Tải bản đầy đủ - 0trang
296



D. Maktav and F. Sunar



Fig. 15.1 Map of the Bỹyỹkỗekmece district (numbers and brackets indicate the administrative

transformation dates of villages into sub-districts)



LANDSAT Thematic Mapper (TM), SPOT P and IKONOS XS & P image data

were also obtained and are described in more detail in Table 15.2.

The multi-temporal LANDSAT TM data (1984 and 1998) were rectified based

on 1:25,000 scale topographic maps using an automated registration process



Admin. unit

Bầekmece

Kavakl

Yakuplu

Hodere

Kraỗ

Gỹrpnar

Esenyurt

MimarSinan

Kumburgaz

Gỹzelce

Tepecik

Tỹrkoba

ầakmakl

Karaaaỗ

Ahmediye

Total



1970

3,913

501

974



371

1,305

923

2,296

928

722

1,607

339

344



435

14,658



1975

5,204

628

1,045

802

435

1,578

1,631

2,232

1,270

999

3,134

505

801

325

473

21,062



1980

8,121

866

1,252

864

544

2,812

6,636

3,138

2,750

2,111

4,805

364

525

451

664

35,903



1985

11,310

1,021

1,664

924

826

3,584

21,290

4,083

2,569

1,366

7,382

436

709

399

802

58,365



Table 15.1 Census data distribution in Bỹyỹkỗekmece (19701997)

1990

22,394

2,170

2,841

1,538

2,239

10,191

70,280

7,690

7,118



12,240

712

1,633

681

1,183

120,516



1997

41,644

24,475

23,878

12,915

10,353

20,702

100,565

15,204

8,329



14,588

2,392

3,675

868

1,300

239,244



Changes (%)

19701985

189.04

103.79

70.84



122.64

174.64

220.66

77.83

176.83

89.2

359.36

28.61

106.11



84.37

25.1



19851997

268.21

2,297.16

1,334.98

1,297.73

1,153.39

477.62

372.36

272.37

224.21



97.62

448.62

418.34

117.54

62.09

309.9



15

Remote Sensing of Urban Land Use Change in Developing Countries

297



298



D. Maktav and F. Sunar

2500



Growth %



2000

1500

70-85



1000



85-97



0



Bầekmece C.

Kavakl

Yakuplu

Hodere

Kraỗ

Gỹrpnar

Esenyurt

Mimarsinan

Kumburgaz

Gỹzelce

Tepecik

Tỹrkoba

ầakmakl

Karaaaỗ

Ahmediye



500



District



Fig. 15.2 Changes in population growth over the periods 1970–1985 and 1985–1997



Table 15.2 Characteristics of the satellite data used in the case study

Satellite

Sensor

Spatial resolution (m)

Date

LANDSAT

TM

30 (except TIR band)

June 12,1984

Apr 16, 1998

SPOT

HRV

10

Apr 16,1989

July 22, 1998

XS

4

Feb 13, 2002

IKONOS

P

1

Feb 13, 2002



(see detail of the registration process in Maktav et al. 2000; Sunar et al. 2000;

Taberner et al. 1999). In the registration algorithm, matching between scenes is carried out using local correlations in the frequency domain. The result is a correlation

map and the location of the elements with maximum correlation provides the necessary x and y shift to give the best fit. With this automated procedure, over 1,600 points

in an almost complete matrix distribution described by a polynomial with a fit to

within ±0.5 pixel RMSE were produced. Because of the incompatibility of the automatic process due to different resolutions of the two different sensors, a first-degree

polynomial equation was used for the geometric registration process of the LANDSAT

TM and IKONOS data standard techniques with 15 ground control points. As a

re-sampling process, cubic convolution was used with ±0.5 and 3 pixel registration

accuracy for LANDSAT TM and IKONOS XS images, respectively. Because of

being same sensor and of seasonal compatibility (April and June) no atmospheric and

radiometric corrections were applied for the LANDSAT TM images.

LANDSAT TM data, excluding the thermal band, were classified separately

using a supervised classification technique. For both dates the following classes

were considered: settlement, fields, lake and sea (Bỹyỹkỗekmece Lake and some

of the Marmara Sea coast), forest, stone quarries, and industrial areas (Fig. 15.3).



