The coastal and marine resources of Abu Dhabi have sustained the local inhabitants for thousands of years. However, rapid development following the discovery and exploitation of mineral oil reserves in the second half of the 20th Century changed the intensity and use of the natural resources. Changes in coastal geomorphology, landuse and vegetation were documented using remote sensed images covering the period from 1972 to 2003. The study indicated that there have been major geomorphologic and landuse changes within the coastal zone of Abu Dhabi as a result of urban and industrial development. Vegetation, both cultivated and intertidal, has increased within the study period. The data collected emphasises the urgency for developing sustainable coastal management strategies in Abu Dhabi and the region.

Introduction

The Abu Dhabi capital area underwent rapid urban and industrial development during the period from 1972 to 2003, and the authors hypothesized that the physical structure and the vegetation of the area was altered as a result of this development. This study represents an effort to quantify these changes in order to better understand the consequences of urban and industrial development in the Abu Dhabi capital area, especially since the development that occurred from 1972 to 2003 continues today. The objectives of the study were:

  1. To use remote sensed satellite data to document overall geomorphological, land use and vegetation changes that have occurred in the Abu Dhabi capital area from 1972 to 2003.

  2. To undertake quantitative analysis on the changes in vegetation resources in the Abu Dhabi capital area.

Remote sensed satellite data of the study area were obtained from various periods from 1972 to 2003 and were geographically corrected to within 20 m using permanent ground features (Figure 1). A specific area of interest centred on the capital area of Abu Dhabi was established (Figure 1), and detailed quantitative analysis of changes within that area's vegetation was undertaken for each period. Overall trends in coastal parameters such as geomorphology and shallow water areas were identified. Methods used to determine changes in natural resources included normalised difference vegetation index (NDVI) and fractal dimension analysis of coastlines. NDVI as a remote sensing tool is usually used for discerning areas of vegetation. However, in this study area, NDVI was useful for differentiating between landform classes because the only vegetation that occurs is intertidal vegetation and cultivated vegetation. To determine between the two, areas of cultivated vegetation, i.e. Abu Dhabi Island, were removed from the analysis. Manual examination of remote sensed data sheets using grid overlays, and ground truthing* enabled more recent detailed changes in the area to be determined. Digital change detection analysis was also performed on the datasets to compare with manual methods. The analysis indicated long-term changes in coastal and marine resources, growth of urban and industrial development both spatially and temporally and development trends.

Methodology

Assessing change in the study area (1972–2003)

Remote sensed data processing and unsupervised classification of the datasets for the various periods was undertaken at Southern Cross University, Australia. Supervised classification, analysis, interpretation, and ground truthing was undertaken in Abu Dhabi. To correct the geographical error, all datasets were geographically matched. This was undertaken by using one dataset (1996) as a reference, and matching all other datasets with that scene. To achieve correction across the entire study area (Abu Dhabi emirate), a series of known features/sites were located in all scenes. In total, six different sites geographically spread across the study area were selected in all the datasets, and these were then checked for geographical error between the various datasets/years (Table 1). Loughland (2006) provides details on the methodology employed in rectifying the various images and defining the area of interest (AOI) for this study. All remote sensed images were standardized at a scale such that each pixel covered 25 × 25 m2.

Normalised difference vegetation index (NDVI)

A Normalised Difference Vegetation Index (NDVI) analysis (Rouse et al., 1973) was undertaken for all AOI subsets (i.e. from each period). This analysis identified two subclasses of vegetation: intertidal vegetation, which is all vegetation occurring within the AOI but off Abu Dhabi (City) Island (i.e. mangrove and saltmarsh); and cultivated vegetation which occurs mostly on Abu Dhabi (City) Island and grows with the assistance of artificial irrigation. The desert areas contain minimal vegetation, and are recorded as bare sand (i.e. class land). The analysis also identified class water (i.e. marine water). These three main classes represented land, water and vegetation in the analysis. Thresholds between vegetation, land and water were obvious by their delineation on the maps and through ground truthing. This particular study site, because of its aridness, allowed NDVI analysis to be performed on different classes such as water, land and vegetation.

Classification of classes within the AOI-NDVI subsets

Abu Dhabi Island (City) covers around 93 km2 (6% of the AOI), and has cultivated green areas consisting of numerous parks, gardens, sports fields, and wooded areas. The large green areas of Abu Dhabi City (Island) resulted in difficulties with the classification of intertidal vegetation adjacent cultivated areas. Therefore, the entire area of Abu Dhabi City (Abu Dhabi Island) was masked from each AOI subset for the purpose of clearer classification of natural intertidal vegetation (Figure 2).

Two data subsets from each year's datasets were created.

