ISPRS Journal of Photogrammetry and Remote Sensing 101 (2015) 137–144

Contents lists available at ScienceDirect

ISPRS Journal of Photogrammetry and Remote Sensing

journal homepage: www.elsevier.com/locate/isprsjprs

Monitoring the coastline change of Hatiya Island in Bangladesh using

remote sensing techniques
Manoj Kumer Ghosh a,⇑, Lalit Kumar b,1, Chandan Roy a,2
a
Geography and Environmental Studies, University of Rajshahi, Rajshahi 6205, Bangladesh
b
GIS & Remote Sensing, School of Environmental and Rural Science, University of New England, Armidale, NSW 2351, Australia

a r t i c l e i n f o a b s t r a c t

Article history: A large percentage of the world’s population is concentrated along the coastal zones. These environmen-
Received 31 July 2014 tally sensitive areas are under intense pressure from natural processes such as erosion, accretion and
Received in revised form 5 December 2014 natural disasters as well as anthropogenic processes such as urban growth, resource development and
Accepted 6 December 2014
pollution. These threats have made the coastal zone a priority for coastline monitoring programs and
sustainable coastal management. This research utilizes integrated techniques of remote sensing and
geographic information system (GIS) to monitor coastline changes from 1989 to 2010 at Hatiya Island,
Keywords:
Bangladesh. In this study, satellite images from Thematic Mapper (TM) and Enhanced Thematic Mapper
Integrated technique
Landsat
(ETM) were used to quantify the spatio-temporal changes that took place in the coastal zone of Hatiya
MNDWI Island during the specified period. The modified normalized difference water index (MNDWI) algorithm
Accretion was applied to TM (1989 and 2010) and ETM (2000) images to discriminate the land–water interface and
Erosion the on-screen digitizing approach was used over the MNDWI images of 1989, 2000 and 2010 for coastline
On-screen digitizing extraction. Afterwards, the extent of changes in the coastline was estimated through overlaying the dig-
itized maps of Hatiya Island of all three years. Coastline positions were highlighted to infer the erosion/
accretion sectors along the coast, and the coastline changes were calculated. The results showed that ero-
sion was severe in the northern and western parts of the island, whereas the southern and eastern parts
of the island gained land through sedimentation. Over the study period (1989–2010), this offshore island
witnessed the erosion of 6476 hectares. In contrast it experienced an accretion of 9916 hectares. These
erosion and accretion processes played an active role in the changes of coastline during the study period.
Ó 2014 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier
B.V. All rights reserved.

1. Introduction (Zhang, 2011). Effective planning and management are the precon-
ditions for sustainable coastal development (Kumar and Ghosh,
Coastal zones are now being gradually recognized as functional 2012). Contribution of the spatio-temporal monitoring of coastal
regions facing intensified natural and anthropogenic disturbances environmental changes is multifold. Monitoring can help to under-
including sea level rise, coastal erosion and sedimentation, and stand the spatial distribution of erosion hazards, predict their
over-exploitation of resources. Almost 70% of the world’s beaches development trends and support research on coastal erosion and
are experiencing coastal erosion. At many coastal regions the rate its counter measures (Jayson-Quashigah et al., 2013).
of erosion is alarming and is considered a serious hazard (Addo Coastline is defined as the line of contact between land and the
et al., 2008). Awareness of the quality of global coastal ecosystems water body at one instant in time (Gens, 2010; Naji and Tawfeeq,
has increased over the recent years. This awareness of coastlines 2011). It is one of the most important linear features on the earth’s
being adversely impacted by multiple driving forces has acceler- surface, which displays a dynamic nature and is an indicator for
ated our efforts to assess, monitor and mitigate coastal stressors coastal erosion and accretion. It could be regarded as the most
unique feature on the earth’s surface. The location and attributes
of coastlines are highly valued by a diverse user community, as
⇑ Corresponding author. Tel.: +880 1711080539. they have never been stable in either their long-term or short-term
E-mail addresses: manojkumer@gmail.com (M.K. Ghosh), lkumar@une.edu.au positions. Human life, cultivation and natural resources are
(L. Kumar), chandanroy@gmail.com (C. Roy).
1
affected by the processes of erosion and accretion along the coast,
Tel.: + 61 2 67735239; fax: +61 2 67732769.
2
Tel.: +880 1720690758.
and rapid coastline changes can create catastrophic social and

http://dx.doi.org/10.1016/j.isprsjprs.2014.12.009
0924-2716/Ó 2014 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved.
138 M.K. Ghosh et al. / ISPRS Journal of Photogrammetry and Remote Sensing 101 (2015) 137–144

