| Geoscience and Environment |
Land Use in Bayan Lepas 1969Land use in Bayan Lepas 2003The Land-use Map 2003Image Processing and Potential RefinementsGo to Bayan Lepas Menu |
|
Land use in Bayan Lepas 1969The land use map shown in Figure 7, below, was produced by analyzing a topographic map for Penang Island, (1977). The map covers an area of 67.4 km2, of which 42.9 km2 (63.6%) was land, 1.8 km2 (2.7%) was marshland, and 22.7 km2 (33.7%) was sea, including the intertidal zone (coastal sand and mudflats), 2.8 km2. The land area was classified as follows: rice fields 5.6 km2 (13.1%); airport 1.5 km2 (3.5%); and other land, 35.8 km2 (83.4%). The area of buildings and associated infrastructure could not be measured directly because such point and line features lack the dimensions needed for measuring area. The residential and commercial areas, including infrastructure but excluding the airport, was estimated at about 0.75 km 2, (See Buildings for the calculation.). The large drains are included in the rice-fields category. The land area was less than in 2003 because in the interval land was reclaimed from the marshes and the sea. (See: Tables showing land-use classes and Techniques for making 1969 land-use map.) |
|
Of total land area and marshes (44.7 km2), 16.9 km2 (37.8%) was below 10 meters, 9.9 km2 (22.1%) was between 10 and 30 meters, and 17.9 km2 (40.0%) above 30 meters. These elevations are shown as overlays on the map. Of the sea area (22.7 km2), the intertidal zone was 2.8 km2 (12.3%). During the period 1970 to 2003, 4.7 km2 of land was reclaimed, including all of the intertidal zone and all of the marshland. Of rice fields, 85% were below 10 meters and the rest were between 10 and 30 meters. (One hectare of rice field appeared to be located over 30 meters, but the elevation was judged to be spurious, resulting from slight misregistration of the images.) Of buildings, 58% were estimated to be located below 10 meters, 38% between 10 and 30 meters, and 4% over 30 meters. The features called "drains" in the land-use map may be more significant than this term implies. See discussion. |
|
Land use in Bayan Lepas 2003This photograph (Figure 8) was taken looking east from the parking lot of a new golf and country club located at G in Figure 9. The golf course lies in the foreground just east of the letter G in Figure 9, where the land use code indicates grassland. In the middle ground on the left of Figure 8, high-rise apartment blocks are flanked on the left by older low-rise-low-income housing areas and on the right newer high-income apartments. Behind the tallest apartment block, buildings can be seen on the top of Bukit Gedong, a hill reserved by the military. The lower eastern slopes behind the hill are sparsely covered by woods though which a road passes to join the main north-south road. Between these converging roads lies a triangular area occupied by factories. Beyond the easternmost of these two roads, the main north-south road, lies the runway of the international airport. Beyond the runway the main area of factories can be seen stretching to the coast and extending north behind Bukit Gedong and as far south as the lower levels of the southern hills. The hill visible beyond Bukit Gedong is an offshore island part of which extends south of Bukit Gedong. This 120-degree panorama image was produced by stitching together seven photographs taken with a hand-held digital camera using the PanaVue ImageAssembler. Demo version available. |
|
|
|
Land-use Map 2003The map in Figure 9 shows the sea, 19.8 km2 (29%), and five kinds of land use and land cover, 47.6 km2 (71%). The total land area is comprised of: buildings and bare soil, 21.0 km2 (44%); grassland, 7.4 km2 (16%); forest, 6.7 km2 (14%); brushland, 3.4 km2 (7%); and other, 9.0 km2 (19%). (See: Tables showing land-use classes.) The land-class "other" was the most difficult to identify. Some locations indicated as "other" were not accessible, such as the north end of the main runway. Both golf courses (G and H) have small ponds that appear to fit, leading to a working hypothesis that "other" refers to water bodies. The ASTER dataset was acquired on November 4, 2003, a time when rainfall was high and declining. (See: Rainfall). However, the field visit took place July 18, 2004, when rainfall was lower, thus field observations were of limited use in determining the type of land use. Inspection of a false-color composite image containing an infrared layer reveals some of the features as surface water bodies, "drains", rivers, and ponds. The reason these pixels are so dark in the ASTER 243 composite is that infrared absorbs light reflected from water, thus pixels representing wet areas appear dark. This characteristic of the infrared makes it useful for anaylysis. (See: 243 composite). Reclaimed land is shown in Figure 9, but it cannot be seen as such in the ASTER images. The area of reclaimed land was defined by comparing the ASTER images with the topographic map based on the 1969 survey. The industrial area outlined by red dots was determined from the 2-4-3 composite image, "looking over the fence", and by inspection of photographs, (1), (2), and, (3). |
|
Image Processing and Potential Refinements
Image ProcessingThe land-use map in Figure 9 was made by using all of the 14 ASTER bands as follows: visible and near-infrared bands (VNIR), images 1 to 3, with 15 meters per pixel resolution; infrared bands (SWIR), images 4 to 9, with 30 meters resolution; and thermal infrared bands (TIR), images 10 to 14, with 90 meters resolution. (See: ASTER characteristics).Some of the images have noise in the form of stripes across the image. This noise was removed using principal components analysis (PCA). The last component or components (PC's) extracted contained most of the noise. These were discarded and the images reassembled from the higher order components containing most of the information but lacking noise. Finally, to classify all pixels in the scene, all three VNIR images (green, red, and near-infrared) were used together with three PC's from the six SWIR bands and two PC's from the five TIR bands, a total of eight images. A computer routine classified similar locations by cluster analysis of the pixels in the eight input images, producing a new image with each location (pixel) represented by a code number shown as a land use category in Figure 9. Potential RefinementsIntuitively, the more information used the more accurate land-use classification should be. If followed, this rule of thumb could lead to less refined classification, because the low resolution images could degrade the information in the higher resolution images. While beyond the scope of the current project, experiments with other processing strategies might yield more refined classification maps.For example, principal components analysis could be applied to the eight images used as input to the cluster routine as described above. Since the extracted principal components would be uncorrelated (orthogonal), there should be no degradation of one band by another. A second approach would be to pre-process the bands before classifacation, using for example the NDVI index. (See:NDVI). This index shows variations in green biomass and chlorophyll and is therefore capable of discriminating between vegetation, soil, and building materials. Thus the NDVI index, uses the highest resolution ASTER images and, in combination with other bands, might provide a more refined land-use map. A third approach would be to pre-process the bands by applying a statistical transformation so that the transformed data conforms to a normal distribution. In theory, this approach would result in more reliable cluster analysis. Conceptually, several processing strategies have potential for refining the process of land use classification, but exploration of their potential is beyond the scope of the present study.
|
| The URL of this site is [www.geoscience2000.info/index.html] |