Land Use Trends Analysis Using SPOT-5 Images and Its Effect on the Landscape of
Cameron Highland, Malaysia
Mohd Hasmadi Ismail1*, Che Ku Akmar Che Ku Othman1, Ismail Adnan Abd
Malek1 and Saiful Arif Abdullah2
1
Forest Surveying and Engineering Laboratory, Faculty of Forestry
Universiti Putra Malaysia, 43400 UPM, Serdang, Selangor, Malaysia
*email: mhasmadi@putra.upm.edu.my
2
Institute for Environment and Development (LESTARI), Universiti Kebangsaan
Malaysia, 43600 Bangi, Selangor, Malaysia
Abstract
A large part of the mountain steep land in Peninsular Malaysia is covered by forests.
Cameron Highland is a mountainous region with a climate favourable to the cultivation
of tea, sub-tropical vegetable and flowers. However rapid economic growth and land use
practices has altered the environment landscape of the area. This study was carried out to
examine the rate of loss and pattern of fragmentation of the tropical mountain forests in
Cameron Highlands. Temporal remotely sensed data (SPOT-5 images) from 2000, 2005
and 2010 were used in a GIS to calculate landscape indices. Results showed increases in
the class area (15,384 ha to 15,691 ha), number of patches (499 to 545) and patch density
(1.8 to 2.0 patches/100 ha). The largest patch index increase (34% to 40%) was
associated with the decrease in the area of mean patch (30 ha to 28 ha). The observed
landscape trends indicate slight increase of forest loss and fragmentation, particularly
during the years 2005-2010 periods. Approximately 2 % of the forest cover in Cameron
Highland had been lost in 10 years, and a proportion of the remaining forests had been
degraded as a result of agricultural practices. Combining landscape ecology and remote
sensing has the potential to provide a significant way in assessing the dynamic of
highland landscapes. It is suggested that conservation efforts should be focused on the
management of the natural system and the management of the external influences
particularly restoration and sustainable forest exploitation in the highland.
Keywords – Cameron Highland, land cover change, landscape pattern, remote sensing.
1.
INTRODUCTION
Global environmental change is a result of land cover change (Skole et al., 1997). The
ability to monitor land cover change at a variety of scales provides essential information
required to assist in sustainable land management. In recent years, land management has
1
moved towards a landscape approach which reflects mix of social, environmental and
economic values. In landscape ecology, landscape scale is divided into ecological
processes and human use through developed infrastructure, ownership and management
resources. Our landscape is continuously changing due to both natural and human
disturbances. Landscape changes often occur gradually over time as a series of small,
localized events. The structure and function of a landscape can be perceived differently at
different scales and it is important for the observer to decide upon appropriate scales for a
study (Turner, 1989). The relationship between human behavior and forest change poses
a major research challenge for development projects, policy makers and environmental
organizations that aim to improve forest management (Jane Southworth et al., 2002).
Landscape mapping is often the first step in many remote sensing projects (e.g., Watson
and Wilcock, 2001; Zha and Ni, 2003). Many landscape metrics used in remote sensing
change detection are based on the ecology and these metrics have been developed for
quantifying landscape structure. Landscape metrics is a number or indices that describe
the landscape configuration and composition to formulate and analyse either individual
patches or the whole landscape. Landscape metrics are very important to detect the
pattern of change that is not readily visible to the human eye or easily detectable by
human analyst. The metrics can be used to assess ecosystem health or as variables for
models that support environmental assessment and planning efforts (Herzog et al., 2001,
Patil et al., 2001). These metrics fall into two general categories: those that quantify the
composition of the map without reference to spatial attributes, and those that quantify the
spatial configuration of the map, requiring spatial information for their calculation
(McGarigal and Marks, 1995; Gustafson, 1998). Using satellite imagery such as SPOT or
Landsat, aerial photography, and geographic information systems, landscape ecologists
are able to examine how the landscape has changed over time and how it is likely to
change in the future. Once landscape changes are identified or predicted, the causes and
the ecological and societal consequences of such changes can be examined.
