Yasser AYAD and Michel GUENET Faculty of environmental planning University of Montreal

The use of Remote Sensing and GIS in the assessment of visual attributes: Case study of the northwestern coastal zone of Egypt

ABSTRACT

Geographic Information Systems (GIS) techniques are widely used to analyse different characteristics of the landscape. The feasibility of using remotely sensed data to delineate landscape units and assess their relative visual quality was interpreted by authors in the beginning of the nineties, which has proven to provide considerable savings in time.

In the present paper, the main objective is to assess the change in the visual attributes of the landscape using techniques of Remote Sensing (RS) and GIS. Thus a study area located in the northwestern coastal zone of Egypt was selected. The landscape of the selected area has witnessed dramatic changes during the past few decades because of the national interest aiming at increasing the economic benefits of coastal zones. This may have direct impact on the visual quality of the landscape. Therefore, an approach of preparing remotely sensed data is carried-out to delineate different physiographic units. Different visual attributes were selected for the present study such as land use/cover diversity, activity (degree of naturalness) and landform. These attributes can be considered of great value in measuring the visual quality of the landscape.

Available data consisted of sets of aerial photographs for 1949 and 1977, and SPOT satellite images (XS and Panchromatic) for 1987 and 1992. The study was carried-out in five major stages: (1) data selection and pre-processing, (2) physiographic units identification, (3) attributes definition, (4) building the databases, and (5) analysis and results representation. Visual attributes were extracted for each date and analyzed in order to assess the nature of visual attributes changes. In conclusion, the study area under investigation is witnessing changes in its visual qualities which has to be taken into consideration in planning and managing its resources. This calls for an urgent act in order to include visual quality assessments in national plans for landscape management especially for regions of high touristic importance. Furthermore, the integration of RS and GIS to study visual attributes evolution has proven to be time-saving tools that can detect and manage different landscape attributes, thus giving the opportunity for decision makers to act promptly to specific planning problems for the benefit of viable and sustainable landscapes.

Introduction

Physical characters of the landscape can be identified by its visual attributes, this relates to the fact that planning and management decisions, culture interaction, and natural processes lead to physical changes that will eventually be seen in the landscape. Scenic landscapes are a major source for human enjoyment and in some cases have been the object of direct public action to preserve their quality (Fabos et al., 1978). Furthermore, the desire for valid means for quantifying the scenic characters of the landscapes has substantially increased with the development of land use planning and its requirement for environmental data on which to base land use decisions (Smardon, 1983, Zube et al., 1976, Litton et al., 1974, and Jackle, 1987).

Many researches have attempted to categorize the literature according to major paradigms of landscape appraisal. Hence, the current study is adopting the expert, behavioural paradigm, which draws on the discipline of psychology, and includes the development of predictive landscape dimensions, both physical and cognitive based on the statistical response of observers to the landscape (Smardon, 1983, Itami, 1989, and Crawford, 1994). It has been proven that some landscape attributes are more preferred than others. The perception of topographic ruggedness, the presence of water bodies, natural vegetation variety, higher densities of tree cover, and more natural landscapes increase the feeling of higher scenic quality.

In this respect, visual attributes and scenic indices are traced in order to identify and to protect landscapes of higher visual amenity. The measurement of such attributes relied on manual methods and extensive field works, which have proven to be a time and money consuming task. However, technologies of Remote Sensing and Geographic Information Systems (GIS) have been fairly introduced in such fields of study. Remote sensing can provide a bird's eye view of the land in a repetitive manner, and thus enable the study of the change and present status of the landscape (Aronoff, 1987). Although it has been used in other disciplines in landscape planning, their capabilities remained undiscovered for visual quality assessments. Several authors have mentioned its importance, but few have tried to assess its relevance to the field. However, Crawford (1994) has recently tested the feasibility of using remotely sensed data in the assessment of landscape visual quality. The author used landscape MSS images in his study and concluded that the use of higher resolution remotely sensed data and digital terrain information can considerably save time and effort.

On the other hand, arid coasts landscapes, which have their own visual characteristics, are perceived from the inhabitants and many of their seasonal visitors as an escape from the tedious life of the cities. Many studies have been carried-out to assess the natural resources of these landscapes, studies vary from the economic to the ecological evaluation of the environment despite that these arid coasts may serve as a major source for national income from tourism and holiday making.

