. The main objectives of this study are: -
images. Two Landsat (TM) images,
acquired in 1984 and 1993, were used Study Area 1- To define the boundaries of urban
to monitor urban growth in Alexandria. growth, which will give a clear
The images were georeferenced Alexandria is the chief port of Egypt vision of the current situation for
(rectified and registered). The and is located in the north and occupies urban growth in Alexandria.
georeferenced images were classified a T-shaped peninsula and strip of land Mapping will show where urban
using a user-defined land use scheme. separating the Mediterranean from expansion has occurred and the
Two land-use maps were produced for Lake Maryout. (Figure.1). The city was size and trend of that growth
each date. The georeferenced, founded in 331 BC by Alexander the within the period of study.
classified images were compared pixel Great and was the capital of Egypt for 2- To produce an urban change map
by pixel to locate and quantify changes over 1000 years. Alexandria witnessed that will show the changes of land
of urban land use during the 1984 to a continuous urban growth from the use in the city within that period.
1993 period. This process of multi- beginning of the Mohammed Ali era 3- To produce up to date satellite
temporal change detection results in a (1805) up to the present time. In 1905, image based land use maps.
map of change: firstly of land use Alexandria’s 370 thousand inhabitants
within the specified period and lived in an area of about 4 km2 Geo-referencing
secondly with structural indication of between the two harbours. Since that
the boundaries, size and extension of time the city has expanded rapidly, Remotely sensed images of Alexandria
urban growth itself. The satellite eastwards and westwards, beyond its for 1984 and 1993 were obtained from
images were rectified, classified, medieval walls. It presently occupies the Center for Remote Sensing
analyzed and compared using Imagine an area of about 300 km2 and has a University, USA. The 1984 image was
ERDAS. The results of the study ten-fold increase in population at 4 rectified to Transverse Mercator
indicate the western direction of urban million, with a density exceeding 1,200 Projection, Ellipsoid, Helmert 1906
expansion in the city and the need to per km2 (Shouk, 2000). Population is using 43-ground control points with a
plan this growth. On the other hand, projected to become 5.4 millions by root mean square error <0.3 pixel The
the eastern and southeastern areas face 2015 (United Nations, 1997) Egyptian-series topographic maps, of
a problem of urban sprawl over the (Figure.2). Because of this, Alexandria scale 1:50,000 (EGSA 1997) were used
cultivated lands; these valuable lands is the second largest urban governorate to establish ground control points. A
are in need of protection by preparing in Egypt. At an international level, the transformation matrix was computed
plans to direct the continuos urban city was ranked 62 in 1996 and it is and tested and the pixels were
growth to the suitable lands. In relation predicted to become rank 54 by 2015 resampled to conform to the new grid.
to urban remote sensing applications, (United Nations, 1997). This enormous A new image was created with the new
the results obtained from this urban growth requires precise detection coordinate information in the header.
application of Landsat images data The 1993 image was registered to 1984

Monitoring Urban Growth In Alexandria-Egypt Using Satellite Image▐


Paper
▌FORUM
image so that they can conform one
with the other. Registration is
important when images need to be
compared pixel by pixel (ERDAS
1997) in a change detection phase.

Images Classification

The selection of the appropriate land


use/cover scheme was not an easy
exercise. A number of classification
schemes have been developed by
specialists who have inventoried a
geographic region. The U.S.
Geological Survey Land Use/Land
Cover Classification System was
developed by Anderson and others and Figure 1. Area of study
is a resource oriented (land cover). In
contrast, the Standard Land Use
Coding (SLUC) Manual or the
Michigan Land Use Classification
System are human oriented (land use) 6000000
(Jensen, 1996). These schemes use
hierarchical class systems as shown on 5000000
Table 1. For the Alexandria study, 4000000
none of these was used because most of
the specified classes could not be 3000000
applied in the area of study (with its
different environmental criteria). 2000000
Future work will endeavor to develop a 1000000
land use scheme, which can be applied
to Egypt and hopefully to the Middle 0
East as a region. Another problem was
07