Roy and Joshi (2002) clearly state that changing the landscape pattern through
fragmentation can disrupt ecological processes that depend on movement within the
2
landscape. Tropical mountain forests are among the most fragile and highly threatened of
all tropical forest ecosystems (Bruijnzeel, 2001). Forest landscape models have benefited
greatly from technological advances, including increased computing capacity, the
development of GIS, remote sensing, and software engineering. The forest ecological
processes and their interactions in forest landscape models can be represented by welldesigned computer software (He et al., 2000).
Previous studies reported that Cameron Highlands face various environmental problems
caused by human activities like agriculture, urbanization, infrastructure development and
deforestation which contribute to degradation of the highland landscape and severe
upland soil erosions (Aminuddin et al., 2005; Che Ku Akmar and Mohd Hasmadi, 2010).
To date there are limited studies on landscape pattern or changes in mountain area in
Malaysia. In this paper, the rate of forest loss and pattern of landscape fragmentation in
tropical mountain forest of Cameron Highland was examined by comparing temporal
SPOT-5 images in year 2000, 2005 and 2010. The landscape structure changes were
assessed based on their spatial configuration over time using selected landscape metric or
indices. The information obtained may be directly or indirectly useful to the management
and development strategies for environmental sustainability of the highlands.
2.
METHODOLOGY
2.1 Study area
The study area covers the western region of the Cameron Highlands district, State of
Pahang, Peninsular Malaysia (Figure 1). The area is located between 40 35’ 55.40” N
latitude and 1010 29’ 07.05” E longitude. The study area covers an area of about 27009.8
ha from the total area of the Cameron Highland district (71,225 ha). The elevation in the
study area ranges between 1070 m and 1830 m above mean sea level .The highland has a
steep slope where 66% is more than 200. The mean temperatures about 240C in th
daytime and 14 0C at night. The average annual rainfall is 2660 mm with two peaks in
May and October. Cameron Highlands is drained by three main rivers namely Sg. Telom,
Sg. Bertam and Sg. Lemoi. Two of the main economic activities in Cameron Highlands
are tourism and agriculture.
3
Figure 1: Location and satellite imagery (inset) of Cameron Highlands,
Peninsular Malaysia.
2.2 Land cover classification
Three SPOT-5 imageries with 20 meter raster grid resolution were acquired for the years
2000, 2005 and 2010 respectively. Each image was geometrically, atmospherically and
topographically corrected. The images were analyzed using ERDAS Imagine 9.1, ArcGIS
9.3 software. The land cover in Cameron Highlands were first defined into five classes;
(1)
water
body,(2)
tea
plantation,
(3)
secondary
forest/shrubs,
(4)
mixed
agriculture/residential/road, and (5) primary forest. The classification of was
automatically generated using supervised-maximum likelihood classifier. Supervise
image classification is a method in which the analyst initially defines small areas, called
training sites, on the image which are representative of each desired land cover category
(Kucukmehmetoglu and Geymen, 2008). The classifiers then recognize the spectral
values or signatures associated with these training sites. After the signatures for each land
use/land cover category have been defined, the software then uses these signatures to
4
classify the remaining pixels. Land use/land covers classification in each images were
generated using combined bands of 4, 3 and 1. Using AOI (Area of Interest) the spectral
signature and spectral separability among classes were selected. The land cover classes
then were verified following ground verification or truthing. The classified images were
finally filtered by using 3 X 3 of median statistical filtering approach to reduce pixel
overlaying of minor or isolated classes.
2.3 Accuracy assessment
Accuracy assessment is an essential component of the investigation to quantify whether
the data quality of one classification method is superior over the others (Sader et al,
2005). Accuracy assessment is the process in which the image is partitioned into a set
number of groups (classes) based on the values of the pixels in one or more image
channels. The accuracy was set at more than 80% as the threshold for accurate values. A
total of 255 reference pixels class were selected in a stratified random sampling approach
for each class to assess the accuracy. Accuracy assessment involves identifying a set of
sample locations (ground verification points) that would be visited in the field. Then the
land cover identified in the field was compared to the one mapped in the image during
supervised classification for the same location by means of error or confusion matrices
(Jensen, 1986; Stehman, 1996). Based on the confusion matrices, different accuracy
measures were calculated: producer’s accuracy, user’s accuracy, and overall accuracy. In
order to summarise the classification results, overall accuracies with 95% confidence
intervals were also generated. KAPPA analysis yields (Khat statistic) were also
calculated to measure agreement or accuracy as suggested by Congalton (1991).