economic effect along these regions. An in-depth understanding of by erosion and deposition of sediments using time series data from
regional coastline dynamics (erosion and accretion) is required and 1989 to 2010 with the help of established remote sensing tech-
necessary for the design of viable land use and protection strate- niques. As mentioned earlier, there are many advanced change
gies to reduce potential loss (Blodget et al., 1991; Chu et al., detection techniques available. The approach used in this study
2006; Jayson-Quashigah et al., 2013; Naji and Tawfeeq, 2011). to identify coastline and detect changes was based on the combi-
Coastal erosion is the process of removal of material at the shore- nation of water indices approach and on-screen digitization. The
line which leads to loss of land as the shoreline retreats landward modified normalized difference water index (MNDWI) (Xu, 2006)
whereas accretion is the product of deposition of material at the was used for discriminating the land–water interface and on
shoreline which leads to gain of land as the coast advances sea- screen digitization was used for extracting the island boundary
ward (Gibb, 1978). Due to the importance of the processes that from the MNDWI images. In this research Hatiya, an offshore island
occur along the coastline, rapid and reliable techniques are of coastal regions of Bangladesh, was selected to determine the
required to monitor and update coastline maps. It is also necessary trend and rate of this change.
to quantify the rates of environmental retreats.
Aerial photo and ground survey techniques are conventionally
2. Methods
used for coastline monitoring. However, these techniques are
expensive and time consuming and require trained staff. Recently,
2.1. Study area
remote sensing and geographical information systems (GIS) have
been widely used to complement the conventional method for
The present study was conducted in Hatiya Island of the Noak-
monitoring coastline change (Ryu et al., 2002; Yamano et al.,
hali district of Bangladesh. This island is located between 22°030 N
2006). Remote sensing provides the capability to monitor the
to 22°250 N latitude and 90°580 E to 91°120 E longitudes (Fig. 1). The
coastline in a cost effective manner. Furthermore many advanced
island is surrounded by the Meghna River and Noakhali mainland
techniques have been developed using digital images for change
in the north, the Bay of Bengal on the south, Shahabajpur River,
detection. Several change detection techniques using different spa-
Monpura and Tajumuddin Thana (village) of the Bhola district to
tial and temporal resolutions of satellite imagery have been
the west and Meghna River and Sandwip Thana of the Chittagong
reported in the literature (Coppin et al., 2004; Zimmermann and
district in the east. Hatiya Island occupies the young Meghna estu-
Bijker, 2004). The most commonly used are band ratio, band differ-
arine floodplain sub-unit. The topography of the island is almost
encing, principal component analysis and post-classification. On-
flat. The average elevation of the island is approximately 2.4 m
screen digitizing techniques have also been used in various other
above the mean sea-level. This area is mainly divided into three
studies for spatial change mapping (Darvishzadeh, 2000; Harvey
physiographic regions: the Lower Meghna estuarine floodplain,
and Hill, 2001; El-Asmar and Hereher, 2011). For example this
the mixed tidal flood plain, and mud and others (Canal, River, Mar-
technique has been applied to TM satellite images for mapping
shy land). Most of Hatiya Island consists of quaternary alluvial
spatial changes in sand dune locations in the Western Desert of
deposits laid down by the Ganges, Brahmaputra and Meghna Riv-
Egypt between 1987 and 2000 (Hereher, 2010). To detect changes
ers and deposited under tidal conditions. According to the soil sur-
in water bodies, it is necessary to pre-process the satellite image in
vey report, the deposited sediments are predominantly silt and are
such a way that water and land appear as contrasting features in
medium to moderately fine textured. The tectonic structure of Hat-
the satellite image. Water indices are applied to the images to dif-
iya Island is closely related to the Bengal fore deep. The Bengal fore
ferentiate land from the water bodies so that on-screen digitizing
deep is a Geosynclinals area, which is an area of both sedimenta-
can be performed easily and accurately. Water indices are mathe-
tion and subsidence.
matical models that enhance the water signals for a given pixel on
As a small island with an area of 46,682 hectares (Dwip
images obtained from the visible/near-infrared portion of the elec-
Unnayan Report, 2008), Hatiya is characterized by limited
tromagnetic spectrum. Two bands are used (usually from the visi-
resources. The economy of this island is based on agriculture and
ble and near-infrared portions of the spectrum) to calculate these
fishing. Hatiya Island is one of many offshore islands on the Ban-
indices (El-Asmar and Hereher, 2011).
gladesh coast and is representative of the shifts in coastline that
Hatiya Island is one of the offshore islands of Bangladesh that
have been occurring due to human activity, river discharge, tides,
has been subjected to severe erosion and accretion (re-deposition
coastal hydrology of the region and the series of floods due to
of sediment) processes. The island has shown significant morpho-
the three mighty river systems: the Ganges (Padma), Brahmaputra
logical changes through the years due to severe erosion hazards
(Jamuna) and the Meghna. In this article Hatiya Island has been
and strong accretion prospects. The coastline of Hatiya Island has
selected as the study area because the island is very densely pop-
changed dramatically during the last three decades due to these
ulated and has been experiencing rapid coastal changes due to the
processes (Kumar and Ghosh, 2012).
dynamics of major river system in the coastal belt of Bangladesh,
For illustrating coastal dynamics, remote sensing has proved its
and this dynamism has resulted into severe vulnerability of the
effectiveness in providing accurate information, but unfortunately,
island, such as erosion and accretion.
no such application has been undertaken for Hatiya Island for
coastline change detection. It is necessary to mention that the
island does not have any official statistics on coastline change pat- 2.2. Image used
terns, which are very important for sustainable coastal manage-
ment of the island. Sustainable development cannot be achieved In this study, satellite images of three time steps were used to
and may lead to the mismanagement of scarce resources in the quantify the changes along the coastline of Hatiya Island: two from
absence of such information. Space-borne remotely sensed data the Landsat Thematic Mapper (TM) (28.5 m spatial resolution)
are considered to be important for coastal islands like Hatiya acquired in December 1989 and December 2010 and one from
because they are not only required to comprehend the past and the Enhanced Thematic Mapper (ETM) (28.5 m spatial resolution)
present conditions of the coastline but also used to achieve sus- acquired in December 2000. The TM and ETM have seven spectral
tainable development and to drive sound environmental planning bands that cover the visible, near infrared, short-wave infrared and
(Dewan and Yamaguchi, 2009; Kumar and Ghosh, 2012). thermal infrared regions of the electromagnetic spectrum: blue
In this study an attempt has been made to map the coastline of (0.45–0.52 lm), green (0.52–0.60 lm), red (0.63–0.69 lm), NIR
Hatiya Island, Bangladesh, and to identify coastline changes caused (0.76–0.90 lm), mid-infrared, MIR (1.55–1.75 lm), thermal
 M.K. Ghosh et al. / ISPRS Journal of Photogrammetry and Remote Sensing 101 (2015) 137–144 139

