Characterizing Tropical Forest in Kinabalu Park, Sabah, Malaysia, Using Landsat-TM Data
Phua Mui How1, Hideki Saito2 and Masa Aniya3
Forest application of remote sensing deals mainly what the canopy characteristics of forest. The advances in remote sensing technology provide and alternative that may be supplementary to the conventional methods, the field survey and the aerial photographs interpretation, for forest information extraction and forest applications at a regional scale. At a regional scale, satellite data can be a very useful source for forest resources mapping and inventorying with the reasonable costs and accuracy. Information on forest resources at a regional scale. The objectives of this study are to examine the characteristics of the tropical forest using Landsat-TM data, and to investigate the relationship between forest variables of the tropical forest and Landsat-TM data, and to investigate the relationships between forest variables of the tropical forest and Landsat-TM and the tasseled cap transformed data.
Kinabablu Park is located in the northern part of the state of Sabah, Malaysia, between 116° 28′ and 116° 45’E and between 6° and 6° 30’N. Having an area of 75,370 h, it is the focus of the tourism activates in the state especially the southwestern part of the park where the highest mountain in Southeast Asia, Mount kinabalu (4096m) is located.
The vegetation in Kinabalu park was delineated inn to 6 discrete altitudinal floristic zones; Lowland (<1200 m), Lower montane (1200-2350m), Upper Montane 92350-2800m), Lower subalpine (2800-3400m), Upper subalpine (3400-3700m) and Alpine (>3700m), based on their great variations in the deominance type, species composition and forest structure (kitayama,1991). Kinabalu’s vegetation zonation is comparable to many other mountains in the region of Malesia such as the Vegetation zonation of Mount Wilhelm in Papua New Guinea. Studying vegetation in Mount Kinabalu can therefore be useful to vegetation studies in othe5r tropical mountains in the resin of Malaysia.
A Landsat-TM image taken on 8th April 1996 was used. Examination of atmospheric effects on the image was carried out by regression analyses between elevation and the remotely sensed data (Itten et al., 1992), and by histogram method (chavez, 1988). Bands 1(0.45-0.52mm), 2 (0.52-0.60mm) and 6(10.40 – 12.50mm) were strongly related to the elevation. Thus, in this study, emphasis is given on bands 3 (0.63-0.69mm), and infrared bands of 4 (0.76-0.90mm), 5 (1.55-1.75mm) and 7 (2.08-2.35mm), which did not show apparent elevation patterns.
Aerial photographs were the principal source for extracting the variables of the tropical forest; crown diameter and percentage crown cover, as well as recognition of forest types. Field surveys were carried out to investigate the forest structures under the canopy layer and to obtain the ground truth data for land cover types in general, for forest types in specific.
Contour lines were digitized for topographic map (1:50,000) and input in to a Geographic Information system as vector data. A digital elevation model ea produced from the vector deal (elevation ) for geometric correction s of though satellite image. Linear distortions of the image due to relief displacement effect were corrected, and the satellite image was transformed to the Universal Transverse Mercator projection. RMS error of the transformation before the correction was 4.56 pixels compared to 2.55 pixels after the correction.
The Landsat-TM image was transformed using tasseled cap transformation after the geometric corrections. The tasseled cap transformation generates spectral features (brightness, greenness and wetness ) that are well known for their relationships with physical characteristics in a satellite image (Crist and kauth, 1986). It was applied in this study because the spectral features can be directly related to the forest variables.
Instead of applying an inappropriate radiometric correction, classification is desired to assure relative homogenous illumination. In this case, since the sum azimuth during the satellite overpass (86.6°) was almost identical to the East (90°), slope-aspect of the image can be classified into four slope-aspect classes; east (42°-131°), south (132°-221°), west (222°-311°) and north (312°-41°). Slope -aspect with in the definition of east slope plots were identified in the satellite image and the mean digital number (DN) was extracted from each plot located in the east slope-aspect class for analysis.
The topographic effect within the east slope-aspect class was assessed by regressing cosine of incident angle (topographic parameter ) with the Landsat-TM data. There is no evidence that the remotely sensed data used in this study (east slope-aspect ) were influenced by the topography effect.