2.4 Landscape pattern analysis
There are more than a hundred of indices for quantifying landscape pattern. A group of
selected metrics (or indices) can be useful to interpret the landscape changes and
considered relative to the type of patches (Apan and Peterson, 1998). Landscape pattern
metrics were run on images for year studied. This analysis is designed to generate spatialtemporal indicators of landscape pattern as part of the scale-pattern-process paradigm
(Walsh et al., 1998). Six landscape metrics were chosen in this study. The six landscape
5
metrics are Class Area (CA), Percent of landscape (PLAND), Number of patches (NP),
Patch Density (PD), Mean patch area (AREA_MN) and Largest Patch Index (LPI). The
selected landscape metrics is presented in Table1. The spatial metrics were generated by
FRAGSTATS Version 3.3, a software package that calculates a number of spatial
structures at three levels; the entire landscape, class and patch levels (McGarigal and
Marks, 1995; McGarigal et al., 2002).
Table 1: The selected landscape metrics used in this study
No.
1.
2.
3.
Landscape
metrics
Class Area
Abbreviation
Percent of
landscape
PLAND
Number
patches
of
CA
NP
Unit
Hectares
(ha)
Percentage
(%)
No
Description
The sum of areas of all patches of the
corresponding patch type.
It is equals the percentage of the landscape
comprised of the corresponding patch type.
%LAND = (CA/TLA)* 100
-TLA (Total Landscape Area)
Pi = proportion of the landscape occupied
by patch type (class) i.
aij = area (m2) of patch ij.
A = total landscape area (m2).
The number of patches for each individual class.
The higher NP indicates greater fragmentation.
NP = N
4.
Patch
Density
PD
Number
per
100 hectares
5.
Mean patch
area
AREA_MN
No
6.
Largest
Patch Index
LPI
Percentage
(%)
N = total number of patches in the landscape.
Equals the number of patches in the landscape,
divided by total landscape area (m2), and
multiplied by 10,000 and 100 (to convert to 100
hectares).
N = total number of patches in the landscape.
A = total landscape area (m2).
It equals to means of patch areas (ha)
Equals the area (m2) of the largest patch of the
corresponding patch type divided by total
landscape area (m2), multiplied by 100 (to convert
to a percentage).
6
aij = area (m2) of patch ij.
A = total landscape area (m2).
3.
RESULTS AND DISCUSSION
3.1 Land use land cover change
The trends of land cover change between years 2000, 2005 and 2010 in Cameron
Highlands were varied for each class. Changes in land cover (Figure 2) were derived
from area estimates using land cover maps (Figure 3).
Figure 2: Estimated area of land cover types in years 2000, 2005, and 2010 in the
Cameron Highlands
7
Figure 3: Spatial variation of land use/land cover Cameron Highlands in years 2000,
2005, and 2010.
Comparing the three classification maps in general, the landscape change is not
significant in the study area. Primary forest had increased slightly from year 2000 to year
2005, but decreased after that toward year 2010. This demonstrated the recovery process
of the secondary forest/shrub in Cameron Highland and also in some part of the
abandoned mixed agriculture area. The expansion of tea plantation is expected as shown
in this study. Cameron Highland is the most famous tea producer in Malaysia where more
than 2000 ha of the highland is covered by tea plantation. According to Jamilah et al.,
(2006), the tea plantation landscape is the most preferred scene in Cameron Highlands.
The scenic view of the tea plantation has been known ever since tea was introduced to
Cameron Highlands in 1929. So, this land use has maintained as the main scenic icon of
Cameron Highlands and had increased in aerial extent gradually. It is noted that the other
mixed agriculture/residential/road are showing a decreased trend in the periods of study.