  1. AOI-NDVI-With-Island: The total AOI, including Abu Dhabi Island (Abu Dhabi City).

  2. AOI-NDVI-Without-Island: The total AOI minus Abu Dhabi Island (Abu Dhabi City).

As each pixel covered 25 × 25 m2, it was possible to calculate the average km2, hectare values, and the percentage (%) that each class contributed to the AOI. The results for the different classes, for all the datasets (all years) for both With-Island and Without-Island, were analysed for changes and trends over 31 years.

Measuring change in the study area

In order to quantify change in the various classes at particular times (from 1972 to 2003), a change detection analysis was undertaken on all the AOI-NDVI subsets. The analysis involved changes between the three identified classes of vegetation, land and water for all the AOI-NDVI-With-Island and AOI-NDVI-Without-Island subsets.

To determine the actual proportion of changes for each class, in each period on Abu Dhabi Island, required the subtraction of the results for the analysis of Without-Island from the results from the analysis With-Island. This data provided an insight into the contribution of changes on the Island, to the overall changes in the AOI.

Change detection analysis had three main steps:

  1. Manual subtraction of the total pixel value for each class and between each period (e.g. 1984 minus 1972) was undertaken for With-Island and Without-Island. This analysis identified clear changes and relationships from one class to another but did not identify the location of the change.

  2. Supervised assessment of change by overlaying AOI and NDVI subset maps with a 5 × 5 km2 grid, and examining changes and trends in the class area of each grid cell, from one scene (as opposed to period) to the next (i.e. 1972 through to 2003, figure 3). This provided an estimate of the change in area of each class. This procedure was also used for undertaking supervised assessment of spatial and temporal landuse changes (i.e. urban and industrial areas). This provided information on the location, rate, and overall trend of new developments in the AOI.

  3. Unsupervised (GIS) assessment of change used image analysis software which subtracts one year's dataset from the next years dataset (as in step 1) and highlights the areas that have changed on a map with a grid overlay as in step 2. During this assessment, tidal error and adjustment error of pixels was indicated. In addition, mixed pixels were indicated, especially on man made surfaces, such as roads after rainfall events. This analysis allows both the total class changes (in pixels), and the location of the class changes to be identified for each period.

Fractal dimension analyses (changes in complexity of the coastline over time)

Fractal dimension is a relatively new field of geometry and Mandelbrot (1967) first described a coastline's infinite length with fractal dimension using a value (D) which indicates the complexity of a coastline. The more complex, the greater the value D, and complexity is influenced by anthropogenic factors such as coastal development.

Fractals have been used in combination with remotely sensed data to describe certain processes or aspects within a landscape over a period of time by comparing images from different dates (Jorge and Garcia, 1997; Nikora et al., 1999; Chust et al., 1999; Li, 2000; Griffith et al., 2000; Jenerette and Wu, 2001; Imbernon and Branthomme, 2001; and Grossi et al., 2001). Fractals also show effects such as human impacts, which usually tend to reduce the overall complexity of an environment (Krummel et al., 1987).

It was hypothesised that anthropogenic development along the coastline of Abu Dhabi Island would result in less complexity (i.e. straight break walls, Corniche promenades and harbours) and therefore a reduction in the D value with increased time (i.e. with increased development). The island of Abu Dhabi has both a north and south coastline that are quite dissimilar. In order to measure the complexity of the two coasts, the fractal dimension analysis was undertaken separately for each coastline.

Results: Change in the study area

Analysis of data obtained from satellite images indicated changes in intertidal and cultivated vegetation, intertidal areas, coastal morphology, and coastal development. The data also indicated development trends in the study area. Despite the many physical changes to the study area, the most striking change was that of vegetation, with increased cover in both intertidal and urban areas (Figure 4).

Fractal Dimension changes on Abu Dhabi Island between 1972 and 2003

The analysis indicated that in 1972 and 2003 that both the north and south coasts of Abu Dhabi Island had different D values, with the south coast having the highest complexity in both datasets (Figure 5).

The analysis indicated that the complexity value (D) increased between the datasets, with increasing complexity being indicated in 2003 (Figure 5).

The fractal dimension analysis indicates that there have been major changes to the coastline of Abu Dhabi Island as a result of development activities. Contrary to our hypothesis, the analysis indicated that anthropogenic activities along the coastline such as reclamation, the dredging of channels and harbours and the construction of marinas and break walls actually resulted in greater complexity of the north and south coastlines.

Periods of most change in the study area

There have been significant changes as a result of development activities in the study area over 31 years. By 2003, Abu Dhabi City was 7.5 times larger than it was in 1972, and Abu Dhabi Island was nearly completely developed, and development was spreading to the northeast on the adjacent mainland.

For the unsupervised change detection analysis, the period identified with the greatest change was 1985-1988. This period recorded the greatest change for four of the seven change scenarios, and these were mostly related to changes in vegetation cover. The period identified by the unsupervised analysis with the least change was 1984-1985, where 97.62% of the AOI recorded no change. In this period there was also relatively little intertidal development activity.