In this paper the main aim is to assess the changes that have occurred to the visual characteristics of the arid coasts landscape. It demonstrates the feasibility of incorporating remotely sensed data in the GIS to achieve this aim, and therefore to establish a model to detect visual changes of arid landscapes.

The study area

A study area located 60 kms west of Alexandria city in Egypt is selected (Figure 1). It expands to about 20 kms westward to include both villages of Burg El-Arab and El-Hammam. It extends 15 kms from the coast line to the south (Figure 2). It represents a transitional area that have witnessed changes in its landscape during the second half of the 20th century. The northwestern coastal zone of Egypt in general, is distinguished by a northern coastal plain and a southern tableland. the coastal plain is characterized by the presence of a number of alternating ridges and depressions running parallel to the coast in the east-west direction.

Three sub-regions were identified for the present study: the coastal region, the saline depression, and the Mariut ridge (Figure 3).

First, the coastal region includes the beach, the coastal sand dunes, the first depression, and Abu Sir ridge. The beach is approximately 200 m wide, it consists of white sand and is bordered from the south by dune rows. The coastal sand dunes are approximately 400 m wide. The mean elevation of this bar is 10 m. The Abu Sir ridge or the first rocky ridge is approximately 900 m wide. It extends parallel to the coastal sand dunes with an average elevation of 30 m. On the edge of the northern slope of the ridge lies Alexandria-Matruh coastal highway. This region is considered as the main source of national income of summer tourism and holiday making activities, several touristic villages are constructed on the coastal sand dunes. Furthermore, many quarries are being exploited from the first rocky ridge.

Second, the saline depression (the second depression) is approximately 1.8 km wide and consists of a periodical water-logged depression. The surface of the depression is mostly below sea level, and is filled mainly with brackish water and saline calcareous deposits of weathered and down wash materials. This region is a transitional zone of travellers between the coast and the internal highway, through which both slopes of the first and the second rocky ridges can be seen. It contains large water bodies and natural vegetation. No serious human activities are witnessed in this region to the shallowness of water table and the saline nature of the soil.

Finally, the Mariut or the second rocky ridge is approximately 1.3 km wide, with an average elevation of 25 m. It is limited by the saline depression in the north and with the non-saline depression in the south, and it can be seen directly either from the first rocky ridge or by travellers through the saline depression to further inland.

Data preparation

All spatial data were obtained from the Department of Environmental Studies at the Faculty of Science, University of Alexandria, Egypt. It consisted of a topographic map of 1977 (scale 1:50000), aerial photographs of 1955 and 1977 (scale 1:25000), and SPOT XS and Panchromatic satellite images of 1987 and 1992.

The contour lines of the study area were extracted from the topographic map of 1977. They were digitized, projected and then converted to a Digital Elevation Model (DEM).

After the preprocessing steps, registration, and geometric correction, the SPOT XS images of 1987 and 1992 (20m x 20m spatial resolution) were classified using the Maximum Likelihood Classification (MLC) algorithm. This resulted in 12 land use/cover classes for each date.

On the other hand, the land use/cover classes for 1955 and 1977 were extracted from the aerial photographs mosaics. The result was digitized, corrected and adjusted to match the same projection system of the classified satellite images for further comparison. All the produced vector thematic maps were then converted to the raster format. The resulting classes for all dates were synchronized which produced the final nine classes presented in Table (1).

**Table 1: The final land use classes**
class name	ID	Description
dunes	1	coastal sand dunes
crops	2	crop-cultivated land
orchards	3	orchards plantation
ploughed	4	land prepared for agriculture
background	5	open land, including pasture land and sparse natural vegetation
urban	6	the urbanized zones including built-up and the surrounding areas
flooded	7	undulated salt-marsh areas
s.m. vegetation	8	all salt marsh vegetation classes
quarries	9	quarries and construction sites

Methodology

Two visual characteristics were used in the present study, namely the "activity" and the visual "diversity" of the land uses. The "activity" describes the degree of naturalness or compatibility of the existing activities with the natural character of the landscape, and is based on physical landscape characteristics. The visual "diversity" on the other hand, is a criteria to measure the variety and the proportional configuration of the land cover in a unit area. It is related to aesthetic principles, in recognition of the fact that people not only respond to the landscape by the quality of the visual characteristics but also by their juxtaposition and combination (Crawford, 1994). Figure (4) traces the major steps for calculating the visual characteristics.