Fig. 1. (a) Location of the study area and (b) provinces of Hatiya Island.

infrared, TIR (10.4–12.5 lm) and MIR (2.08–2.35 lm) (NASA, 2011; channels. The images were re-sampled to a 30 m pixel size using
USGS, 2013). All the images used in this study were taken in the the nearest neighbor resampling method. A first-order polynomial
dry season so that the coastline could be easily identified, as the transform algorithm was used to ensure that each permanent
water level does not remain very high in the dry season. feature was at exactly the same location in all the images.

2.3. Image pre-processing 2.4. Change detection

To remove the influence of the atmosphere and topography on Two different approaches were applied to identify the coastline
the remotely sensed data, geometric and radiometric corrections and detect changes: (1) land and water interface discrimination for
were carried out. The images were corrected for atmospheric inter- coastline identification using a water index algorithm and (2)
ference caused by haze, dust, smoke, etc. using the dark-object coastline digitization for mapping erosion/accretion patterns along
subtraction method (Chavez, 1996). A DEM was used for topo- Hatiya Island.
graphic correction, and the sun azimuth and sun elevation data The MNDWI water index algorithm was generated in the Idrisi
extracted from the image’s header file were used for radiometric Environment using the modeler function of Idrisi Selva software
correction (Rahdary, 2008). All images were converted to top-of- (Clark Labs, 2012) by combining the green and mid-infrared bands.
atmosphere reflectance values as per the suggestion of Chander The green band (0.52–0.6 lm) is sensitive to water turbidity differ-
and Markham (2003) so that a standardized measure could be ences as well as sediment and pollution plumes because it covers
obtained for comparison between images. To ensure that images the green reflectance peak from leaf surfaces. It can be useful for
taken at different dates could be radiometrically compared, a discriminating broad classes of vegetation. The mid-infrared band
relative radiometric normalization was performed by normalizing (1.60–1.70 lm) exhibits a strong contrast between land and water
the variation in solar illumination and atmospheric conditions features due to the high degree of absorption by water and the
(Mas, 1999; Coppin et al., 2004). All the images were geo-rectified strong reflectance by vegetation and natural features in the range.
using twenty-one ground control points (GCP) with a root mean Thus, the MNDWI algorithm, which is a combination of green and
square error (RMSE) of 0.002047 pixels. GCPs were collected from mid-infrared bands, is ideal for discriminating between land and
road intersections, prominent geomorphologic features and river water at their interface.
140 M.K. Ghosh et al. / ISPRS Journal of Photogrammetry and Remote Sensing 101 (2015) 137–144

Table 1
Description of Landsat scenes and corresponding reference data.

