GISdevelopment.net ---> AARS ---> ACRS 1997 ---> Poster Session 1

A Study Of Land Use/Land Cover Changes From 1988 To 1994 In Nilai and Surrounding Areas

Mazlani Muhammad
Malaysian Center for Remote sensing ( Macres)
Letter Box 208,Lot CB 100, 5th Floor City Square Center
Jalan Tun Razak, 50400 kuala Lumpur , Malaysia
Tel: 603-2645640 Fax : 603-2645650
E-mail : sda@macres.gov.my

Abstract
Nilai consists of several new towns and potential areas of development . it also consists of diversity of land cover/land use such as industrial commercial, residential, agriculture and natural vegetation. Since it is in strategic location , the Government decided to develop Nilai and it's surrounding areas to new townships and two major projects - New Government Administrative Center and Kuala Lumpur International Airport. The aim of this project is to determine the changes of land cover/land use using change detection techniques such as Normalized Difference Vegetation Index, principal component Analysis, Image Ratioing and Image Differencing. Two sets of Landsat TM data acquired on 17 April 1988 and 2 April 1994 were used in this project. Results showed that development occurred at Bandar Baru Nilai and most of the settlement or ' kampongs ' areas were converted to plantations.

Introduction
Remote sensing technology has been used for many years to acquire data such as for land use or land cover application. It provides a useful aid to map and monitor the changing condition of land surfaces . it allows large scope for analysis, classification and interpretation of the image data, whereby it gives a lot of information form a single image. even though remote sensing cannot solve or manage land surfaces problems , but it provides the required information which when integrated with some other data, it will gives a between ability to do the planning and implementation of a certain project.

Forster ( 1983); Martin and Howarth ( 1989) found that images multi-spectral satellite( Landsat MSS , TM, SPOT ) can give information on change detection for example form rural to urban at lower cost with better accuracy than ground surveys an analysis of aerial photographs . this is also supported by Jensen et al. ( 1983) " satellite-based remote sensing is a potentially powerful tool for monitoring of land use or land cove changes with high temporal resolution and at lower cost than using traditional land survey methods " .

Milne ( 1995) said that it is an alteration in the surface components of the landscape. Some of the more common types of changes detectable on remotely sensed data are associated with he clearing of natural vegetation, increased cultivation, urban expansion, the changing surface level in bodies of standing water, vegetation regeneration after wildfires and soil disturbances resulting form mining, landslides and the overgrazing of animals . if there is change in land cover, there is an alteration of spectral reflectance's recorded by the sensors. Sabins ( 1987) stated that change detection images provide information about seasonal or other changes then the information is extracted by comparing two or more images of an area that were acquired at differences times.

Study Area
The study area, Nilai is located 25 km south of Kuala Lumpur ( capital of Malaysia ) and 20 km north of seremban (state capital of Negeri Sembilan ). Erinco ( 1992) found that Nilai contains the diversity of land use- industrial and commercial, residential which includes settlement or ' kampungs ' , agriculture ( rubber and palm oil ) as well as natural vegetation. The original images of 1988 and 1994 area as Plate 1 and Plate 2 respectively. The cover types were characterized based on their spectral response ( the categories are presumed and may not represent exact land use or cover types in the area because no field work or surface sampling and topographic map is available ). Industrial, commercial and settlement or ' kampungs' - cyan ( due to the reflectance of blue and green from the concrete, cement, asphalt etc.); developing residential-brown ( due to the reflectance of NIR and visible red of sparsely vegetated areas ); agriculture-red ( due to the NIR reflectance ); natural vegetation - bright red (due to the NIR reflectance ) ; natural vegetation - bright red( due to the dominance of NIR ); estate with plantation - pink red ( due to red reflectance (NIR) is more than blue and green ).


Plate 1: Nilai and surrounding areas in 1988
R G B = 4 3 2


Plate 2: Nilai and surrounding areas in 1994
R G B = 4 3 2

Aim and Objective
The aim of this project is to use high resolution remotely sensed data, Landsat TM, to identify the changes of land use/land cover of Nilai and surrounding areas over time. This will be examined by comparing the data acquired at two different dates: 17 April 1988 and 2 April 1994. from this analysis , the land cover/land use change classification map will be produced. Since the images were acquired on almost the same date , there will be no difference due to solar elevation , solar irradiance and solar zenith angle ( Forster, 1995).