15



Remote Sensing of Urban Land Use Change in Developing Countries



299



Fig. 15.3 Sample pictures for some of land use classes utilized in the case study (fields, stone

quarries, settlement, and industrial areas)



The percentage of their distribution over the 14 administrative units of Istanbul was

calculated from land use classification results for both dates. Classification accuracy analysis for each classification was performed to calculate the confusion

matrix and Kappa coefficient using 50 ground truth sample points, which were

selected independent of training areas.

Using the multi-temporal LANDSAT TM images, change detection methods

were applied to reveal changes in land use between 1984 and 1998. In this study,

both the image differencing and image ratioing methods were utilized. Band 3

(0.63–0.69 mm) of the first LANDSAT TM data (1984) was subtracted/divided

from band 3 of the second LANDSAT TM data (1998). Band 3 was selected

because it is believed to be the best band for cultural/urban feature identification.

The resulting difference image was then classified to reveal changes in areas of

different land uses between the years 1984 and 1998. In addition, the NDVI was

used to evaluate urban change in a focused area, Mimarsinan – one of Istanbul’s

sub-districts known of main land use changes, between 1998 and 2002.



15.5



Results and Discussion



The population growth between 1970 and 1985 in some sub-districts of

Bỹyỹkỗekmece exploded after 1985 to reach a growth rate of over 1000% during

the period 1985–1997 (Table 15.1 and Fig. 15.2). Attractive coastal sub-districts



300



D. Maktav and F. Sunar



such as in Kavakl, Yakuplu, and Kraỗ, for example, witnessed a population growth

rate of 70–125% during the period 1970–1985 and increased to 1,150–2,300% during the period 1985–1998. Even in Hoşdere, which was only a village in 1997, the

increase in population was approximately 1,300% between 1985 and 1997.

Likewise, the increase in Esenyurt was 221% between 1970 and 1985 then reached

372% in the period 1985–1997. The population in this sub-district has been continuously increasing from 1985 (21,000) to 1997 (100,000) and reached a greater

population than the coastal sub-districts and city centre. The main reason for the

increase in population is related to the huge amount of migration from other parts

of Turkey rather than natural population growth. It is obvious that such a population

explosion would cause great land use changes in the area.

The results obtained from the classified LANDSAT TM image of 1984

(Fig. 15.4a, b, Table 15.3) showed that 93.4% (21,013.8 ha) of the Bỹyỹkỗekmece

was covered with agricultural fields. For the fields located in Bỹyỹkỗekmece, its

sub-districts and villages the proportion was approximately 86–98% (except in

Ahmediye). Before 1984 the whole area was covered with watermelon, muskmelon, grain, and sunflower fields, with farming and agriculture being the main

land use activities. On the other hand, in Bỹyỹkỗekmece there were only 3.6%

settlement areas, having the densest building in Mimarsinan with 10.2% and a

minimum proportion of settlement in Ahmediye with only 0.6%. The district had

virtually no forested areas (only 2.4%). In this district, there were some stone

quarries located in a 40.2 ha area and its percentage of coverage area within the

total district was only 0.2%. There were no industrialized areas established prior

to 1984.

According to the results obtained from classified LANDSAT TM data dated

1998 (Fig. 15.5a, b, Table 15.4) the percentage area of the fields averaged 67%.

Fields in Kumburgaz and Türkoba located on the west of the Bỹyỹkỗekmece Lake,

and Karaaaỗ on the northeast of the lake covered 83–86% of the areas, but only

50–78% in other parts of the district (with the exception of Ahmediye, where they

are only 29.3% of the area).

In Bỹyỹkỗekmece, settlements covered 5,242 ha, which is about 1/4 of the total

district. Industrialized areas comprising 3–5% of the study area were mostly located

at the eastern part of the Bỹyỹkỗekmece Lake in Kavakl, Yakuplu, and Kraỗ, and

in Mimarsinan at the southwestern part of the lake. All of these sub-districts have

shores or coastlines, except Kraỗ. Forest areas covered approximately 6% of the

Karaaaỗ, but they were less than 3.2% in all other areas. In the whole district, percentage of the forest areas average 1.7%, with half of it in the Karaaaỗ. Percentage

of land used for stone quarries in the district was 1.5% (340 ha).