Year Landsat scene Reference data and sources

Acquisition date Path Row
1989 07, December, 1989 136 45 LGED map, 1989 (published on 1994)
2000 21, December, 2000 136 45 LGED map, 2000 (published on 2001)
2010 09, December, 2010 136 45 Google Earth™ image acquired on 07 December, 2010

MNDWI was estimated as (Green  MIR)/(Green + MIR), where The image classification resulted in kappa index of 0.83, 0.85
Green and MIR are the reflection in the green and mid-infrared and 0.96 and overall accuracies of 93%, 94% and 96% for the images
bands of the TM and ETM images, respectively. The MNDWI algo- of 1989, 2000, and 2010, respectively (Table 2). The MNDWI-gen-
rithm was applied over the TM and ETM images of 1989, 2000 and erated images had accuracies of 88%, 90% and 95% for the images
2010. Afterwards, a Boolean approach was used over the MNDWI of 1989, 2000, and 2010, respectively (Table 3), suggesting that
images to create two classes: land and water. the reliability of MNDWI technique is robust. User and producer
To detect the changes along the coastline, classified images accuracies were also calculated for each class for all the images
were overlaid and on-screen digitizing of coastline was undertaken and the accuracies were consistently high, ranging from 82% to
to create the coastline layers. Layers were overlain together so that 98%.
the coastline position could be seen at each date. Coastline posi- The changes in coastline positions are shown in Fig. 3(a)–(c)
tions were highlighted to infer the erosion/accretion sectors along while the extent of erosion and sedimentation are given in Tables
the coast, and the coastline changes were calculated. The coastline 4–7. The locations of major erosion and accretion areas were
of Hatiya Island is subject to normal high and low tides. Seasonal identified from satellite data for each study period. Between these
variations in the monthly mean sea level along this coast are high. erosion- and accretion-prone segments, there were also some
The sea level is highest in June, July and August, which corresponds unaffected locations.
to the monsoon season in Bangladesh and the maximum discharge The location of the areas of the coastline change between 1989
of all rivers. The sea level is lowest during January, February and and 2000 is shown in Fig. 3(a). Four erosion dominant locations
March due to the winter season when rainfall is very low. Large (sites A, B, C, and D) and one accretion dominant location (site E)
fluctuations in the mean sea level due to weather conditions are were identified from satellite data along Hatiya Island’s coastline.
also observed along this coast. During the monsoon the heights The black line in the map represents the coastline in 1989, whereas
of the low and high tides range between 0.9 and 3.5 m. In contrast, the blue3 line represents the position of the coastline in 2000. Site A,
during winter, the heights of the low and high tides range between which extends for 26.98 km, experienced the maximum erosion of
0.5 and 2.8 m. In this study the coastlines were not corrected for 2995 hectares, with a coastline displacement ranging between 0.7
variations in tide levels, and it was assumed that these variations and 3.5 km. Sites B, C and D, which extend for 36.50, 22.84 and
were low compared to the scale of coastline shifts. 3.2 km experienced erosions of 1396, 664 and 31 hectares, with
the coastline displacement ranging between 0.3 and 0.7, 0.4 and
0.65, and 0.15 and 0.28 km, respectively. On the other hand, site E,
2.5. Method used for validating the results which extends for 32.3 km, witnessed an accretion of 1315 hectares
with a coastline displacement ranging between 0.8 and 2 km.
For validation of the MNDWI method and land/water classifica- Fig. 3(b) shows the coastline change between 2000 and 2010.
tion, reference data used in accuracy assessment are described in Two erosion dominant locations (sites A and B) and four accretion
Table 1. For the image of 1989 and 2000, The Local Government dominant locations (sites C, D, E and F) were identified from satel-
Engineering Department (LGED) map of 1989 and 2000 were used lite data along Hatiya Island’s coastline during this study period.
as reference. The high spatial resolution images provided by Goo- The position of coastlines in 2000 and 2010 in the map are repre-
gle Earth™ were used as reference for the image of 2010. The dates sented by black and blue lines, respectively. Sites A and B, which
of the reference data and the analyzed images were closely extend for 18.16 and 9.3 km, witnessed an erosion of 2039 and
matched to minimize bias in the surface water boundaries that 320 hectares and also experienced a coastline retreat ranging
could arise because of large differences in time. For the validation between 0.4 and 2.6 and between 0.5 and 0.9 km, respectively. In
of MNDWI method the ‘‘true’’ boundaries between land and water contrast, the locations that experienced prolonged accretion are
were digitized manually on-screen from all the reference data. sites C, D, E and F, which are 8.6, 11.8, 23 and 40 km long. In sites
Then total areas were calculated for each newly generated C, D, E and F, seaward coastline advancement accounts for 1.5–2.1,
reference data and compared with classified Landsat images. To 1.2–2.3, 2.5–4.5 and 0.9–2.3 km, respectively, whereas these four
assess the accuracy of the land/water classification for all classified sites (C, D, E and F) experienced an accretion of 875, 2213, 5602
maps kappa coefficient (Xu, 2007) was used. Error matrices were and 1649 hectares, respectively.
developed to evaluate the efficacy of the classification. To assess The change of the position of the coastline during the twenty-
the thematic accuracy for each classified map a minimum of 55 one year time period (between 1989 and 2010) is shown in
pixels were selected from each category using the stratified Fig. 3(c). In this period, two erosion dominant locations (sites A
random sampling method and then checked with reference data. and B) and four accretion dominant locations (sites C, D, E and F)
Producer’s and user’s accuracy for each class were also calculated were identified. Sites A and B experienced erosion of 4920 and
along with the overall accuracy and kappa coefficient. 1050 hectares and also witnessed a coastline retreat which ranged
between 0.7 and 6.5, and 0.4 and 1.3 km. On the other hand, the
locations exposed to extended accretion are sites C, D, E and F.
3. Results
The amounts of accretion in these sites were 500, 2848, 4501