Methods
The changes detection techniques to be used are Normalized Difference Vegetation Index ( NDVI) , Principal Component Analysis ( PCA), Image Differencing and Image Rationing . the two images used in this project were geometrically corrected by using image to image registration. As forester ( 1990) said , " image to image registration is two to three times more accurate than image to map registration ' and also the first order is adequate ( Forster, 19950.

Normalised Difference Vegetation Index ( NDVI)
For the change detection analysis, the bands for both years were combined as below:

Table 1. band combination for year 1998 and 1994
Year Old band New band
1994 1 1
1998 2 2
1994 3 3
1988 4 4
1988 3 5
1994 2 6
1994 4 7

The NDVI was performed by ( Band 4- band 3)/ ( Band 4+ Band 3) and the images is plate 3. from the image, areas where vegetation has been cleared between 1988 and 1994 is white, ' developed ' areas in 1988 and 1994 is black and the areas where ' development ' occurred in 1994 is grey in colour .


Plate 3: Change Detection using NDVI

Principal component Analysis ( PCA)
Principal component was performed for each data sets- Landsat TM 1988 and 1994. the results of the PCA are given in Table 2 and Table 3 below. It shows that from the first four PCA, there were some changes occurred in Nilai an surrounding areas.

Table2: Eigenmatrices and eigenvalues for year 1988
TM components
Bands 1 2 3 4 5 6 7
1 0.42 -0.01 -0.26 0.33 0.78 0.17 -0.02
2 0.43 0.03 -0.23 0.29 -0.56 0.61 0.06
3 0.43 -0.09 -0.23 0.30 0.27 -0.75 -0.17
4 0.13 0.94 0.26 0.11 0.01 -0.10 0.12
5 0.42 0.09 0.06 -0.62 0.12 0.10 -0.64
6 0.29 -0.31 0.87 0.25 0.01 0.03 -0.01
7 0.43 -0.10 -0.02 0.50 0.02 -0.11 0.73
Eigen value 26849 5391 3323 1457 488 267 153

Table 3: Eigenmatrices and eigenvalues for year 1994.
TM component
Band 1 2 3 4 5 6 7
1 0.41 -0.09 -0.13 0.47 -0.76 0.08 -0.01
2 0.42 -0.05 0.07 0.31 0.49 0.69 -0.003
3 0.42 -0.14 0.03 0.37 0.40 -0.71 -0.04
4 0.10 0.96 -0.14 0.13 0.03 -0.07 0.15
5 0.41 0.18 0.36 -0.42 -0.11 -0.01 -0.69
6 0.34 -0.10 -0.83 -0.43 0.07 0.004 -0.01
7 0.42 -0.05 0.38 -0.41 -0.10 -0.04 0.70
Eigen value 26696 5507 2024 1728 1728 429 232

Image Differencing
Image differencing was performed by subtracting B41994 for B41988 and the image is as Plate 4. Most of the changes were due to the vegetation changes either by cultivation or land clearing for development purposes by the state government. Light areas indicate that land cover changed from vegetated areas in 1988 to ' developed ' areas in 1994; from less dense ' kampungs' in 1988 changed to more densed ' kampungs' in 1994 and there were more green vegetation in 1988 than in 1994. the black areas were the areas the vegetation increased I 1994. these were due to the new settlements with young plantations in 1988 became matured in 1994, or either the ' kampungs' areas converted to the plantation areas. The gray areas were the areas the experienced no change in 1988 and 1994.


Plate 4: Change Detection using image Differencing

Image Ratioing
For change detection analysis, band 4 ( NIR0 was used to see the changes in vegetation cover in the study area ( Band4 1994 / Band 41994) and the image is as Plate 5. from the output images, it is clearly shown that the white region were the regions that have more vegetation in 1988 than in 1994. the grey regions were the unchanged regions between 1988 and 1994. whereas the black regions were the regions where it was more vegetated in 1994 than in 1988. this was due to the fact that in 1988, the new settlement areas with young plantations became matured plantations in 1994 and also some of the 'kampungs' were converted to the plantations area.