17

27

37

47

60

76

86

96

15
the spatial resolution of remotely
19

19

19

19

19

19

19

19

19

20
sensed data, which in Alexandria
prevents detailed classification. Five
classes were therefore chosen for Figure 2. Population growth in Alexandria (1907-2015)
application: - (1) Water bodies
including, The Mediterranean Sea and analyst’s experience of the study area, classification using the ISODATA
Maryout lake. (2) Green land, but this failed to provide adequate Method was performed using five
including all agricultural lands separation between the target classes classes. Unfortunately, the result was
regardless of type of crops and their because of the small spectral variation not acceptable because of overlapping
level of intensity, this class includes between important classes. between urban fabric and Green land
also “ urban green lands” such as parks classes in some parts. Because of this,
and sporting clubs. (3) Fish farms and The ISODATA clustering algorithm is another unsupervised classification was
Salinas. (4) Urban Fabric, including all an Iterative Self-Organizing Data performed using a larger number of
buildings and built-up areas regardless Analysis technique that is widely used classes (16). The maximum iteration
of use or pattern. (5) Desert/Sand, in unsupervised classification. It is (maximum times that the ISODATA
including the western desert and the iterative in that it performs the utility should recluster the data) was 10
coastal sands. classification and recalculates statistics. and the convergence threshold
It uses minimum spectral distance to (maximum percentage of pixels whose
A supervised classification approach assign a cluster for each candidate cluster assignments can go unchanged
was applied initially relating to the pixel (ERDAS, 1997). Unsupervised between iterations) was 0.950.

Monitoring Urban Growth In Alexandria-Egypt Using Satellite Image▐


Paper
▌FORUM

Topographic maps of the study area Level I Level II


helped in combining and recoding the
classes reduced the number of classes 1. Urban or built-up land 11. Residential.
from 16 to 5. Figure 3 and Figure 4
indicate the land use categories 12. Commercial and services.
(classes) in Alexandria (1984, 1993).
13. Industrial.
Accuracy assessment for the 14. Transportation, communications, and
classification was made using utilities.
Egyptian-series topographic maps of
scale 1:50,000 (EGSA, 1997). 55 15. Industrial and commercial complexes.
samples were taken randomly from
each class according to a rule of thumb 16. Mixed urban or built-up land.
presented by Congalton (Douglas,
2000) which defines a minimum of 50 17. Other urban or built-up land.
samples to be collected for each land 2. Agricultural land 21. Cropland and pasture.
cover class in the error matrix. The
error matrix shown in table 2 reveals an 22. Orchards, groves, vineyards,
overall accuracy of 88% in image nurseries, and ornamental areas.
classification for 1993 image. The K
(hat) coefficient (Jensen, 1996) was 23. Confined feeding operations.
calculated to obtain the accuracy,
which includes omissions and 24. Other agricultural land.
commission. 3. Rangeland 31. Herbaceous rangeland.
Change Detection 32. Shrub and brush rangeland.

The selection of an appropriate change Table 1: Some classes from Anderson land-use and land-cover classification system for use
detection algorithm is very important with remote sensing data.(Thelin, 2000)
(Jensen, 1996). Many algorithms can
be used to perform change detection.
At least seven change detection
algorithms are commonly used (Jensen,
1996) but these techniques have
disadvantages such as (1) Parts do not
provide quantitative information on the
size of land use change from one land
cover category to another, (2) The
analyst cannot label the change classes
in parts (3) Some need a number of
complex steps to perform change
detection.

In this study, two remotely sensed


images (1984,1993) were rectified,
registered and classified into five
classes. The Image Algebra Change
Detection algorithm was used with
small modification; in the usual
method, the amount of change between
two images was identified by band
ratioing or image differencing (Jensen,

Figure 3. Land use categories in Alexandria (1984)

Monitoring Urban Growth In Alexandria-Egypt Using Satellite Image▐


Paper
▌FORUM
subtracting the rectified imagery of one
set of data from that of another. In the
modified method, the rectified and
classified images were subtracted. The
subtraction results in positive and
negative values in areas of radiance
change and zero values in areas of no
change (Jensen, 1996).