The MNDWI and classified images for 1989, 2000 and 2010 are
shown in Fig. 2. The most striking result is the large change in the 3
For interpretation of color in Fig. 3, the reader is referred to the web version of
coastline due to the strong erosion and sedimentation impacts. this article.
 M.K. Ghosh et al. / ISPRS Journal of Photogrammetry and Remote Sensing 101 (2015) 137–144 141

Table 3
MNDWI model generated and reference area of Hatiya Island during the years 1989,
2000, and 2010.

Year Reference area (LGED and Obtained area through Accuracy

Google Earth™) in hectares classification in hectares (in %)
1989 40,121 45,593 88
2000 37,261 37,261 90
2010 47,044 49,521 95

Fig. 2. MNDWI and classified (land and water) images of (a) 1989, (b) 2000 and (c)
2010.

Table 2
The kappa coefficient and overall classification accuracy for the images of Landsat TM
1989 and 2000, and ETM 2010.

Classified map Kappa coefficient Overall classification accuracy (in %)

1989 0.83 93
2000 0.85 94
2010 0.96 96 Fig. 3. Coastline change of Hatiya Island Bangladesh, based on image analysis from
(a) 1989 to 2000, (b) 2000 to 2010 and (c) 1989 to 2010.
142 M.K. Ghosh et al. / ISPRS Journal of Photogrammetry and Remote Sensing 101 (2015) 137–144

Table 4
Erosion and accretion of selected sites in Hatiya Island, Bangladesh (1989–2000). See Fig. 3(a) for the location of the sites.

Site Erosion (in hectares) Accretion (in hectares) Overall change (in hectares) Dominant process
A 2994.84 113.67 () 2881.17 Erosion
B 1396.44 0 () 1396.44 Erosion
C 664.38 0 () 664.38 Erosion
D 30.69 0 () 30.69 Erosion
E 506.34 1315.17 (+) 808.17 Accretion

Table 5
Erosion and accretion of selected sites in Hatiya Island, Bangladesh (2000–2010). See Fig. 3(b) for the location of the sites.

Site Erosion (in hectares) Accretion (in hectares) Overall change (in hectares) Dominant process
A 2038.95 0 () 2038.95 Erosion
B 320.22 0 () 320.22 Erosion
C 127.17 875.25 (+) 748.08 Accretion
D 0 2213.19 (+) 2213.19 Accretion
E 32.31 5601.87 (+) 5569.76 Accretion
F 141.84 1649.34 (+) 1507.50 Accretion

Table 6
Erosion and accretion of selected sites in Hatiya Island, Bangladesh (1989–2010). See Fig. 3(c) for the location of the sites.

