Microsoft Word - 122
BIOTROPIA VOL. 13 NO. 2,2006 : 122 -131
ASSESSMENT OF COASTAL LAND USE CHANGES
IN BANTEN BAY, INDONESIA USING DIFFERENT CHANGE
DETECTION METHODS
PUVADOL DOYDEE¹ AND VlNCENTIUS SlREGAR²
'Department of Fishery Management, Faculty of Fisheries, Kasetsart University,
Bangkok, 10900, Thailand
2SEAMEO BIOTROP, Jl. Raya Tajur Km 6, P.O Box 116, Bogor Indonesia E-mail:vincents@biotrop.org
ABSTRACT
Many types of the coastal land use in Banten Bay have been assessed in order to know the change as evidently
detected by Landsat imagery in 1994 and 2001. Image processing such as, supervised classification and various
change detection techniques are performed to the satellite images. Red Green method showed the best result for
detecting the coastal land use change. This method is suitable for detecting the increasing areas of the paddy fields and
settlement. Image Differencing method is capable to detect the increasing areas in agriculture, decreasing in fishponds
and natural areas. Image Ratioing method can be considered suitable for detecting the increasing area of fishponds,
decreasing of paddy fields and agriculture areas. Each coastal land use type has increased, except for the natural
area/brushwood. Most of agriculture and paddy fields areas have been converted to fish ponds.
Key words: coastal land use/land cover, change detection method, remote sensing
INTRODUCTION
As a transition area between land and sea, coast is characterized essentially by geologic
nature of the land. Sometimes, coastal zone is defined depending on the project purposes, for
example LOICZ project (Holligan and de Boois 1993). This zone in many cases is relatively
unstable and fragile (Carter 1989), but became an important place for the human activities,
such as culture, economic, fisheries, industry, tourism and social (Rais and de Boois 1997).
More than 60 % of the human population lives in coastal area and it will increase in the future,
hi many cases, coastal area is the best place or the only available space for city development,
agriculture, fisheries, industry and tourisms. This condition will cause the increasing demand for
space for such activities. However, in coastal zone a land use change is obviously a phenomenon
(LOICZ 1994; Williams et al 1997; NOAA 1998) This change is an essential matter that should
be monitored for planning or avoiding any further changes that can damage or harm the
environment. Remote sensing technique is one of the effective tools in monitoring
phenomena that change continuously overtime and cover large areas (Aronoff 1991). One
of the major applications of remote sensing technique is change detection method. Many ready-
made algorithms are available to perform that method. Despite of their differences in identification,
an accurate spatial registration of the various dates of imagery is a requirement for all methods
used.
Corresponding author: vincents@biotrop.org 122
Asessment of coastal land use changes - P. Doydee and V. Siregar
This study is intended to assess the capability of four methods of change detection in detecting
the change in coastal land use, and to compare the result of those methods with each other. Location
of this study is the coastal zone of Banten Bay.
Banten Bay is an area most frequently used for various researches, since in this region many
activities are taken place, such as agriculture, fishing and mariculture, harbor, and industry. There is
no doubt that these activities will influence the coastal habitats and resources.
MATERIALS AND METHODS
The study area is located in Banten Bay at the northern coast of Banten Province
(Figure 1). Banten bay has an area of about 120 km2 with water depth of less than 13 m.
Geographically, this bay lies on 05°55' S-06°05' S and 106°05'E -106°15'E and the distance
from Jakarta is about 175 km. This bay is also characterized as a lowland area (Saptarini
2000).
Data used comprised satellite data of Landsat-5 TM (with the path/row 123/064 and
acquisition date of April 6, 1994), and Landsat-7 ETIVT (with path/row 123/064 and acquisition
date of August 7, 2001, NASA, 2002), topographic and geomorphological maps. Different
software have been used such as image processing ER Mapper (ER Mapper 1997), Banten Bay
Information System 3.0 and Geographic Information System.
Supervised classification method used has the objective to identify and classify the features in
an image quantitatively (Richards 1995; Lillesand and Kiefer 1994; Congalton and Green 1999;
Yusuf 2001). This involves the analysis of multi-spectral image and the application of
statistically based decision rules for determining the coastal land use type of each pixel on an
image (Middlekoop and Lif 1991).
Many different techniques have been developed for change detection. These methods
involve change extraction and change classification. (Coppin et al. 2002; Jensen 1986). The use
of change detection techniques in this study comprised four methods, including Red Green,
Image Differencing, Image Ratioing and Principal Component Analysis methods. Image analysis
was performed to assess the coastal land use change between two different dates by overlaying the
images of Landsat-5 TM in 1994 and Landsat-7 ETM+ in 2001, and applying the change
detection techniques as stated above.