The subtraction was done using Model


Maker in Imagine 8.3 and the output
(change map) was tested and proved
using a Change Detection option in
Imagine 8.4 and the result was the
same. Figure 5 shows the change map
for the area of study.

Results and Discussion


Table 3 and figure 6 show the size and
percentage of land use change in
Alexandria area. Generally, the
changes can be divided into two
categories; (1) Increased change
(positive) which includes urban fabric
and fish farms / saline and (2)
Decreased change (negative) which
includes green lands, desert and water
bodies. The summary of these changes
has been put in a form of “ from-to” as
Figure 4. Land use categories in Alexandria (1993) shown in table 4.

1996). Image differencing involves

Overall accuracy = 242/275 = 88%


The K (hat) coefficient indicates an overall adjusted classification accuracy of 85% for the 1993 image.

Monitoring Urban Growth In Alexandria-Egypt Using Satellite Image▐


Paper
▌FORUM

Classification Green Fish farms/sa Urban Sand/ Water bodies Accuracy Row
land fabric desert total
Green land 47 0 7 2 0 85.5 56
Fish farms/sa 3 46 0 0 2 83.6 51
Urban fabric 5 0 48 3 0 87.3 56
Sand/ desert 0 0 0 48 0 87.3 48
Water bodies 0 9 0 2 53 96.4 64
Column total 55 55 55 55 55 275
Table (2) The accuracy assessment of land use unsupervised classification for Alexandria remotely sensed images

Based upon the analyst’s experience of


the area of study, the areas of change
between 1984 and 1993 (Figure.3,
Figure 4 and Figure 5) may be
summarized as follows: -

1- The New Dekhila Port,


established in the west of the city,
covers a great area of the land and
areas of the sea.
2- Sidi Kerir oil refinery station was
developed about 22 Kilometers
west of the city, and occupies a
large area of the sandy coastal
strip.
3- A great part of Maryout Lake has
been drained for development to
accommodate urban expansion
demands. A big fish farm was also
developed in the far end of the
lake. Expansion of saline areas has
also been developed in different
parts of the lake to produce salt.
4- Most of urban expansion goes
toward the western parts of the
city; meanwhile, part of the
agricultural land in the east and
the south east of the city has been Figure 5. Land use change in Alexandria (1984-1993)
lost for increased urbanization.
5- A wide area of the desert and the Class 1984 1993 Size of % of change
coastal plain has been developed (Hectare) (Hectare) change(-/+)
as a new location for urban Urban fabric 61964.3 79768.8 + 17804.5 28.7
housing. Green lands 128770 127105 - 1665 1.3
Fish 12685.3 16306.8 + 3621.5 28.5
Conclusions and Future Work farm/Saline
Desert 313878 297738.5 - 16139.5 5.1
The results of this study show the trend Water bodies 175441 171451.5 - 3989.5 2.3
and size of urban growth in
Alexandria; most of the new urban Table 3. The size of land use classes and the change in Alexandria area (1984-1993)
development in the city goes towards

Monitoring Urban Growth In Alexandria-Egypt Using Satellite Image▐


Paper
▌FORUM

From To 350000
Water bodies Urban fabric
Desert / Sand Urban fabric 300000

Water bodies Fish farm / saline 250000

Water bodies Green lands

Hectares
200000
1984
Green lands Urban fabric 1993
Table 4. Summary of land use change in Alexandria area (1984-1993) 150000

100000
the western parts where the big areas of temporal urban mapping model, the
needed land available. However, most Urban Growth Model, was developed 50000

of this urban expansion is not planned by Clark (Clarke, 1998). This model 0
by the government so that, the rest of develops a historical perspective of