Site Erosion (in hectares) Accretion (in hectares) Overall change (in hectares) Dominant process
A 4920.48 0 () 4920.48 Erosion
B 1049.94 0 () 1049.94 Erosion
C 192.32 500.49 (+) 307.17 Accretion
D 0 2848.41 (+) 2848.41 Accretion
E 146.88 4500.99 (+) 4354.11 Accretion
F 165.15 2066.04 (+) 1900.89 Accretion

Table 7
Area of Hatiya Island, Bangladesh and its rate of change in different years (1989–2010).

Year Area in hectares Change in area from 1989 Change rate (in %) Remarks
1989 45,593 – – Base map
2000 41,402 4191 9.19 Erosion in reference to 1989
2010 49,521 3928 8.62 Accretion in reference to 1989

Note: Area of the whole island has been considered.

and 2066 hectares, whereas seaward coastline advancement of the northern and western parts of the island, while deposition took
these sites varied between 0.4 and 1.3, 0.5 and 4.4, 1.2 and 5.1, place in the southern and eastern parts.
and 0.3 and 1.8 km, respectively. During the twenty-one year study period, approximately 6476
hectares were eroded in the study area. Erosion along the coastline
4. Discussion occurred at different rates. Most erosion was encountered in the
northern part of the island (site A, Fig. 3(c)) that is very close to
Hatiya Island is a part of the lower Meghna estuary and consists the Meghna river mouth, and the rate of erosion decreased gradu-
of quaternary alluvial deposits of silt, sand and clay. As a result of ally in the western part of the island (site B, Fig. 3(c)). As a result of
river discharge, tides and coastal hydrology of the region, morphol- erosion, one administrative unit of the study area, the Harni union
ogy of the island is changing rapidly. Hatiya Island is an erosion (located in site A, Fig. 3(c)), was totally eroded during the study
and accretion prone region. Channel migration of Meghna Old period. This erosion process collapsed the settlement area, causing
River has been very rapid, causing severe erosion of Hatiya Island damage and loss of property. Agricultural land was reduced by this
in recent times (1989–2010). It is evident that significant changes erosion process and thus made many people homeless and jobless
have taken place in the coastline of the island, and remote sensing in affected areas like Harni union (ICZM, 2003).
data also confirm this trend. It has been observed that erosion and Accretion has taken place in the southern and eastern parts of
accretion processes are mainly responsible for the coastline change the island. Most accretion was encountered at sites D, E and F
in the study area. (Fig. 3(c)). Approximately 9916 hectares were accreted during
Analysis of coastline change for Hatiya Island shows that the the study period (1989–2010) in the study area, which had a very
locations of coastlines have changed remarkably between 1989 positive impact for seaward advancement of the coastline. Land
and 2000 and between 2000 and 2010 (Fig. 3(a) and (b)). During use pattern of the study area was also positively influenced by this
the first time interval, erosion was the dominant physical force accretion process (Kumar and Ghosh, 2012).
in bringing about changes in the coastline in the study area, but Due to erosion and accretion processes, the coastline of Hatiya
results suggest that the situation was different for the 2000– Island experienced massive changes during the study period. Sites
2010 time period when accretion was dominant. The study has A and B experienced a negative (erosion) coastline change that ran-
also revealed that during 1989–2010, more erosion took place in ged between 0.7 and 6.5 and between 0.4 and 1.3 km, respectively.
 M.K. Ghosh et al. / ISPRS Journal of Photogrammetry and Remote Sensing 101 (2015) 137–144 143