Prior to the application of change detection techniques , the images utilized have to be
normalized using atmospheric correction technique of histogram adjustment in order to
eliminate error due to different atmospheric conditions of the images of different periods
(Wallace and Campbell 1996). Image of 1994 was rectified to image of 2001 which has been
already corrected (Geodetic datum WGS 84 and Map projection SUTM 48).
123
Coastal Land use in 1994 and 2001
The type of coastal land use in 1994 and 2001 are grouped into 5 classes such as, paddy
fields, fishponds, settlement, agriculture and natural areas (brushwood). The maps of coastal
land use were prepared from the result of supervised classification using the band
combination of 542/RGB (Figure 2.)
Change Detection Techniques Red Green Method
This method involves displaying simultaneously one dataset in green and one dataset in red,
namely image in 1994 was displayed as red and image in 2001 was displayed as green. The red
color means that these areas only exist in image 1994 and the green color in image 2001. For
example in Lontar district (Figure 3) as zoomed-in, paddy fields in 1994 have been converted to
fishponds in 2001. The red color is represented as paddy fields before and those that have been
converted into fishponds in 2001. This could be confirmed by using geomorphological map and
thematic coastal land use map resulted from supervised classification.
124
Asessment of coastal land use changes - P. Doydee and V. Siregar
Image Differencing Method
The digital number value of image in 2001 was subtracted from digital number
value of image in 1994. The basic concept of this method is that, in the raster data,
each band has its digital value (DN). Therefore, if it is subtracted from the same DN
value of each pixel in both images, the output is equal to 0. The output might be
minus or plus in the case that those pixels do not have the same DN values, this
means that there is a change of value between the two images. This change might be
due to the change of characteristic of the object. The output of this method was
displayed as pseudo color area (Figure 4-left). Prior to data overlay, the raster data
should be converted from raster cells into vector polygon and then overlaid that
vector (change portion) with the original image so that the raster and vector polygon
could be distinguished (original image), the result is presented in Figure 4 (right).
Figure 4. The result of image differencing method.
125
BIOTROPIA VOL. 13 NO. 2, 2006
Image Ratioing Method
Actually, this is a simple method as Image Differencing method where in this
method the formula is changed from subtraction to division. The data from two
registered images are divided pixel by pixel. Therefore, the areas in different dates
that have the same value of DN is equal to 1, it means that those areas do not
change. Bands of TM1, TM2 and TM3 are highly correlated. The lower the
correlation between two bands the greater the information content. Ratio of TM2
and TM5 provide some subtle wetland information (Jensen 1986). The result of this
method is presented in Figure 5.
Figure 5. The result of
image ratioing method Principal Component Analysis Method
The PC#2 has maximum variance of the original data set, therefore it is used
for analysis. PC#2 data consist of two bands inputs, where input 1 and input 2 were
images of 1994 and 2001, respectively. The result showed that the bright blue color
appears covering the areas (Figure 6-left). The output should be converted from
raster cells into vector polygon in order to facilitate the interpretation (Figure 6-
right).
Figure 6. The result of principal component analysis method
126
Asessment of coastal land use changes - P. Doydee and V. Siregar
Change Analysis
The entire study area was 46 785.69 ha consisting of 23 437.26 ha terrestrial zone and
23348.43 ha water. The classification of coastal land use (type) between two dates was processed
by post-supervised classification method (Cho 2000) and the result is presented in Table 1 .
Table 1 shows that the area of each land use type has increased, except for the natural
area/brushwood which shows the contrary. The main factor caused by this change is the
increasing numbers of the population and human activities in this area. From the field check, it was
observed that some parts of agriculture and paddy fields are converted into fishponds. It is
known that in 1990's fishponds were very promising, and for that reason most of villagers
and farmers converted their agriculture area and paddy fields into fishponds in the coastal zone.