This due to the transformation of this class into other land covers type such as secondary
forest and tea plantation. Meanwhile, the water body showed a minor increase (less than
100 ha) from years 2000 to 2010.
8
3.1 Accuracy assessment
The overall accuracies of the supervised classification of the satellite images accounted
for 94% ( for year 2000), 91% ( for year 2005) and 88% ( for year 2010), respectively
(Table 2). Meanwhile the Kappa statistics were 0.90, 0.80, and 0.79 respectively for the
years 2000, 2005 and 2010 classifications.
Table 2: Accuracy assessment and Kappa statistic for Land cover classification (years
2000, 2005 and 2010)
Image
Accuracy
Assessment
Kappa (K^)
statistic
SPOT-5 Year 2000
94%
0.90
SPOT-5 Year 2005
91%
0.84
SPOT-5 Year 2010
88%
0.79
The KHAT statistic was calculated for determining the statistical significance or
classification accuracy between the years. The computed KHAT statistics is a better
indicator of percentage because the values approached to 1. The overall image accuracies
for each date, following verification from field data, were well above the generally
accepted 85% standard of accuracy for image classifications (Foody, 2003), with Kappa
statistics above 0.75 (Nagendra et al., 2003). The generally accepted overall accuracy
level for land use maps is 85% which is approximately equal in accuracy for most
categories (Jensen, 1986; Mohd Hasmadi and Kamaruzaman, 2008)
3.2 Analysis of landscape patterns
Spatially, changes in landscape of Cameron Highlands are constantly dynamic. The
changes are less than 5% across the study period. Table 3 shows the landscape metric
patterns in years 2000, 2005 and 2010, respectively. The total number of forest fragments
slightly decresed from 1.8474 in 2000 to 1.666 in 2005, and then increased to 2.0177 in
2010. Areas of reforestation were significantly larger than areas of deforestation, across
all dates. Patch size was a good indicator of economic activity.
9
PLAND is increased for tea plantation and water body from 10% to 11% and from 0.12%
to 0.35%, respectively. Meanwhile mixed agriculture, residential and roads had decreased
over the years (23% in 2000 to 20% in 2010). CA and PLAND is a good indicator of
landscape composition. PLAND is an important characteristic in a number of ecological
applications. Both metrics have same characteristic when spatial extent does not change
or are opposite. These indices relate to numbers/total areas for LULC classes. NP
defined total patches for every class and the major patch value in the study is from mixed
secondary forest and shrubs being about 1802. These patches number depends on the
pixel/window size from satellite image resolution. In the study, a low patch numbers is
exhibited by water body class (10 patches only). The highest PD shown in the study was
by secondary forest and shrubs ranging from 4.8 to 6.6. Meanwhile, LPI measures the
largest patch of landscape fragmentation in percentage values and the largest patches
shown in this study was primary forest class (40). The performance of LPI is generally
not sensitive to varying spatial aggregation, thus this index can serve for landscape
composition but not suitable for indication of fragmentation.
Table 3: Landscape structure of the Cameron Highland in 2000, 2005, and 2010
Primary
forest
Secondary
forest and
shrubs
Land cover type
Mixed agriculture,
residential,
roads
Tea
plantations
Water bodies
2000
2005
2010
15384.4
16240.4
15691.8
2185.6
2021.9
2553.5
6468.1
5782.6
5546.8
2939.9
2924.4
3121.5
31.8
40.5
96.2
2000
2005
2010
56.9122
60.2359
58.1235
8.1951
7.281
9.3544
23.8588
21.4083
20.5035
10.9076
10.9299
11.6637
0.1263
0.1449
0.3549
2000
2005
2010
499
450
545
1306
1619
1802
564
530
437
1010
1159
1441
10
40
110
2000
2005
2010
1.8474
1.666
2.0177
4.8351
5.9938
6.6713
2.088
1.9622
1.6179
3.7392
4.2908
5.3349
0.037
0.1481
0.4072
2000
2005
30.8068
36.1563
1.6949
1.2147
11.4264
10.9106
2.9171
2.5473
3.4124
0.9786
Landscape
metric/year
CA
PLAND
NP
PD
AREA_MN
10
2010
28.8069
1.4022
12.6732
2.1863
0.8714
2000
2005
2010
34.4468
38.8871
40.6688
0.3957
0.2341
0.3753
13.5959
4.9977
5.9545
2.0065
1.568
1.7055
0.1003
0.0539
0.1152
LPI
Class Area (CA), Percent of landscape (PLAND), Number of patches (NP), Patch Density (PD), Mean patch area
(AREA_MN) and Largest Patch Index (LPI).