In contrast, sites C, D, E and F experienced a positive (accretion) northern, eastern and southern parts of the island, where major
coastline change that varied between 0.4 and 1.3, 0.5 and 4.4, 1.2 erosion and accretion had occurred. During the study period, sites
and 5.1, and 0.3 and 1.8 km, respectively (Fig. 3(c)). These findings A and B, where the trend is towards erosion, witnessed a coastline
prove that the coastline has been changed in the study area during retreat, whereas sites C, D, E and F, where the trend is towards
the study period. accretion, experienced a seaward coastline advancement.
Both the erosion and accretion processes are continuously Continuous monitoring of coastlines is important for under-
changing the landform of the study area. Old and mature soils standing the changes in the coastline of Hatiya Island. Further-
are eroding away and new land comprising unconsolidated soils more, studies are recommended that link land use patterns,
are accreting. Mature soils are suitable for year round cultivation, hydrological data, human activities and other processes taking
while new soils are generally not fertile. Salt contents are also high place in the coastal and marine areas of the island. To monitor
in this soil. Although the net balance is towards accretion, loss of changes in the future, results of this study should be integrated
old fertile land is considered a national loss (ICZM, 2003). In with other datasets such as land use patterns, hydrological data
addition, land cover patterns have changed dramatically in these and human activities. These results can also be shared with coastal
erosion and accretion prone areas by following these morphologi- managers, relevant stakeholders, the public and policy makers for
cal processes. During the study period land under agriculture and use in decision-making on such issues as land use planning and
settlements have increased in all the unions of the Hatiya Island identification of low-cost methods, which could be used in the
except Nalchira, Sukhchar and Harni unions, where erosion process context of integrated coastal management.
plays an active role to oppose such change. On the contrary accre- This study demonstrates the applicability of remote sensing and
tion process is more dominant than erosion in the southern part of GIS techniques for monitoring the coastlines of vulnerable islands
the island, where medium forest/vegetation and forest classes have such as Hatiya. Integration of aerial photography with high resolu-
increased due to the reforestation program of the Bangladesh For- tion imagery from other satellites (e.g., Quickbird) could further
est Department, and mud and marshy land have decreased due to improve the accuracy of the models.
land reclamation policy of the Bangladesh Government (Kumar
and Ghosh, 2012; Dwip Unnayan Report, 2011).
It should be noted that the older datasets (TM 1989 and 2000)
References
have resulted in relatively lower overall accuracies and kappa coef-
ficient values compared to the newer datasets. The quality of the Addo, A.K., Walkden, M., Mills, J.P., 2008. Detection, measurement and prediction of
reference datasets are an issue for older images because the shoreline recession in Accra, Ghana. ISPRS J. Photogramm. Remote Sens. 63 (5),
543–558.
training sites were developed from paper-based maps which might
Anderson, J.R., 1976. A land use and land cover classification system for use with
result in misleading classification. These could be the reasons for remote sensor data. Geological Survey Professional Paper 964. U.S. Geological
the lower accuracy of the classified map for the years 1989 and Survey, Washington, D.C..
2000 compared to the map of the year 2010. An examination of Blodget, H.W., Taylor, P.T., Roark, J.H., 1991. Shoreline changes along the Rosetta-
Nile promontory: monitoring with satellite observations. Mar. Geol. 99 (1), 67–
the accuracies of the land/water classification maps revealed that 77.
all the classified images had overall accuracies of more than 85%, Chander, G., Markham, B., 2003. Revised Landsat-5 TM radiometric calibration
which meets the level (85%) that the U.S. Geological Survey has procedures and postcalibration dynamic ranges. IEEE Trans. Geosci. Remote
Sens. 41, 2674–2677.
recommended for acceptability of classification results Chavez, P.S., 1996. Image-based atmospheric corrections-revisited and improved.
(Anderson, 1976). Photogramm. Eng. Remote Sens. 62 (9), 1025–1035.
Coastal areas are very sensitive to sea level rise, extreme Chu, Z.X., Sun, X.G., Zhai, S.K., Xu, K.H., 2006. Changing pattern of accretion/erosion
of the modern Yellow River (Huanghe) subaerial delta, China: based on remote
weather events such as cyclones, increased levels of precipitation sensing images. Mar. Geol. 227 (1), 13–30.
and warmer sea temperatures. Changes in the global climatic Clark Labs. 2012. IDRISI help system <http://www.clarklabs.org/support/IDRISI-
patterns are expected to worsen the situation by influencing the Taiga-Help-System.cfm> (Accessed 19.07.14).
Coppin, P., Jonckheere, I., Nackaerts, K., Muys, B., Lambin, E., 2004. Digital change
above-mentioned issues that the coasts are sensitive to. Under detection methods in ecosystem monitoring: a review. Int. J. Remote Sens. 25
the changing climatic scenario several factors, such as rapid glacier (9), 1565–1596.
melt in the Himalayas and enhanced monsoon precipitation, are Darvishzadeh, R. 2000. Change detection for urban spatial databases using remote
sensing and GIS. In: Proceedings of ISPRS 2000–320. Retrieved from <http://
likely to contribute to increased flooding risk along the coast of
www.isprs.org/proceedings/xxxiii/congress/part7/313_XXXIII-part7.pdf>.
Bangladesh. In addition to this, increases in the sea level are Dewan, A.M., Yamaguchi, Y., 2009. Using remote sensing and GIS to detect and
expected to cause higher levels of storm surges along the coast. monitor land use and land cover change in Dhaka Metropolitan of Bangladesh
This will increase the magnitude and coverage of area of inunda- during 1960–2005. Environ. Monit. Assess. 150 (1), 237–249.
DWIP Unnayan. 2008. Dwip Unnayan Sangstha Yearly Report, Bangladesh Ministry
tion. Climate change will have adverse effects on low lying offshore of Environment.
islands like Hatiya Island. Climate change will not only add new DWIP Unnayan. 2011. Dwip Unnayan Sangstha Yearly Report, Bangladesh Ministry
factors that would impact on the coastline of the Hatiya Island of Environment.
El-Asmar, H.M., Hereher, M.E., 2011. Change detection of the coastal zone east of the
mainly through inundation but also make the existing factors more Nile Delta using remote sensing. Environ. Earth Sci. 62 (4), 769–777.
destructive by increasing the level of discharge through the rivers. Gens, R., 2010. Remote sensing of coastlines: detection, extraction and monitoring.
A continuous satellite image-based monitoring program of the Int. J. Remote Sens. 31 (7), 1819–1836.
Gibb, J.G., 1978. Rates of coastal erosion and accretion in New Zealand. NZ J. Mar.
coastal region will be useful to better understand the dimensions Freshwat. Res. 12 (4), 429–456.
and rates of coastline change along Hatiya Island. Harvey, K.R., Hill, G.J.E., 2001. Vegetation mapping of a tropical freshwater swamp
in the Northern Territory, Australia: a comparison of aerial photography,
Landsat TM and SPOT satellite imagery. Int. J. Remote Sens. 22 (15), 2911–2925.
Hereher, M.E., 2010. Sand movement patterns in the Western Desert of Egypt: an
5. Conclusions environmental concern. Environ. Earth Sci. 59 (5), 1119–1127.
ICZM (Integrated Coastal Zone Management Plan Project). 2003. Government of the
People’s Republic of Bangladesh, Ministry of Water Resources, Water Resources
Position of the coastline and its changing patterns have been
Planning Organization (WARPO), 119pp.
mapped in Hatiya Island by integrating remote sensing and Jayson-Quashigah, P.-N., Addo, K.A., Kodzo, K.S. 2013. Medium resolution satellite
geographic information system (GIS) techniques for a twenty-one imagery as a tool for monitoring shoreline change. Case study of the Eastern
year time period (1989–2010). The results show that significant coast of Ghana <http://ics2013.org/papers/Paper3676_rev.pdf>.
Kumar, L., Ghosh, M.K., 2012. Land cover change detection of Hatiya Island,
changes in the coastline took place during the study period. The Bangladesh, using remote sensing techniques. J. Appl. Remote Sens. 6 (1),
highest amount of coastline displacement was observed in the 063608.
144 M.K. Ghosh et al. / ISPRS Journal of Photogrammetry and Remote Sensing 101 (2015) 137–144