Three major classes of activities were recorded (Table 2), by which the degree of naturalization is identified according to the proportional distribution of its corresponding land use classes in a block of 100 x 100 meters (5 x 5 pixels). The proportional distribution is calculated with the equation (1):

PP = 1/x_i * SUM (N_i) equation (1)

Where PP is the proportional distribution, x is the total number of pixels in each block (5 x 5 pixels for the present case), N_i is the activity of ID i.

**Table 2: The activity classes and their definition**
ID	Code	Activity	Class_names	Class_ID
3	NA	Natural	Dunes Background classes Flooded areas Salt-marsh vegetation	1 5 7 8
2	SN	Semi-natural	Crops Orchards Ploughed fields	2 3 4
1	AR	Artificial	Urban areas Quarries and construction sites	6 9

The final output was produced to translate the existing land uses in the study area with its corresponding "degree of naturalness". Each resulting class was thus given a specific score which describe its importance (Table 3) (Crawford, 1994).

**Table 3: The combination of different *activity* classes**
ID	Code	Score	Meaning
3	NA	3	Only natural classes
2	SN	2	Only semi-natural classes
23	SNNA	1	Combination of natural and semi-natural classes
123	EQ	0	Equal share between all classes (natural, semi- natural, and artificial)
13	ARNA	-1	Combination of artificial and natural classes
12	ARSN	-2	Combination of artificial and semi-natural classes
1	AR	-3	Only artificial classes

Thereafter, the Shannon Diversity Index is also calculated for the same 100 x 100 meters block according to the equation (2):

SHDI = - SUM [ P_iln (P_i) ] equation (2)

Where SHDI is the Shannon Diversity Index, i is the class, P_iis the proportional abundance of a class i in a specific study area. The value of P_ialways vary between 0 and 1 therefore the logarithm of P_i will always be a negative value (unless for 0 which will give infinity, and for 1 which will give 0).

The presence of the negative sign outside the summation function is needed in order to retain a positive result for the index. A SHDI equal to 0 means that there is no diversity within the 100 x 100 meters block (only one class exists), and the greater the value of the index the more diverse the site. The SHDI calculation procedures are depicted in figure (4).

For each sub-region, the proportional distribution is calculated for the 100 x 100 meters block for each class separately, and the first part of the equation (3) (P_i ln (P_i) ) is therefore calculated. The calculated results are then summed to produce one grid that contains the values of the SHDI. The cell size of the resulting grid is of 100 x 100 meters. These values are therefore classified in order to represent four scores of diversity (Table 4).

Table 4: The diversity classification and their corresponding proposed scores

Score	Code	Meaning
0	ND	No Diversity
1	LO	Low diversity
2	IN	Intermediate diversity
3	HI	High diversity

The results of both the "degree of naturalness" (Table 3) and the diversity classes (Table 4) were merged in order to sum the scores of each block. The produced grids represented the distribution of the visual quality for each location. Finally a scalar of 4 was added to the results in order to redistribute scores to values that vary from 1 to 10. Score 1 represents an assumed lowest visual quality, which is represented by no diversity (score 0) and artificial activity (score -3) (i.e. 0 + (-3) + 4 = 1). On the other hand, score 10 represents an assumed highest visual quality in which the diversity is the highest (score 3) and the activity class is natural (score 3) (i.e. 3 + 3 + 4 = 10).

Results

Region (1): The coastal region

As it can be noted in figure (5), the region under consideration is being transformed to a higher diversified landscape, it dramatically changed from natural to artificial utilization in the later years. The natural areas have witnessed a loss of about 6.6% whereas the artificial/natural combination areas increased dramatically from zero to 37.5% over the 37 years period under investigation. Four combinations are strikingly changing, the intermediate-diversity areas that include artificial and natural classes jumped from zero to 22.2% of the total area, whereas the intermediate-diversity areas combining semi-natural and natural classes occupied 25.2% of the total area in 1955 have totally disappeared in 1992. On the other hand, the same trend can be found in the higher-diversity locations with respect to their smaller occupational area.