Plate 5: Change Detection using image Ratioing

Classification of change and No-change Areas.
The combination of bands from the different dates will allow the determination of changes in the area. The band combination was taken from table 1- band 4 ( NIR 1988), band 7 ( NIR 1994) and Band 1( blue 1994). They were represented as R GB respectively and the image is Plate 6. this combination was found to be best demonstrated the land cover changes since vegetation as a high response in near infra-red ' development ' has a high response blue. From the image, it shows that areas which were vegetation in 1988 and in 1994, has a high response in NIR1988 and NIR1994 ; areas which were ' developed ' in 1988 and in 1994, has a high response in B1994; areas which were vegetation in 1988 but ' developed in 1994, has a high respo9nse in NIR1988 and B1994; and areas which were vegetation in 1988 and became more vegetated in 1994 has a high response in NIR1994 but very little in NIR1988.;


Plate 6: Band Combination - 4 (NIR 1988) 7 (NIR 1994) 1 (B 1994)

Using the band combination R,G,B= 4 7 1 the image could be classified into areas which remained ( unchanged ) vegetation in 1988 and 1994; areas which remained ( unchanged development in 1988 and 1994; areas which changed from less vegetated in 1988 into more vegetated ( plantations ) in 1994.

The classification firstly done by displaying the image using a combination of bands 4 7 1 . supervised classification was used because we can train the classifier to recognize the spectral classes but not the unsupervised classification ( MACRES,_______) . to differentiate the changes, a representative pixels form each area was chosen and then used to classify the image and the classified map is plate 7. the area which remained ( unchanged ) vegetation in 1988 and 1994 is yellow ; areas which remained ( unchanged ) development in 1988 and 1994 is blue; areas which changed from vegetation in 1988 into development in 1994 is magenta ; and areas which changed from les vegetated in 1988 into more vegetated in 1994 is green in colour.


Plate 7: Land Cover/Land Use Change in Nilai and Surrounding Areas from 1988 to 1994

Conclusion
This project shows that there changes in the land cover/land use in Nilai between the years 1988 and 1994. some areas remained vegetated in 1988 and 1994- these were mainly natural vegetation, oil palm and rubber plantations. The only development that occurred in the study area is at the center-right in the image, kniown as Bandar Baru Nilai which consists of light industrial, commercial and residential areas. A number of areas at the left of the images were changed form less vegetated in 1988 into more vegetated in 1994. these areas were settlement or ' kampungs ' areas that were converted to plantations.

Reference:
  • Erinco Sdn Bhd., 1992. preliminary Environmental Impact Assessment study for the Proposed Bandar Baru Nilai, Petaling Jaya, Malaysia .
  • Forester, B.C. , 1983. some Urban Measurements from Landsat Data. Photogrammetric Engineering and Remote sensing, Vol 49, pp 1693-1707.
  • Forester, B.C. , 1990. Application of Normalized vegetation Index Difference for Urban Change Monitoring -Urban Development on the Fringe of Sydney. Proceedings of the 23rd International symposium on Remote sensing of Environment, Bangkok, Thailand, Vol.2. pp 215-220.
  • Forester, B.C. , 1995. introduction to Remote sensing and Image Analysis. Notes of the Short course Conducted for Malaysian Center for Remote sensing 's Research Officers, Organized Jointly by University of New south wales and MACRES on 30 January- 18 February 1995.
  • Jenson , J.R., et. Al, 1983. Urban/Suburban Land Use Analysis, In Manual of Remote sensing, edited by R.N. Colwell. Fall Church , Virginia , American Society of Photogrammetry, Vol. 2. pp 1571-1666.
  • MACRES,________. Lecture Outline for Short Course . Malaysian center fro Remote sensing, Kuala Lumpur, Malaysia .
  • Martin , L.R.G., and P.J.H Howarth, 1989. Change Detection Accuracy Assessment Using SPOT Multispectral Imagery of the Rural -Urban Fringe. Remote sensing of Environment, Vol. 30, pp 55-66.
  • Milne, A.K., 1995. Change Detection . Notes of a Short course on Introduction to Remote sensing and Images Analysis, Organized Jointly by the University of New South Wales and MACRES on 30 January- 18 February 1995.
  • Sabins , F.F., 1987. Remote sensing Principles and Interpretation. W.H. Freeman and Co. USA.