Mas, J., 1999. Monitoring land-cover changes: a comparison of change detection Xu, H., 2007. Extraction of urban built-up land features from landsat imagery using
techniques. Int. J. Remote Sens. 20, 139–152. a thematic oriented index combination technique. Photogramm. Eng. Remote
Naji, T.A.H., Tawfeeq, R.J., 2011. Detection of shoreline change in AL-Thirthar Lake Sens. 73 (12), 1381–1391.
using remotely sensed imagery and topography map. IBN AL-HAITHAM J. Pure Yamano, H., Shimazaki, H., Matsunaga, T., Ishoda, A., McClennen, C., Yokoki, H.,
Appl. Sci. 24 (1). Kayanne, H., 2006. Evaluation of various satellite sensors for waterline
NASA (National Aeronautics and Space Administration). 2011. Landsat 7 science extraction in a coral reef environment: Majuro Atoll, Marshall Islands.
data users handbook. Geomorphology 82 (3), 398–411.
Rahdary, V., 2008. Land use and land cover change detection of Mouteh wildlife Zhang, Y., 2011. Coastal environmental monitoring using remotely sensed data and
refuge using remotely sensed data and geographic information system. World GIS techniques in the modern Yellow River delta, China. Environ. Monit. Assess.
Appl. Sci. J. 3, 113–118. 179 (1–4), 15–29.
Ryu, J.-H., Won, J.-S., Min, K.D., 2002. Waterline extraction from Landsat TM data in Zimmermann, G., Bijker, W. 2004. Monitoring the Amazon with different spatial and
a tidal flat: a case study in Gomso Bay, Korea. Remote Sens. Environ. 83 (3), temporal resolution. In: The XX ISPRS Congress Proceeding. Retrieved from
442–456. <http://www.cartesia.org/geodoc/isprs2004/comm4/papers/481.pdf>.
USGS. 2013. Landsat 5 History <http://landsat.usgs.gov/about_landsat5.php>
(Accessed 24.09.14).
Xu, H., 2006. Modification of normalised difference water index (NDWI) to enhance
open water features in remotely sensed imagery. Int. J. Remote Sens. 27 (14),
3025–3033.