4.
CONCLUSIONS
This research provides evidence to the usefulness of multi-temporal remote sensing
approaches and landscape metric to monitor the rate of loss and pattern of fragmentation
in the tropical mountain forest of Cameron Highlands. It can be concluded that in general,
changes in land use occurred over the study period. Increase in the number of forest
patches was above 2% between year 2000 and 2010 and substantial decrease was shown
in the mixed agriculture/residential/road patch. The latter indicates an improvement in
the forest sucession in Cameron highlands. The positive trends in forest cover changes
and expansion provide some evidence of the ecological sustainability of the area. The
decreasing number of patches by 23% to 21% in mixed agriculture/residential/road
during the study period has, however, raised some questions regarding the agricultural
practices in the highland. This study however, has provided important insights into the
dynamics of land cover in forested area and other major land uses of Cameron Highands.
Quantitatve information form this study reinforces understanding of the relationship
between goverment policies and forest conditions. It is urgent to define political will and
conservation plan that minimises landscpae degradation in the near future in the the
Cameron Highland. This study was only a preliminary steps toward understanding the
landscape and properties of the Cameron Highland environment. Further studies are
needed to improve the scientific knowledge in developing sustainable agriculture and
tourism development strategies in Cameron Highlands, as well as the ability to quantify
the biogeochemical and hydrological processes changes as resulted from landscape
changes in the highlands. The use of multi-temporal remotely sensed data to conduct
landscape trends analysis represents an exciting opportunity not only to conduct change
detection analysis, but also to advance the disciplines of both landscape ecology and
remote sensing in a spatial context.
11
Acknowledgement
We gratefully acknowledge financial support from the Ministry of Higher Education
Malaysia through Fundamental Research Grant Scheme FRGS-5523434 and Malaysian
Remote Sensing Agency for providing satellite data.
REFERENCES
Aminuddin,B.Y., Ghulam,M.H and Wan Abdullah,W.Y. (2005). Sustainability of
Current Agricultural Practices in the Cameron Highlands, Malaysia. Water, Air,
and Soil Pollution: Focus, 5: 89-101.
Apan, A.A. and Peterson, J.A. (1998). Probing tropical deforestation: the use of GIS and
statistical analysis of georeferenced data. Applied Geography, 18: 137-152.
Bruijnzeel, L. A. (2001). Hydrology of tropical montane cloud forests: A reassessment.
Land Use Water Resource. Res., 1: 1-18.
Che Ku Akmar C. K. O., and Mohd Hasmadi I. (2010). Land use in Cameron Highlands:
Analysis of its changes from space. Proceeding of the World Engineering
Congress: Geometrics and Geographical Information Science, 2-5 August 2010,
Grand Margherita Hotel,Kuching, Sarawak, Malaysia, pp: 190-195
Congalton, R.G. (1991). A review of assessing the accuracy of classifications of remotely
sensed data. Remote Sensing of Environment, 37: 35–46.
Foody, G.M. (2003). Remote sensing of tropical forest environments: towards the
monitoring of environmental resources for sustainable development. Int. J.
Remote Sens. 24 (20): 4035–4046.
Gustafson, E.J. (1998). Quantifying landscape spatial pattern: what is the state of the art?
Ecosystems, 1:143–156.
He, H. S., DeZonia, B. E., and Mladenoff, D.J. (2000). An aggregation index (AI) to
quantify spatial patterns of landscapes. Landscape Ecology, 15 (7): 591-601.