In this example, the overall performance of classification is a compound of the

accuracies of the individual classifications which, in turn, depend largely on the

consistency, homogeneity and separability of the original training classes and how

representative they are (Coppin and Bauer 1996). Hence, classification accuracy

analysis was performed after each classification process with error matrix and

Kappa analysis using 50 randomly selected test points which are independent of

training areas from the existing field maps (Table 15.5).



15



Remote Sensing of Urban Land Use Change in Developing Countries



301



b

3000



Settlement



Area (ha)



2500



Field



2000



Lake + Sea



1500



Forest



1000

500



Ahmediye



Türkoba



Kumburgaz



MimarSinan



Esenyurt



Tepecik



Stone Quarry



0



Industrial area



District

Fig. 15.4 Analysis results for the Landsat TM image of 1984. (a) Classified 1984 LANDSAT TM

imagery, (b) calculated areal extents of the land use classes



In Bỹyỹkỗekmece, land use for settlements over these years increased by

almost 20%, from 3.6% to 23.3%. Table 15.6 shows a comparative analysis of the

classification results and displays the extreme increase in the built area within the



Table 15.3 Land use results obtained from classified 1984 LANDSAT TM data

Settlement

Fields

Lake-sea

1984

ha

%

ha

%

ha

%

Bầekmece

88.1

4.1

2,034.5

93.5

19.0

0.9

Kavakl

48.4

4.9

919.3

93.1

0.6

0

Yakuplu

85.6

5.9

1,348.6

93.3

3.0

0.2

Hodere

40.9

2.2

1,735.1

92.7

06

0

Kraỗ

24.0

2.7

858.7

97

0

0

Gỹrpnar

120.0

6.6

1,662.8

91.7

13.9

0.8

Esenyurt

94.4

3

2,976.4

96

0

0

MimarSinan

85.3

10.2

719.6

86.2

26.2

3.1

Kumburgaz

69.8

3.8

1,765.5

95.2

6.3

0.3

Tepecik

58.6

4.2

1,250.8

90.5

15.8

1.1

Tỹrkoba

11.6

1

1,132.9

97.1

0

0

ầakmakl

14.8

1.4

1,061.9

98.3

0

0

Karaaaỗ

62.3

2.1

2,801.8

95

3.7

0.1

Ahmediye

5.5

0.6

745.9

78.2

11.3

1.2

Total

809.4

3.6

21,013.8

93.4

99.9

0.4

Forest

ha

30.9

19.0

3.0

94.

2.1

11.9

30.6

2.5

9.9

36.4

21.9

3.9

82.6

190.9

540.3

%

1.4

1.9

0.2

5.1

0.2

0.7

1

0.3

0.5

2.6

1.9

0.4

2.8

20

2.4



Stone quarry

ha

%

2.8

0.1

0.6

0.1

4.9

0.3

1.4

0.1

0

0

4.8

0.3

0.6

0

1.3

0.1

2.9

0.2

20.4

1.5

0

0

0

0

0

0

06

0.1

40.2

0.2



Ind.

ha

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

%

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0



Total

ha

2,175.3

987.4

1,445.1

1,872.7

884.8

1,813.6

3,102.0

834.9

1,854.4

1,381.9

1,166.4

1,080.6

2,950.3

954.3

22,503.6



%

9.7

4.4

6.4

8.3

3.9

8.1

13.8

3.7

8.2

6.1

5.2

4.8

13.1

4.2

99.9



302

D. Maktav and F. Sunar



15



Remote Sensing of Urban Land Use Change in Developing Countries



303



Area (ha)



b

3000



Settlement



2500



Field



2000



Lake + Sea



1500

Forest



1000



Stone Quarry



500

0

Türkoba



Tepecik



Kumburgaz



MimarSinan



Esenyurt



Yakuplu



Industrial area



District

Fig. 15.5 Analysis result for the Landsat TM image of 1998. (a) Classified 1998 LANDSAT TM

