Application of Simulated
Radarsat Sar Data in Land Use/Land Cover Mapping in Cameron Highlands,
Malaysia Laili
Nordin Abstract Department of Fisheries Malaysia Department of Agriculture Jalan Sultan Salahuddin 50632 Kula Lumpur Ku Mohd Noh Ku Ramli 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 : 603-3645646 A pilot project investigation the capability of using simulated Radar set SAR data together with Landsat TM imagery and other relevant data in land use/land cover mapping is being out in Highlands are of Peninsular Malaysia. This paper reports some results of the study that has being carried out in land use/land cover mapping to date. Ground data was collected on land use/land cover at a number of selected sites on high as well as low terrain area of Cameron Highland. Some of this features has been identified Radarsat SAR imagery : on the lower terrain of Cameron highland area, extensively occurring features such as oil palm plantations, rubber plantations, residential area, natural forest, tin mining areas including the ponds, and infrastructures such as north-south highway, main roads and railways line can also be recognized. However, on the high terrain, of Cameron Highlands, land use features that generally occurred as small units of high value vegetables and flowers cultivated in the valleys and tea plantations, could not be identified. It could be discriminated if radar data acquired at a steeper incidence angles. Therefore, more SAR data or digital contour data and further processing are required to test if smaller units of the land use features could be discriminated. Introduction Optical remote sensing data such as Landsat Thematic Mapper and SPOT have been successfully in curtains parts of Malaysia in a number of applications, particularly in land use/ land cover mapping, land use change detection etc. however, there are some limitations; difficulty in acquiring cloud-free imageries in a regular basis: and also difficulty in performing spectral classification for certain types of land use/land cover. Radar remote sensing data has a number of advantages in comparison with optical remote sensing data: the radar data acquisition is not constrained by poor weather conditions . thus, radar data of site on observation can be captured at any time of the year. In additions, radar data differentiate between land use/land cover types with different surface roughness, volumetric backscattering and moisture regime. These parameters are use full in separating land use/land cover types. A pilot project of using simulated radar sat SAR data, is being carried out, covering an area in Cameron Highlands District, with the objective of assessing the capability simulated radarsat data in mapping of land use/land cover. However, this paper only the preliminary findings of land/use cover mapping. Objective The objective of this study is to assess the suitability of simulated Radasat SAR data for land use/land cover mapping in high and low terrain of Cameron Highlands , Malaysia. Study Area The study area covers a portion of Cameron highlands District in the state of Pahang. It lies approximately along the track from a point at 101 30'E to a point at 3 43'N, 101 30'E which covers areas of approximately 2000km square, Appendix 1. the undulating terrain of the area covers a variety of information including forest types, land use for agriculture etc. Materials and Methodology Simulated Radarsat Data Globe SAR data of two polarizations acquired on November 8, 1993 by the Canadian center for Remote sensing (CCRS) using convair -580 equipped with C-band synthetic aperture radar sensor system. The strip of the SAR data, covering a portion of Cameron highland district of 4096 pixels by 22311 lines by 16 bits per pixel size. The images have pixels size of approximately 25m by 28m on the ground. The original SAR data of the area acquired were sent to back to Canada and processed at CCRS in Ottawa for higher quality products. The processing commented were antenna pattern correction, slant range to ground range correction, slant range to ground range correction, 3 by 3 Lee filtering, and enhancement. The produced was simulated radarsat SAR data of 1 bites per, pixel, multi file format (MFF). Single band, 1734 pixel by 7678 lines samples , and standard simulation mode. The simulated radarsat SAR data was registered to geocoded Landsat TM for further processing, Appendix 2. Firstly, the SAR imagery was enhanced and printed hardcopy for visual interpretation. Landsat Thematic Mapper Landsat TM data of path 127 row 57 acquired on 26th February 1993 available at the Departent of Agriculture, Peninsular Malaysia was used in this project. The geo-coding to RSO coordinate of the scene was done the Malaysia center of Remoter sensing ( MACRES). Subsequently, one visible ( band 3) and two infra -red band ( band 4 and band 5) of which sensitive to plant species, vegetation biomass and amount of water in plants were used. The three bands (RGB) were