Herzog, F., Lausch, A., Muller, E., Thulke, H.H., Steinhardt, U. and Lehmann, S. (2001).
Landscape metrics for assessment of landscape destruction and rehabilitation.
Environmental Management, 27: 91–107.
12
Jamilah, O., Ahmad, M.A., Manohar, M., and Zaliah, S. (2006). Consideration of Visual
Aesthetics Quality for Landscape Management Decisions. Proceeding of the 2nd
Seminar on the Environmental, Centre for Built Environment, Kulliyyah of
Architecture & Environmental Design, IIUM, pp: 206-301.
Jane, S., Harini,, N., and Catherine T. (2002). Fragmentation of a Landscape:
incorporating landscape metrics into satellite analyses of land-cover change.
Landscape Research, 27(3): 253–269.
Jensen J. R. (1986). Introductory digital image processing—A remote sensing perspective
.Englewood, New Jersey: Prentice Hall
Kucukmehmetoglu, M. and Geymen, A., 2008. Measuring the spatial impacts of
urbanization on the surfacewater resource basins in Istanbul via remote sensing.
Environ. Monit. Assess., 142:153-169.
McGarigal, K., and Marks, B. (1995). FRAGSTATS: Spatial pattern analysis program for
quantifying landscape structure, Vol. 2.0. Forest Science Lab, Oregon State
University, Corvalis. Portland, USDA Forest Service, Pacific Northwest Research
Station; General Technical Report PNW-GTR-351.
McGarigal, K., Cushman, S.A., Neel, M.C., Ene, E. (2002). FRAGSTATS: Spatial
pattern analysis program for categorical maps. Computer software program
produced by the authors at the University of Massachusetts, Amherst, Available
at: www.umass.edu/landeco/research/fragstats/fragstats. html. Assessed on 7 July
2009.
Mohd Hasmadi,I. and Kamaruzaman,J.(2008). Satellite data classification accuracy
assessment based from reference dataset. International J. on Computer &
Information Science & Engineering, 2(2): 96-102.
Nagendra, H., Southworth, J., Tucker, C. (2003). Accessibility as a determinant of
landscape transformation in Western Honduras: linking pattern and process.
Landsc. Ecol. 18: 141–158.
Patil, G.P., Brooks, R.P., Myers, W.L., Rapport, D.J. and Taillie, C. (2001). Ecosystem
health and its measurement at landscape scale: Toward the next generation of
quantitative assessments. Ecosystem Health, 7:307–316.
13
Roy, P.S. and Joshi, P.K.(2002) Forest cover assessment in North-East India-the potential
of temporal wide swath satellite sensor data, Int. J. Remote Sensing, 23 : 48814896
Sader, S.A., Ahl, D. And Wen-Shi, L. (1995). Accuracy of Landsat TM and GIS rulebased methods for forest wetland classification in Maine. Remote Sensing of
Environment, 53 : 133-144.
Skole, D., Justice, C., Townshend, J. and Janetos, A. (1997): A land covers change
monitoring program: strategy for an international effort. Mitigation and
Adaptation Strategies for Global Change, 2: 157-75.
Stehman, S. V. (1996). Estimation of Kappa coefficient and its variance using stratified
random sampling. Photogrammetric Engineering and Remote Sensing, 26: 401407.
Turner, M.G. (1989). Landscape ecology: the effect of pattern on process, Annual Review
of Ecological Systems, 20:171–197.
Walsh, S. J., Butler, D. R and Malanson, G. P. (1998). An overview of scale, pattern,
process relationships in geomorphology: a remote sensing and GIS perspective.
Geomorphology, 21 (3&4): 183-205.
Watson, N., and Wilcock, D. (2001). Pre-classification as an aid to the improvement of
thematic and spatial accuracy in land cover maps derived from satellite imagery.
Remote Sensing of Environment, 75 (2): 267-278.
Zha, Y., Goa, J., Ni, S. (2003). Use of normalized difference built-up index in
automatically mapping urban areas from TM imagery. International Journal of
Remote Sensing, 24 (3): 583-594.
14