imagery, (b) calculated areal extents of the land use classes



Table 15.4 Land use results obtained from classified 1998 LANDSAT TM data

Settlement

Fields

Lake-sea

Forest

1998

ha

%

ha

%

ha

%

ha

Bầekmece

4,56.1

21

1,488.3

68.4

1,38.3

6.4

40.9

Kavakl

3,55.1

36

591.2

59.9

0.2

0

1.4

Yakuplu

5,93.4

41.1

736.1

50.9

1.1

0.1

6.6

Hodere

4,78.8

25.6

1,243.9

66.4

1.1

0.1

59.4

Kraỗ

317.8

35.9

509.7

57.6

0

0

4.4

Gỹrpnar

4,36.8

24.1

1,330.7

73.4

1.0

0.1

10.4

Esenyurt

1,346.8

43.4

1,636.0

52.7

0.1

0

1.1

MimarSinan

344.8

41.3

437.1

52.4

17.7

2.1

0.2

Kumburgaz

299.6

16.2

1,531.3

82.6

3.0

0.2

11.9

Tepecik

158.9

11.5

958.4

69.4

2,13.1

15.4

14.3

Tỹrkoba

155.9

13.4

974.6

83.6

0.1

0

27.1

ầakmakl

199.9

18.5

841.1

77.8

0.2

0

12.7

Karaaaỗ

83.1

2.8

2,521.7

85.5

1,11.8

3.8

1,84.3

Ahmediye

15.2

1.6

279.4

29.3

630.5

66.1

7.4

Total

5241.9

23.3

15079.4

67

1,118.1

5

382.1

%

1.9

0.1

0.5

3.2

0.5

0.6

0

0

0.6

1

2.3

1.2

6.2

0.8

1.7



Stone quarry

ha

%

34.9

1.6

7.3

0.7

49.8

3.4

45.3

2.4

7.8

0.9

7.4

0.4

66.2

2.1

9.0

1.1

4.9

0.3

28.6

2.1

4.8

0.4

6.2

0.6

44.6

1.5

21.8

2.3

338.2

1.5



Ind.

ha

16.0

32.3

58.2

44.2

45.4

27.3

51.9

26.1

3.8

8.7

3.8

20.6

4.9

0

343.9

%

0.8

3.3

4

2.4

5.1

1.5

1.7

3.1

0.2

0.6

0.3

1.9

0.2

0

1.5



Total

ha

2,175.3

987.4

1,445.1

1,872.7

884.8

1,813.6

3,102.0

834.9

1,854.4

1,381.9

1,166.4

1,080.6

2,950.3

954.3

22503.6



%

9.7

4.4

6.4

8.3

3.9

8.1

13.8

3.7

8.2

6.1

5.2

4.8

13.1

4.2

99.9



304

D. Maktav and F. Sunar



15



Remote Sensing of Urban Land Use Change in Developing Countries



Table 15.5 LANDSAT TM image classification accuracy analysis

1984a

Ground truth data (%)

Class

Forest

Settlement

Stone quarry

Field

Forest

100.0

0.00

0.00

31.33

Settlement

0.00

87.74

0.00

0.00

Stone quarry

0.00

12.26

100.0

13.67

Field

0.00

0.00

0.00

55.00

Lake + sea

0.00

0.00

0.00

0.00

Total

100.0

100.0

100.0

100.0



305



Lake + sea

0.00

0.00

1.55

0.00

98.45

100.0



Total

12.28

6.60

12.35

11.45

57.32

100.0



1998b

Ground truth data (%)

Class



Forest



Settlement



Stone

quarry



Industry



Forest

76.00

0.00

0.00

0.00

Settlement

0.00

97.17

13.85

43.90

Stone quarry

0.00

0.00

21.54

0.00

Industry

0.00

2.83

64.62

56.10

Field

24.00

0.00

0.00

0.00

Lake + sea

0.00

0.00

0.00

0.00

Total

100.0

100.0

100.0

100.00

a

Kappa coefficient: 0.8201; overall accuracy: 88.8%

b

Kappa coefficient: 0.8953; overall accuracy: 92.5%



Field



Lake + sea



Total



0.00

0.00

0.00

0.00

100.00

0.00

100.0



0.00

0.00

0.00

0.00

0.00

100.00

100.00



1.67

32.43

1.23

6.57

19.89

38.21

100.00



sub-districts Esenyurt (1,252.3 ha, 40.4%), Yakuplu (507.8 ha, 35.2%), Mimarsinan