According to Table 2, there are two types of changes, that is the increasing and the decreasing
area. Obviously, the total areas of coastal land use converted was 7706.79 ha .These changes
not only happen from one type of land use to other types, but also in the area of the same land use
type. For example of decreasing area, it can be noted as follows: Paddy fields as in 1994 have been
converted into different land use types, such as fishponds (812.46 ha), settlements (70.16 ha), and
agriculture areas (712.66 ha), respectively in 2001. Fishponds as in 1994 have been converted
only into settlement (8.15 ha) in 2001, while for settlement there is no change in terms of land
use type and the area between the two periods. Agriculture as in 1994, has been converted into
paddy fields (1015.06 ha), fishponds (761.39 ha) and settlement (317.18 ha), respectively in
2001. Natural areas (brushwood) in 1994 have been converted into paddy fields (998.51 ha),
fishponds (727.27 ha), settlement (320.19 ha) and agriculture (1963.76 ha) in 2001. Meanwhile, the
increasing areas of the land use types from 1994 to 2001 are as follows: Paddy fields are increased
as
127
BIOTROPIA VOL. 13 NO. 2,2006
much as 2013.57 ha from agricultural (1015.06 ha) and natural areas (998.51 ha); the area of
fishponds are increased to 2301.12 ha from paddy fields (812.46 ha), agricultural (761.39 ha)
and natural areas (727.27 ha); settlement areas are increased 715.68 ha which comprised paddy
fields (70.16 ha), fish ponds (8.15 ha), agriculture (317.18 ha) and natural areas (320.19 ha)
and; agriculture areas are increased to 2,676.42 ha that is from paddy fields 712.66 ha and
natural areas about 1903.76 ha. The total of land use type (7706.79 ha) then is considered as a
reference for selecting the best method of change detection to be applied.
Using the four converted detection methods, the converted areas of each method were
as follows: the Red Green method (7094.97 ha), Image Differencing method (6185.25 ha),
Image Ratioing method (8490.96) ha and Principal Component Analysis (PC#2) (5920.47 ha),
respectively. From the result presented in Table 3, it could be concluded that Red Green method
is better for detecting the coastal land use change compared to other methods. Saptarini
(2000) has also concluded that Red Green method is simple and easy method to detect the change
of areas in coastal zone of Banten bay. This method shows that the total number of converted
areas closely agrees with the number of converted areas of the reference.
As stated above, the total of each land use type has increased from 1994 to 2001, except
for natural areas/brushwood which show the contrary.
128
Asessment of coastal land use changes - P. Doydee and V. Siregar
The increasing and decreasing areas of each type of land use were derived from four change
detection methods. For increasing area in general, Red Green method is the best compared to the
other methods. But, if the increasing area of each type of land use is observed, we found that
only PC#2 has no land use types which has a wide close to the reference (Table 4). It means that
each change detection method is suitable to detect certain land use type.
Table 4. The increasing area of each land use type using the four change detection methods compared to the reference.
Type Red Green Image Differencing Image Ratioing PC#2 References
Paddy fields 1,969.86 1,680.39 2,088.15 1,802.88 2,013.57
Fishponds 2,161.82 1,806.74 2,394.97 1,988.55 2,301.12
Settlement 804.84 439.41 827.13 493.20 715.68
Agriculture 2,158.44 2,258.72 3,180.70 1,635.85 2,676.42
Total 7,094.97 6,185.25 8,490.96 5,920.47 7,706.79
Therefore, it could be concluded that Red Green method is suitable for detecting the
increase of paddy fields and settlements. Differencing method is better in detecting the increasing
agriculture areas. Meanwhile, the Ratioing method could be considered as the best method to detect
the increasing fishpond areas (Table 5).
For decreasing areas, all the change detection methods used showed that the area of each type of
land use is far less under the area of reference (Table 6).
Meanwhile, if the decreasing area of each type of land use provided by the change
detection methods used is compared with the reference, it could be
129
BIOTROPIA VOL. 13 NO. 2,2006
concluded that Image Differencing method is better for detecting the decreasing area of
fishponds and natural areas compared with other methods and image Ratioing method
could be considered as the better method for monitoring the decrease of paddy fields and
agriculture areas (Table 7).
This study showed that no specific change detection method could be used to detect all land use
types. Each method could detect better a specific land use type compared to the other
methods.
CONCLUSIONS
Supervised classification of remotely sensed data (Landsat data in!994 and 2001) of the
coastal zone of Banten Bay provides 5 land use types such as paddy fields, fishponds,
agricultural area, settlement and brushwood (natural area). These land use types served as the
reference for land use change during those two periods. To detect the change of land use types
in those areas, several change detection methods are performed to the Landsat satellite data.
Change analysis showed that among various change detection methods used, the Red Green
method has provided the best result for detecting the coastal land use change. This method is
suitable for detecting the increasing areas of the paddy fields and settlements. Meanwhile,
Image Differencing method is capable to detect increasing areas in agriculture, decreasing in
fishponds and natural areas. Image Ratioing method could be considered suitable for detecting
the increasing area of fishponds, decreasing of paddy fields and agriculture. Each coastal land use
type in the study areas had increased, except for the natural areas, and most of agriculture and
paddy fields areas have been converted to fishponds. Certain change detection methods used has
shown a good performance for detecting the change of certain coastal land use types in Banten
bay, however, several tests in different areas are needed to be conducted in order to know the
level of accuracy of each method used.
