Printer Friendly Format

Page 2 of 3
| Previous | Next |

Monitoring Mangrove Forests using Remote Sensing and GIS

2.3 Method
The research involved two main steps. In the first step, a broad classification of the general land cover, including mangroves was made (Figure1). This indicated the best approach to dealing with the various types of imagery in order to detect the mangrove deforestation. The second step concentrated on the specific problem of detecting changes in the mangrove areas (Figure2). It examined different approaches for monitoring the nature of the changes in order to produce maps showing the current and former conditions in the study area.



Figure 1. Method
3. Results and Discussions
The classes identified on each image using supervised classification are summarized in Table 2..

Table 2 Classes detected in each image through supervised classification and visual interpretation

Landsat MSS	SPOT	Landsat TM	ERS-1	JERS-1	Radarsat
Broad-leaved Nypa Water Ponds - - agriculture - - - -	Broad-leaved nypa water dry pond orchard mixed agriculture - - - -	broad-leaved nypa water fish pond orchard mixed agriculture clear-cut half-cut swampy -	- nypa water fish pond - mixed - clear cut - - oil pipeline	- nypa water fish pond - mixed - clear-cut half-cut - oil pipeline	- nypa water fish pond - mixed - clear-cut half-cut oil pipeline

Figure 2. Mangrove forest area change detection procedure
Three types of mangrove deforestation were found to occur in the study area:

1. deforestation caused by the establishment of agriculture and/or orchards;
2. deforestation caused by the establishment of oil pipeline networks;
3. deforestation caused by the establishment of shrimp ponds.

Deforestation resulting from the establishment of agriculture crops or orchards can be detected on the Landsat TM image of 1994 and on the SPOT_XS of 1987. Radar data (JERS-1, ERS-1 and Radarsat) cannot, however, detect this type of deforestation. Mangrove deforestation caused by the establishment of oil pipeline networks cannot be detected on any of the optical images. This is because the width of the oil pipelines is around 15m and because water and swampy vegetation surround them. The spectral and spatial resolutions therefore prevent them from being imaged on optical images. However, pipelines are visible on all the radar images because they behave as corner reflectors as a vertically standing object on the ground. Mangrove deforestation caused by establishment of shrimp ponds can be detected on all optical and radar images, although ponds under construction, in which only about half of Nypa palm have been cut, can be detected only on the Landsat TM, JERS-1 and Radarsat images. ERS-1 image was not able to detect the process of deforestation in Nypah palm because of its short wavelength, small incidence angle and it is VV polarization. However, JERS-1 radar image is collected with longer wavelength, HH polarization and medium incidence angle. While Radarsat image is collected using the same short wavelength of ERS-1, but with HH polarization and a medium incidence angle. It is believed that incidence angle (e.g., in this case) played an important role to detect the process of deforestation in the Nypah palm. As Table 2 shows that SPOT_XS image was not able to detect this process of deforestation in the Nypah palm because this image does not cover any areas with this deforestation process from Nypah palm to fish ponds.

The accuracy of each classification was assessed by comparison with data collected at sample points on the ground and compiling error matrices. Based on data collected at sample sites in the field, the classification accuracy of each image to detect deforested mangrove area was as follows:

Landsat MSS	1982	5 classes	76%
SPOT_XS	1987	7 classes	89%
Landsat TM	1994	9 classes	88%
ERS	1996	6 classes	83%
JERS	1996	7 classes	85%
Radarsat	1997	7 classes	84%v