|
|
|
Forestry
|
Separation (Recognition) of tree species or species composition in an old growth forest
Plantation in Peninsular Malaysis using the Vegetation Index approach
Method
-
Study Site
The study site covers the are of old growth forest plantation of the Forest Research Institute Malaysia (FRIM) situated within the Bukit Lagong Forest Reserve. The forest reserve is situated 11 km north-west of Kuala Lumpur within the latitude of 3013' to 3015'N and longitude of 101036' and 101039' E (Figure 1). The plantation area was established as early as 1927. The plantation area was divided into compartments in which various species wee planted; some compartments with mono-species while others were mixed. Thus, throughout the study site species composition is varied. However, the species are not widely and randomly distributed. They form rather clumpy patches, some with a very distinct boundary. In terms of vertical structure of the forest, vegetation in this plot has reach the 'near climax' situation n which it has a physical appearance of a primary forest with some of the older trees having reached a height of more than 30 m.
Historical planting records were available for the whole plantation area. Using these records and the geocoded vector map of the area, a thematic map of species distribution with the result of VI.
-
TM data and preprocessing
Two sets of TM data of 1988 and 1993 wee used in this study. There was no technical reason why the 2 images were chosen. These wee the only available daa which were free of cloud cover and had no apparent haze effect. Prior to analysis, the image were geometrically corrected. The control points were selected from common points recognizable on the images and topographic map. Seven points wee used for the 1988 image and 13 points for the 1993 image. The image were corrected using 2nd and 3rd degree polynomials for the 1988 (RMS error = 0.37 pixels) and 1993 (RMS error = 1.03 pixels) images respectively. The pixels were resampled by the nearest neighbour method to maintain their original data. The vector map was then reistered to the images. Masking was done so as to include only the forested areas. Neither radiometric correction nor correction for terrain effects was carried out for both images. Theoretically, VI will normalize radiometric and terrain effects.
-
Vegetation Indices
For this preliminary study, 6 VIs were used for each image following Malingreau (1989). A detailed discussion on all the 6 indices are presented in the said paper. The 6 are:
-
Normalised Different Vegetation Index (NDVI)
| NDVI = |
TM4-TM3
--------------
TM4-TM3 |
- Transform Vegetation Index (TVI)
- Ratio Vegetation Index (RVI)
- Simple subtraction Vegetation Index (SVI)
SVI = TM4-TM3
- F1 nad f2 (Malingreau) Indices
| F1 = |
TM2
--------------
TM4 x TM3 |
| F2 = |
TM4xTM3
--------------
TM2 |
VI values for the above mentioned indices were calculated for the two images. The resultant indices of each of the 6 VIs for both images were then overlaid onto the ground trouth map. The species that showed high correlation to ground trouth data were recognized and their 'indices or indices range' was identified. The resultant VI images were reclassified based on the specific 'indices or indices range' for the 6 species recognized. Confusion matrix was used to evaluate the accuracy of VI values to map the 6 species.
|
|
|
|
|
|
|
