cylindrica), Keora (Sonneratia apetala), Baen (Avicennia
officinalis), Dhundal (Xylocarpus granatum), Goran (Ceriops
decandra) and Golpata (Nypa fruticans).
Time-series Landsat data from 1972-2010 were used in this
study. The list of data-sets is enumerated in Table 1.
Table 1. Landsat scenes used in the study
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B8, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
Time- | Data-sets Path- Date Use
frame row
1972- | Landsat 148044 | 19.12.1972 | Image
73 Multispectral classification
Scanner 147045 | 20.02.1973 | and change
(MSS) analysis
148045 | 21.02.1973
Time-series
change
analysis
along the
lines
1980 | Landsat 147045 | 15.01.1980 | Time-series
MSS change
148045 | 21.02.1980 | analysis
along the
1989 | Landsat 137045 | 12.01.1989 | lines
Thematic
Mapper
(TM)
2000 | Landsat 137045 | 28.02.2000
Enhanced
Thematic
Mapper Plus
(ETM+)
2010 | Landsat TM | 137045 | 30.01.2010 | Image
classification
138045 | 06.02.2010 | and change
analysis
Time-series
change
analysis
along the
lines
2.2 Methodology
Orthorectified Landsat images were downloaded from United
States Geological Survey (USGS). Landsat images of 1972-73
and 2010 were classified based on the spectral subclasses
identified on the digital images. Those classes were later
merged to form desired land cover categories (Table 2).
Maximum likelihood or minimum distance algorithm was
applied in the classification procedure. Mangrove forest change
map was prepared by overlaying classified images of 1970s and
2010. The change map is the basis for computation of change
statistics and generation of change matrix.
Time-series change analysis was done along the selected
transect lines. Five transect lines were selected each for the
eastern and western parts of Sundarbans located in Bangladesh.
Eastern part is located at the east of Kunga River and western
part is located at the west of the river. The lines were selected
randomly from the eroded places noticed on 1973-2010 change
image. The analysis was done in a decadal time-scale. The
trends of erosion are investigated by digitizing on-screen the
visible forest boundary on the digital Landsat scenes.
Table 2. Land cover definitions used in image classification
Land cover Definition
categories
Mangrove The class represents natural mangrove
forest that excludes coastal plantation,
Other The class consists of other vegetation
vegetation and includes homestead forest
vegetation, crops, coastal mangrove
plantation.
Bare land This is a contemporary class. Most of
these areas belong to seasonal cropland.
The class includes sandy soil in the
offshore islands.
Water The class represents all type of water-
bodies including sea, river, shrimp
farms, inland water bodies, agricultural
land flooded with water.
3. RESULTS AND DISCUSSION
3.1 Changes in 1973-2010
The overall changes in the Bangladesh part of Sundarbans are
presented in Figure 1. Red color shows mangrove forest loss
due to coastal erosion during the last four decades (1973-2010).
Erosion is vivid in most of the peripheral zone of mangrove
forest.
The change statistics reveals (Table 3) that the mangrove
forests have been lost about 144 sq. km during the period of
investigation. The loss of mangrove because of erosion was
233 sq. km, while conversion of sea-water to mangrove forest
was 104 sq. km. This is alarming since the loss of forest due to
land erosion is higher than the forest aggradations. Land
accretion and forest succession are also noticed in some places.
But this phenomenon is usually concentrated only in some
spots, not widely spread as coastal erosion
3.2 Time-Series Change Analysis
Time-series changes for the selected transect lines of
Sundarbans forest have been analyzed (Figure 2 and 3) for
eastern and western part of Bangladesh Sundarbans.
Photographs of the eroded places obtained during the field-
mission of January 2012 are presented in Figure 4. The nature
and patterns of coastal erosion responsible for forest loss are
vivid in the pictures.
The rate (m/year) of erosion at the different points of the
selected transect lines is presented in Table 4. The erosion rates
are variable and it is very difficult to get a conclusive result
from the analysis of those points whether the erosion has been
accelerated in the recent past due to man-made or climate
induced factors. The average rates of erosion for the eastern
and western parts are 14 m/year and 15 m/year respectively
obtained form the ten selected transect lines. Further studies
should examine it using different methods.