In contrast, the greatest increase in the extent of changes in built development occurred from 1984–1985 (Figure 6) where the average percentage of grid cells recording new development was 1.25% per year, compared to an average of 0.46% per year for the entire study period. This increase was related to the expansion of both urban areas on Abu Dhabi Island, and industrial areas south east of Abu Dhabi Island.

The greatest increase in the area of built development was estimated from 1996–1998 (Figure 7), where the area of built development increased by around 62 km2, (an average of 31 km2 per year).

In contrast, the average rate of increase for the entire study period was around 6 km2 per year. The increase in the 1996–1998 period was probably related to additional built development off Abu Dhabi Island. The overall results of the analysis for built development indicated that both the area and extent of development increased constantly throughout most of the study period, and began to stabilise towards the end of the study.

Figure 8 indicates intertidal changes that took place throughout the study period with the greatest increase in intertidal development actually occurring at the end of the study from 2002-2003 and this consisted mostly of large-scale dredging and reclamation activities (Figure 9).

Similar to the results for built development (Figure 6 and Figure 7), the analysis for changes in intertidal development also indicated that the general area and extent of new intertidal development was constant throughout most of the study period, with the exception at the end of the study period, where intertidal development increased significantly (Figure 8 and Figure 10).

Conclusions

This study provides information on changes that have occurred in the study area since 1972. Remote sensed images covering the study area indicated that there have been dramatic changes in vegetation, with both terrestrial (cultivated) vegetation and intertidal vegetation such as mangrove increasing. The change detection analysis also indicated that the study area has undergone physical changes as a result of dredging and reclamation of extensive shallow water areas, creating major alterations to the morphology of the coastline.

Note

*

Ground truthing involved visiting anomalies (i.e. mixed or confused pixels) and then with a radio differential GPS recording the land cover at those anomaly sites. The mixed pixels were then classified based on the majority of land cover at the site.

This radiometric rectification was undertaken using histogram matching software (ERDAS Imagine Histogram match module software). Datasets were rectified to UTM (Universal Trans Mercator) coordinates.

Abu Dhabi city contains mixed parklands and urban areas adjacent intertidal vegetation and often it was difficult to separate intertidal vegetation from adjacent cultivated vegetation. By masking out the City Island, the remainder of the study area was ONLY natural intertidal vegetation.

References

Chust, G., Ducrot, D., Riera, J. L. L. and Pretus, J. L. L.
1999
.
Characterizing human-modeled landscapes at a stationary state: a case study of Minorca, Spain
.
Environmental Conservation
,
26
(
4
):
322
331
.
Griffith, J. A., Martinko, E. A. and Price, K. P.
2000
.
Landscape structure analysis of Kansas at three scales
.
Landscape and Urban Planning
,
52
:
45
51
.
Grossi, L., Zurlini, G. and Rossi, O.
2001
.
Statistical detection for multiscale landscape patterns
.
Environmental and Ecological Statistics
,
8
(
3
):
253
267
.
Imbernon, J. and Branthomme, A.
2001
.
Characterization of landscape patterns of deforestation in tropical rainforests
.
International Journal of Remote Sensing
,
22
(
9
):
1753
1765
.
Jenerette, D. G. and Wu, J.
2001
.
Analysis and simulation of land-use change in the central Arizona-Phoenix region, USA
.
Landscape Ecology
,
16
:
611
626
.
Jorge, L. A. B. and Garcia, G. J.
1997
.
A study of habitat fragmentation in Southeastern Brazil using remote sensing and geographic information systems (GIS)
.
Forest Ecology and Management
,
98
:
35
47
.
Krummel, J. R., Gardener, R. H., Sugihara, G., O'Neil, R. V. and Coleman, P. R.
1987
.
Landscape patterns in a disturbed environment
.
Oikos
,
48
:
321
324
.
Li, B. L.
2000
.
Fractal geometry applications in description and analysis of patch patterns and patch dynamics
.
Ecological modeling
,
132
:
33
52
.
Loughland, R. A.
2006
.
Towards sustainable management of Abu Dhabi's marine and coastal resources
,
PhD Thesis
Lismore NSW, , Australia
:
Southern Cross University
.
Mandelbrot, B. B.
1967
.
How long is the coast of Britain? Statistical self-similarity and fractional dimensions
.
Science
,
156
:
636
638
.
Nikora, I. V., Pearson, C. P. and Shankar, U.
1999
.
Scaling properties in landscape patterns: New Zealand experience
.
Landscape Ecology
,
14
:
17
33
.
Rouse, J. W., Haas, R. H., Schell, J. A. and Deering, D. W.
1973
.
Monitoring vegetation systems in the Great Plains with ERTS
309
317
.
Third ERTS Symposium, NASA SP-351 I