|
|
|
Change detection and assesment using multi temporal satellite image for North-East Mediterranean Coast
2. Study area
The Çukurova Delta is located in the south of Adana on the south-eastern Turkish Mediterranean coast. It represents an internationally important coastal strip ecosystem, 110 km long.
The Çukurova Delta is the largest coastal river basin in Turkey and created by the sediments from the Berdan, Seyhan and Ceyhan Rivers and Taurus Mountains over two thousand years. The southern part of Tarsus City (now inland) was a Roman harbour demonstrating that extensive eroded material has been carried from Taurus Mountains [1, 2] and deposited in the delta. With a 110 km long coastal strip, Çukurova Delta encloses an area of about 5000 km².
Whilst the upper Delta basin has important fertile soils for agriculture, the lower part is still an internationally important coastal ecosystem containing biologically diverse habitats and biotopes.
The wetlands and coastal ecological areas in the Delta are, from west to east: Tuzla, Akyatan, Agyatan and Yumurtalik Lagoons. The Delta contains valuable habitats for nearly 200.000 water birds over-winter every year. Two hundred and sixty eight bird species have so far been recorded [3].
As well as having importance for wildlife, the coastal lagoons are also of economic importance for "Dalyan Fishing" [4]. Dalyan fishing is a system based on the catchments of the mature fish flocks by using the tights during the period when the fish flocks move between sea and lagoon. However, it is done only during fishing season and is a utility coming from the nature.
The Delta is also a breeding and nesting area for three globally threatened sea turtles; Caretta caretta, Chelonia mydas, and Trionix triunguis. [5, 6, 7, 8]. This delta is important not only for sea organisms, but also for some endemic vegetation. A type of Aleppo pine (Pinus halepensis), not common in Turkey, can be seen in Çamlik Lagoon, and it is surrounded by sand dunes and lagoons with wetlands. This site has a high floristic diversity [9]. There are also some endemic halophytic plants in the Delta [10, 11, and 12].
The area of dunes is decreasing mainly because of agricultural activities, but Turkey's largest natural coastal sand dunes still survive in the Delta [11, 12, 13].

Figure 1. Study area
3. Methods of Study
Most change detection techniques fall into five general categories: manual, write function memory insertion, image enhancement, multi-date data classification and comparison of two independent landcover classifications [14, 15].
In this study, post-classification change detection method was utilised.
Post-classification change detection is used to compare two independently prepared classified images. Two supervised classifications are produced using the same information classes to facilitate a comparison of two images. This procedure not only allows areas of no change to be identified, but in areas where change has occurred, the nature of the change can be determined [16]. However, in traditional supervised classification change detection, changes must be known in order to be sampled [17].
Landsat imageries of dates 1992 and 2000 were considered for digital image processing. The features of Landsat satellite images were given in Table 1.
Table 1. The features of the Landsat satellite images used in the study
| Landsat Thematic Mapper 4 – 5 |
| Spectral Resolution (m) |
Spatial Resolution (m) |
Temporal Resolution (day) |
Radiometric Resolution |
| Band 1 (0.45 – 0.52) |
30 x 30 |
16 days |
8 bits |
| Band 2 (0.52 – 0.60) |
30 x 30 |
16 days |
8 bits |
| Band 3 (0.63 – 0.69) |
30 x 30 |
16 days |
8 bits |
| Band 4 (0.76 – 0.90) |
30 x 30 |
16 days |
8 bits |
| Band 5 (1.55 – 1.75) |
30 x 30 |
16 days |
8 bits |
| Band 6 (10.4 – 12.5) |
120 x 120 |
16 days |
8 bits |
| Band 7 (2.08 – 2.35) |
30 x 30 |
16 days |
8 bits |
The satellite images used in the study were purchased from Landsat as their geometric and radiometric corrections being already processed.
Unsupervised classification process ISODATA (Iterative Self Organizing Data Analysis Technique) was applied to image data sets. A preliminary thematic raster layer was created, which gave similar results to using a minimum distance classifier. From these data, signatures were created. This thematic layer can be used for analyzing and manipulating the signatures before actual classification takes place [18]. ISODATA algorithm produced 12 spectral clusters after the generalization on the fieldwork, and six landuse/land cover classes were determined: agriculture, wetland, sand dune, vegetation, shallow water, deep water.
|
|
|