ACKNOWLEDGMENTS
The authors would like to acknowledge the funding support of the SEAMEO Regional
Center for Graduate Study and Research in Agriculture (SEAMEO SEARCA), and
technical support and assistance of Mr. Nyoman Sukmantalya in providing all the data needed.
130
Asessment of coastal land use changes - P. Doydee and V. Siregar
REFERENCES
Aronoff, S. 1991. Geographic Information Systems. A Management Perspective, WDL Publications,
Ottawa, 294 p. Carter, R.W.G. 1988. Coastal Environment. An Introduction to the Physical, Ecological and
Cultural
Systems of Coastlines, Academic Press, Toronto, 617 p. Cho, S.H.2000. Digital Change Detection by Post-
classification Comparison of Multi-temporal Remotely
sensed data. University of Idaho, Departement of forest resources. Journal of the Korean
Society of Remote Sensing col 16, No 4, p. 367-373. www.ksrs.or.kr/iournal full/1 6-
4%20ftp/8josung.pdf Congalton, R. G. and K. Green. 1999. Assessing the Accuracy of Remotely Sensed Data:
Principles and
Practices, Lewis publishers, Boca Raton, 137 p Coppin, P., Lambin, E., Jonckheere, L., and B. Muys 2002.
Digital change detection methods in natural
ecosystem monitoring. In Analysis of Multi-temporal remote sensing images. Bruzzone,L.and
Smiths,P.eds. World Scientific, New Jersey. 440 p. ER Mapper 5.5 .1997. ER Mapper Workbook - level One,
Training Workbook, 25 August 1997,
California, 382 p. Holligan, P.M. and H. de Boois. 1993. Land-Ocean Interactions in the Coastal Zone
(LOICZ). Global
Change Report (25). IGBP. Stockholm, 50 p, URL: http:// ww\v,
cs.utk. edu/~l ucas/nubl icat ions/publ ications , html Jensen, J.R. 1986 Introductory Digital Image Processing. A
Remote Sensing Perspective. Prentice-Hall,
Englewood Cliffs, 3 18 p. Lillesand T. M. and R. W. Kiefer .1994. Remote Sensing and Image Interpretation (3 rd
Edition), John
Wiley and Sons, 750 p. LOICZ .1994. Land-Ocean Interactions in the Coastal Zone, Core Project of the
International Geosphere
Biosphere Programme of the International Council of Scientific Unions, p. 6. Middelkoop, H. and J. Lif. 1991.
Implemention on temporal relationship in: Knowledge Base System of
satellite image, Photogrammetric Engineering and Remote Sensing. Vol.57, July 1991, p. 937-
945. NASA .2002. Landsat 7 Science Data Users Handbook, URL: http://ltpwww.gsfc.
nasa.gov/IAS/handbook/handbook_htmls/chapter8/chapter8.html
NOAA .1998. Shoreline, Coastal zone Glossary, URL: http://www.csc.noaa.gQv/ sho reline/glossarv.html Rais, J., Dutton,
M., Pantimena, L., Plouffe, J. and R. Dahuri, 1997. Integrated Coastal and Marine
Resource Management, Proceedings of International Symposium, 25-27 November 1997,
Batu, ITN, BAKOSURTANAL and Proyek Pesisir, Malang. p. 273-276. Richards, J.A. 1995. Remote Sensing
Digital Image Analysis, Springer- Verlag, Berlin, p. 54-57. Saptarini, D. 2000. Coastline changes detection using remote
sensing technique: Banten Bay study case,
Master in Information Technology for Natural Resources Management, Bogor Agriculture
University (1PB), Bogor, 53 p. Wallace, J. and N. Campbell. 1996 Evaluation of the Feasibility of Remote
Sensing for Monitoring,
National State of the Environment Indicators, Technical Paper Series, Department of the
Environment, Canberra, p. 30. Williams, S.J., K. Dodd K, and K.K. Gohn .1997. Coast in Crisis, USGS
Homepage. URL:
http://www.usgs.gOv/http://www.usgs.gov/ Yusuf, D.N. 2001. Remote Sensing and GIS for Landuse Change
Analysis, Case study in Kab. Bandung,
Master in Information Technology for Natural Resources Management, Bogor Agriculture
University (IPB), Bogor, 68 p.
131
122.pdf
123.pdf
124.pdf
125.pdf
126.pdf
127.pdf
128.pdf
129.pdf
130.pdf
131.pdf