transformed into intensity, hue, and saturation (HIS) and following that intensity was replaced with radar imagery. Finally, the radar band, hue, and saturation were converted to RGB for the purposed of hard copy printing for knowledge based visual interpretation, Appendix 2. Land Use /land cover Vector Data 1990. Land use/land cover map of Cameron Highlands of scale 1:50,000 was used as a reference in this study. The map was complied by the Department of Agriculture based on aerial photography interpretation of scale 1:40,000 of 1980. subsequently, the map was updated based on visual interpretation of Landsat TM 1990. the map was digitized by using software ARC/INFO. Topography Base Map Topography based map of 1:50,000 scale prepared the department of survey and mapping Malaysia was used to rectified the Landsat TM imagery to RSO co-ordinate. Consequently, the SAR imagery was geocoded to Landsat TM imagery ( image to image). Field survey data The first field trip was done on November 12, 1993, after three days Globe SAR data acquistion ( November 8, 1993), to collect ground data such as land use type, soil color, soil moisture, rainfall ( station I second field trip, after the remote sensing data processing, was carried out on 19th of August, for the purpose of verifying visual interpretation of the hard copy of simulated Radarsat SAR and a combination of Radarsat SAR & Landsat TM image map printed of scale 1:50,000. Product of image processing The steps of processing the simulated SAR and Landsat TM data were shown in appendix 2. the software used were Ergovista ( for SAR) and Erdas imagine 8.2 ( for Landsat TM ) to produce hard copies as follows:
Initial interpretation of the simulated Radarsat SAR data and color composite of simulated Radarsat SAR and Landsat TM was based on hard copies. Oil palm and rubber plantations were the major land use in the southern portion of Cameron Highland. They were comprises of various stages of growth. On SAR imagery, mature oil palm can be discriminated form matured rubber plantation based on difference degree of backscattering. Mature oil palm plantation demonstrated higher backscattering . mature oil palm plantation demonstrated higher backscattering than mature rubber plantation of which due to 'rougher' texture of oil palm canopy. While the nature rubber plantation has lower backscattering than mature oil palm plantation; it could be as a result of less surface' roughness ' of rubber canopy compared to oil palm. However, young oil palm ( planting distant of 30 feet by 30 feet ) and young rubber ( planting distant 10 feet by 8 feet ) plantation could not be discriminated on SAR as well as on color composite of SAR and Landsat TM imageries as the backscattering was mainly form cover crops such as leguminase family planted before young oil palm and young rubber were transplanted to the plantation, for the purpose of biologically control of weed growth and soil improvement. Forest and tin mining area are the major land cover types covering by the SAR data on the southern portion of Cameron highlands. On the SAR imagery , forest appeared darker tone than the rest of land use/land cover. So it could be discriminated easily. In addition, forest can be recognized clearly because it has a ' rougher ' texture and most of them are found on high terrain areas . on the other and, tin mining can be identified on the SAR imagery due to its physical characteristics, i.e it has ponds and sand ( tin-tilling ). The ponds, accrued sporadically in the tin mining areas appeared darker within the bright areas of sand. The infrastructures such as roads, power lines , railway, can be identified on the SAR imagery because they passed in this area through forest. Oil palm plantations, rubber plantations, which results in a significant height differences of SAR reflecting canopy at the infrastructures location. On the high terrain of camberon highlands , major land use/land cover area small patches of settlements, vegetables and flower farms, and secondary forest and primary forest. Forest on the high terrain was identifiable on the imagery. However, the settlements , flower and vegetables farms in the valleys can not be recognized on the imageries. It could be due to the SAR data captured at wide incidence angle. Summary In the study of application of simulated Radarsat SAR data in land use/land cover mapping ahs shown some results. On the low terrain of Cameron highland, mature oil palm plantations and mature rubber plantations, tin mining areas, natural forest, infrastructure can be discriminated on hard copy of the SAR imagery. However, on the high terrain of Cameron highlands, flower and vegetable farms ( found in the valleys, except natural forest), can not be identified. It could be due to the SAR data captured at wide incidence angel, Perhaps, the flower and vegetable from could be discriminated if the SAR data captured at the steeper incidence or processed with DEM for further investigation. |