D e

es
t
ic

s
these areas have to be planned properly urban development by combining a

er
nd

in
br

di
es
al
fa

sh n la

bo
/S
to be protected from the disadvantages variety of data sources into an

an

rm

er
e
re
rb

at
fa
G
U

W
of urban growth in developing integrated, multi-scale, and multi-

Fi
countries. In addition, there is resolution database. The database
significant expansion in the eastern and provides the baseline information to Figure 6. The increase/decrease for the
different land use classes in Alexandria
southeastern areas of the city where model and predict regional patterns of (1984-1993)
cultivated land is, so that a new plan is urbanization and also highlights the
needed to direct the growth to the profound changes to the landscape that period of time (Timmermans, 1997).
western parts to protect these areas. have incrementally developed over This model has been successfully
time (USGA, 1998). The model has applied in several real world situations
In relation to urban remote sensing successfully predicted urban expansion in the USA. These and other models
applications, the results obtained from at regional scales. In its prototype will be tested and evaluated within the
this application of Landsat (TM) phase, the USGS project estimated the ongoing study. Information will be
images data indicate that the use of this regional impact of urbanization on the analyzed, to detect the factors and
type of high resolution satellite data are San Francisco Bay area and limitations of urban growth in the city,
very useful for mapping land use and Washington-Baltimore region (Clarke, and finally, to determine the results and
change detection analysis especially in 1998). effects of that growth.
urban areas which face a continuos
growth, the traditional methods can not A GIS decision support system will be Acknowledgements
cope with this dynamic phenomenon. designed using available models to
manage the predictable urban growth in The author would thank very much
This study so far reveals the need for the city of Alexandria. There are many Prof. Farouk Elbaz and Dr Mary Pax of
an Urban Land Use/Cover urban models that have been used in Center for Remote Sensing, Boston
Classification scheme, which can be urban applications (Engelen et al, University for their great help in
applied in Egypt and the Middle East 1997) have developed a fairly generic providing the satellite images for the
region. The proposed scheme could simulation shell that is usable in study.
help the studies, which use a remotely designing and running constrained
sensing data as a main input. cellular automata simulations
Temporal urban mapping relies on (Timmermans, 1997). It features a References
modern mapping techniques, such as built-in GIS system and a set of tools
remote sensing and geographic allowing the user to develop
information systems (GIS), to capture interactively his or her application. The Clarke, K. C. (1998) Clarke Urban
information from both historical and ALLOT (‘A Land Location and Growth Model, University of
modern records. At the GIS stage, the Optimization Technique’) model is a California, Santa Barbara, Department
outputs of a remote sensing phase and flexible computer model which has of Geography,
the different kinds of ancillary data will been designed to develop economically http://www.ncgia.ucsb.edu/projects/gig
be used within a model to predict the and environmentally sound alternative /model.htm (Feb., 2001)
prospective urban growth in the city. A land development patterns in a short

Monitoring Urban Growth In Alexandria-Egypt Using Satellite Image▐


Paper
▌FORUM
Ward, D. and Murray A. T., (2000)
“Monitoring Growth in Rapidly
Urbanizing Areas Using Remotely
Sensed Data.” Professional Geographer
52(3): 371-386.

EGSA (1997) Egyptian Series -


Alexandria. Cairo, EGSA

Erdas, I., (1997) ERDAS Field Guide.


Manchester, Manchester Computing.

Jensen, J. R. (1996) Introductory


digital image processing: a remote
sensing perspective. USA, Prentice-
Hall, Inc

Shouk, Y. H. a. F. A. (2000) Human


impacts on Alexandria’s marine
environment. Paris, © UNESCO
http://www.unesco.org/csi/pub/source/
alex8.htm, (Feb., 2000)

Thelin R. J. G. a. G. P. (2000)
Classification and Mapping of
Agricultural Land for National Water-
Quality Assessment, USGS,
http://water.wr.usgs.gov/pnsp/circ1131/
table1.html (Feb., 2001)

Timmermans H., Ed. (1997) Decision


support systems in urban planning.
London, E & FN Spon.

United Nations, (1997). Urban


Agglomerations 1996,
http://www.undp.org/popin/wdtrends/u
rb/urbpab.htm (Feb., 2001)

USGS (1998) Temporal Urban


Mapping,
http://edcwww2.cr.usgs.gov/umap/uma
p.html (March 2001)

Monitoring Urban Growth In Alexandria-Egypt Using Satellite Image▐

You might also like