Mangrove forest zonation by using high resolution satellite data
Methodology
The remotely sensed data used in this study are the CCTS of LANDSAT-5 TM (Scene 128-51) and SPOT-Panchromatic (scene 266-325) acquired on the same date of 20 December 1988 with I hour difference. Aerial infrared photograph at a scale of 1:20,000 which were taken is October 1985 were also used as ground truth for training area selection. The processing of digital performed on Dipt. Aries II at RSD/NRCT.
The SPOT-Panchromatic image was first geometrically registered to UTM topographic maps (scale 1:50,000) using a 1st degree polynomial interpolation and a cubic convolution transformation method. Then, the LANDSAT-TM image was registered to the corrected panchromatic image and resampled from the pixel size of 30 metres to 10 metres using the same transformation method.
The combination of SPOT-Panchromatic and TM 4 and TM 5 images of LANDSAT was selected for this study base on their range of electromagnetic wave-bands which include visible, near infrared and short wave infrared. This should yield better results to both false color composite and automatic
Classification with the fact that the visible region (panchromatic 0.51-0.73 micron) is the chlorophyll absorption band while the near infrared region (TM 4 0.6-0.90 micron) (TM 1.55 - 1.75 micron) responds tot eh humidity condition of soil and vegetation.
Linear enhancement technique applied to the false color composition of panchromatic (blue), TM 5 (green) and TM 4(red) to gain maximum contract of the image. Training area up to 16 classes of land cover in this study area with 7 classes of mangrove forest were selected fro digital classification using maximum forest were selected for digital classification using maximum likelihood classification method.
Results
The false color composite of TM bands 4-53/R/G/B with linear stretch enhancement differentiates mangrove forest from other feature while the FCC of TM 4, TM 5 and SPOT-Panchromatic that was linearly enhanced, show clearly details of both spatial and spectral differences (fig.2). The boundary of each land cover type can be sharply seen. This is particularly true for the linear features such as roads, shrimp dikes, canals etc.
Major classes of land cover can be visually identified. These include 1) mangrove forest which appears in different shades of red depending upon vegetation composition, eg. dense Rhizophora formation in bright red (A) and Melaleuca formation in pale orange (B); 2) aquaculture area which can be clearly identified with its spatial characteristics (C); and 3) open land where light blue color represents paddy fields after harvesting (D) while yellowish green color is moist bare soil (E).
From the digital classification, a total of 16 classes of land cover could be identified within the study area of 454 sq. km. The results are summarized in Table 1 and Fig. 3.
In this study, 7calsses of mangrove forest could be distinguished corresponding to the dominant species as follows :
1) and 2) Rhizophora formation (class I of Vibulsresth et. al., 1985). It is normally distributed along the coast and river banks and could be classified based on its density into 2 classes, namely dense and open.
3) Avicennia formation (class I of Vibulsresth et.al 1985) it forms in narrow strip along the river banks and is dominated by Avicenna ap.
4) and 5) Mixed formation (class 2 and 3 of Vibulsresth et. al 1985). This formation is distributed extensively in the inner zone next Rhizophora formation and could be subdivided into 2 classes as dense and open formations.
Figure 2 False color composite i age of TM 4 (red), TM 5 (Green)
and SPOT-Panchromatic (blue) of mangrove forest at
Khlung, Chanthaburi, on 20 December 1988.