Figure 5: The change in the diversity/activity classes for the first region

To translate this into visual quality meanings, the suggested schema demonstrate a notable transition from higher to lower scores (Table 5). The 37 years period from 1955 to 1992, as depicted in the transition matrix, has witnessed notable changes. In 1955, 66.67% of the total area had a score of 7, 19.09% had a score of 8, and 8.24% had a score of 9, other scores had less than 5%. In 1992, these percentages were redistributed all over the scores from 1 to 10. However, it is important to note that the percentage of scores of 8 and higher were almost unchanged, but on the other hand, areas of scores of 1 to 3 and 5 increased and formed 27.76% of the total area in 1992, which was only 0.59% in 1955.

Table 5: Transition matrix of the proposed visual quality classes from 1955 to 1992 (region 1)

Region (2): The saline depression

It has to be noted that in this region, the artificilization process did not affect, to some extents, its visual attributes. Notable changes are present in the higher-diversity natural areas which have increased from 3.32% in 1955 to 11.63% in 1992 (Figure 6), and in the intermediate-diversity semi-natural/natural combination areas, which have decreased from 16.95% in 1955 to 0.1955% in 1992. On the other hand, natural areas with no diversity increased in 1977 to 28.8% and started to decrease to achieve 12.19% in 1992. This indicate a transformation in the homogeneous natural areas to more fluctuating, heterogeneous areas of combination between either natural or semi-natural classes. Consequently, the loss of 10.72% of the visual quality rate 7 demonstrate a portion of 5.448% increase in rate 6 and from rates 8 to 10 (Table 6).

Figure 6: The change in the diversity/activity classes for the second region.

From these results it can be concluded that the saline depression, is maintaining most of its visual qualities, despite the introduction of some construction projects in the late eighties. The rate of decreased qualities is not alarming for this region. This may be attributed to the difficulties and the cost inadequacy of the exploration of the salt marsh areas.

Table 6: Transition matrix of the proposed visual quality classes between 1955 and 1992 (region 2)

Region (3): The Mariut ridge

Due to the change between 1955 and 1977 in the total area occupied by agricultural activities, the natural areas with no diversity were absent in 1955 (Figure 7), they re-appeared in 1977 (19.4%) and decreased to 4.09% in 1992. Changes in two classes are also notable, first the increase in the intermediate diversity classes of artificial/natural combinations from zero in 1955 to 13.89% in 1992, and second the decrease in the intermediate diversity classes with semi-natural/natural combination (from 44.54% in 1955 to 1.02% in 1992). The increasing trends of the artificial classes in different diversity values are also identifiable.

Figure 7: The change in the diversity/activity classes for the third region

The visual quality rates, on the other hand, indicate a serious loss of 22.47% of rates 6 and higher and a gain of 4.502% of rates 9 and 10 (Table 7). A transition matrix having the greater values in its diagonal (no change) or having values deviating to its upper portion of the diagonal, represents higher visual quality rates redistribution in the upcoming dates. In this region, however, values representing the 1955 rates were redistributed in almost all the lower portion of the diagonal. It can be also noted that this loss is concentrated all over the built-up areas. Planning criterion for better visual appeal may propose a less densified and less concentrated built-up and industrial zones, and more integration with natural and semi-natural classes, which may enhance visual qualities by increasing the diversity and adequately distributing its classes.