(259.4 ha, 31.1%), Kraỗ (293.7 ha, 33.2%) and Kavakl (306.6 ha, 31.1%), most

of which are located on the eastern side of Bỹtyỹkỗekmece Lake. An opposing

trend could be observed in the settlements of Ahmediye and Karaaaỗ, where the

built area in each district increased only by about 1%, which corresponds to a

total area of less than 100 ha. Analysis of the land classified as field in

Bỹyỹkỗekmece revealed a loss of 26.4% of agricultural fields over the period of

14 years. The areas with maximum loss of fields were again located on the eastern side of Bỹyỹkỗekmece Lake: Esenyurt (1,340.4 ha, 43.3%), Yakuplu (612.5

ha, 42.4%), Kraỗ (349 ha, 39.4%), and Kavakl (328.1 ha, 33.2%). Apparently,

there is a correlation between the loss of fields and the increase in settlement

areas (Fig. 15.6).

As for fame lands, it is evident from the analysis that villas and new apartment

houses in Bỹyỹkỗekmece rapidly depleted agricultural lands, an observation supported by ground truth data. Using the same analytical methods, it was found that

Ahmediye experienced a 49% loss of fields but only a 1% increase of settlements.

Interpreting the two different years of LANDSAT TM images, one can easily detect

that a significant portion of land in Ahmediye had been submerged by the water of

the Bỹyỹkỗekmece Lake over the study period. The reason for this interesting event

is the enlargement of the Bỹyỹkỗekmece Lake from a lagoon to a lake following



Table 15.6 Comparative analysis of the tables obtained from the classified LANDSAT TM images

Population

Settlement

Field

Lake + sea

19851997

19841998

19841998

19841998

%

ha

%

ha

%

ha

%

Bầekmece

268.2

367.9

16.9

546.2

25.1

119.3

5.5

Kavakl

2,297.2

306.6

31.1

328.1

33.2

0.1

0

Yakuplu

1,335.0

507.8

35.2

612.5

42.4

1.9

0.1

Hodere

1,297.7

437.9

23.4

491.2

26.3

0.4

0.1

Kraỗ

1,153.4

293.7

33.2

349

39.4

0

0

Gỹrpnar

577.6

316.8

17.5

332.1

18.3

12.9

0.7

Esenyurt

472.4

1252.3

40.4

1,340.4

43.3

0.1

0

Mimarsinan

272.4

259.4

31.1

282.5

33.8

8.5

1

Kumburgaz

224.2

229.8

12.4

234.3

12.6

3.3

0.1

Tepecik

97.6

100.3

7.3

292.4

21.1

197.4

14.3

Tỹrkoba

448.6

144.4

12.4

158.3

13.5

0.1

0

ầakmakl

418.3

185.1

17.1

220.9

20.5

0.2

0

Karaaaỗ

117.5

20.8

0.7

280.1

9.5

108.1

3.7

Ahmediye

62.1

9.7

1

466.6

48.9

619.2

64.9

Total

309.9

4,432.5

19.7

5934.4

26.4

1,018

4.6

Forest

19841998

ha

%

10

0.5

17.6 1.8

3.6

0.3

35.3 1.9

2.4

0.3

1.5 0.1

29.5 1

2.3 0.3

1.9

0.1

22.1 1.6

5.2

0.4

8.8

0.8

101.8

3.4

183.5 19

158.3 0.7



Stone quarry

19841998

ha

%

32.1

1.5

6.6

0.6

44.9

3.1

43.9

2.3

7.6

0.9

2.6

0.1

65.6

2.1

7.8

1

2.1

0.1

8.2

0.6

4.8

0.4

6.2

0.6

44.6

1.5

21.2

2.2

298

1.3



Industry

19841998

ha

%

16.8

0.8

32.3

3.3

58.2

4

44.2

2.4

45.4

5.1

27.3

1.5

51.9

1.7

26.1

3.1

3.8

0.2

8.7

0.6

3.8

0.3

20.6

1.9

4.9

0.2

0

0

343.9

1.5



306

D. Maktav and F. Sunar