Table 7: Transition matrix of the proposed visual quality classes between 1955 and 1992 (region 3)

Discussion

From the obtained results, it can be concluded that there is a general loss in the visual resources of the study area in general. In the first region, for example, specific attention should be given to the visual qualities, because it is accessible by several tourists from different parts of Egypt. These visitors usually have different cultures, and they either pass by this area heading to western zones (Matruh village, Sidi Barrani, El-Alamein, ...etc.), or settle for periods varying from few days to the whole summer period (about 3 months). These visitors are mainly attracted by the beach and consider it as an escape from the tedious rhythm of the life in the city. In brief, the visual qualities of this region adds to the natural resources of the area and therefore can be given higher priorities in the planning process. The artificial areas increased dramatically in both the first and the third regions with a relative decrease in the natural and semi-natural classes. The resulted scores for these regions depicts a loss in its visual resources. On the other hand, the second region, the saline depression, is still retaining its visual qualities, specific attention should, therefore, be taken into consideration to prevent any loss in the future, because even though the salt-extraction industries are not heavily exploited in this region, but its significant impacts may be found in areas more in the east. The impacts of such utilization and the dynamics of the salt marsh areas can be suggested for detailed future studies.
On the other hand, the presented attributes described the visual resources of the study area, more detailed attributes and landscape characters can also be utilized and can also be merged to form a more comprehensive method for detecting visual changes, and therefore to provide a basis for landscape and regional planning. More detailed studies can be carried-out to explore the visual characteristics of arid coasts landscape, other attributes such as color, water bodies edge variety, natural vegetation, ...etc. can be included. Other studies are also encouraged to explore the visual potentialities of the arid coasts landscapes.
The inclusion of the visual analysis and assessments in landscape and regional planning has to be considered in the national policies, furthermore, the visual characteristics of the landscape has to be frequently monitored. Remote sensing data can help in detecting their changes in a regional scale which can provide a basis for local planning strategies for selected sites of higher management priorities.
The present study encourage the use of remotely sensed data and image processing techniques to develop models in order to detect and analyze the visual attributes of the landscape, especially those of physical nature. Hence, It detected a limitation due to scale problems, the detection of the natural vegetation of the arid coasts is still questionable with satellite remote sensing. However, their input to GIS can provide valuable information to visual resources management.

Acknowledgement

The authors are thankful to Dr. Mohammed Ayyad and Dr. Boshra Salem, Department of Environmental studies, Faculty of Science, Alexandria University, for their help in different data preparation stages, and for providing the appropriate satellite images and aerial photographs used in this study.

References

Aronoff, S. (1987). Geographic Information Systems: A management perspective. WDL publications, Ottawa.

Crawford, D. (1994). Using remotely sensed data in landscape visual quality assessment. Landscape and Urban Planning. 30:71-81. ELSEVIER.

Fabos, I.Y., Greene, C.M., and Joyener Jr., S.A., (1978) The METLAND landscape planning process: Composite landscape assessment, alternative plan formulation and plan evaluation: part 3 of Metropolitan landscape planning model: Massachusetts Agricultural Experiment Station and I.S. Department of Interior Office of Water Research and Technology.

Itami, R.M., (1989). Scenic perception: Research and application in U.S. visual management systems. In: Landscape evaluation: Approaches and applications. Dearden, P., and Sadler, B. (Eds). Western geographical series volume 25. University of Victoria.

Jakle, J.A., (1987). The visual elements of landscape. The University of Massachusetts Press.

Litton, R., Tetlow, R.J., Sorensen, J., and Beatty, R.A., (1974). Water and landscape: An aesthetic overview of the role of water in the landscape. Water Information Center Inc.

Smardon, R.C. (1983). The future of Wetlands: Assessing visual-cultural values. Allenheld, Osnmun Publishers, Inc.

Zube, E.H., Brush, R.O., and Fabos, J.Y. (Eds) (1975). Landscape assessment: values, perception, and resources. Dowden, Hutchinson and Ross, Inc.

AYAD Y.M. (M.Sc. Environmental studies)
Ph.D. Candidate, Faculty of Environmental Planning
University of Montreal, Montreal, Canada
ayady@ere.umontreal.ca
http://mistral.ere.umontreal.ca/~ayady

Dr. GUENET M. (Ph.D. Geography)
Adjunct Professor, Institute of Urbanism,
Faculty of Environmental Planning,
University of Montreal, Montreal, Canada
guenetm@ere.umontreal.ca

Faculte de l'amenagement,
Universite de Montreal
C.P. 6128
Succ. Centre-Ville, Montreal, QC, H3C-3J7
Canada

Tel.: (514)343-2373
Fax: (514)343-2338