International Journal of Remote Sensing

ISSN: 0143-1161 (Print) 1366-5901 (Online) Journal homepage: https://www.tandfonline.com/loi/tres20

Mapping mangroves using LISS-IV and Hyperion

data in part of the Indian Sundarban

Biswajit Mondal, Ashis Kumar Saha & Anirban Roy

To cite this article: Biswajit Mondal, Ashis Kumar Saha & Anirban Roy (2019): Mapping
mangroves using LISS-IV and Hyperion data in part of the Indian Sundarban, International Journal
of Remote Sensing, DOI: 10.1080/01431161.2019.1630784

To link to this article: https://doi.org/10.1080/01431161.2019.1630784

Published online: 17 Jun 2019.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tres20
INTERNATIONAL JOURNAL OF REMOTE SENSING
https://doi.org/10.1080/01431161.2019.1630784

Mapping mangroves using LISS-IV and Hyperion data in part

of the Indian Sundarban
a a b
Biswajit Mondal , Ashis Kumar Saha and Anirban Roy
a
Department of Geography, Delhi School of Economics, University of Delhi, Delhi, India; bDepartment of
Environment, Government of West Bengal, West Bengal Biodiversity Board, Kolkata, India

ABSTRACT ARTICLE HISTORY

Mangroves in different part of the globe including world’s largest Received 16 November 2018
halophytic population in Sundarban are under tremendous pressure Accepted 8 April 2019
due to global warming, sea level rise, natural disasters and ever-
increasing influence of human population. In such an alarming situa-
tion conservation of mangrove population has been recommended by
the experts. Species-level mapping of mangroves is one of the impor-
tant steps for sustainable conservation of mangrove ecosystem. This
study demonstrates the efficiency of a combination of Hyperion hyper-
spectral and IRS Resourcesat-2 LISS-IV multispectral data for discrimi-
nating and mapping some mangrove species at Lothian Island and
Saptamukhi Reserve Forest located in the western part of Sundarban
Biosphere Reserve of India. The spectral signature drawn on Hyperion
data helped to extrapolate field samples of various pure pixels of the
mangrove plant species using a region growing approach. Those
training samples were used to apply Object-based Image Analysis
(OBIA) classification on 5 m spatial resolution LISS-IV data. Mostly
dominant mangrove species like Aegialitis rotundifolia, Aegiceras corni-
culatum, Avicennia community, Ceriops community, Excoecaria agallo-
cha, Lumnitzera racemosa and Phoenix paludosa have been
successfully classified along with other associated landuse/landcover
in these islands. The methodology used in the study has a huge
potential in identification, mapping and management through
restoration of mangroves in difficult terrain like the Sundarban.

1. Introduction
Mangroves are taxonomically diverse assemblage of woody spermatophytes, with con-
vergent adaptive radiation, growing tropical wet coastal ecosystems with saline tidal
influence and loose substrates (Snedaker 1978). According to Tomlinson (1986), man-
grove is an intertidal ecosystem or highly adopted plant families that lives in the coastal
environment. Generally, mangroves having special adaptive features like vivipary, pneu-
matophores, salt excretion through salt glands are considered to be true mangroves,
whereas, the plants co-existed with mangroves without such special adaptive feature are
the mangrove associates which are non-arborescent, herbaceous, sub-woody and clim-
ber species and can also extend their population in the adjoining tidal periphery of

CONTACT Ashis Kumar Saha [email protected] Department of Geography, Delhi School of

Economics, University of Delhi, Delhi 110007, India
© 2019 Informa UK Limited, trading as Taylor & Francis Group
2 B. MONDAL ET AL.

mangrove habitat (Naskar and Mandal 1999). The mangroves not only play the pivotal
role in maintaining the ecological security of wet coastal estuarine ecosystem, but also
provides immense ecological and economic services to mankind with the livelihood
support and maintenance of coastal stability by checking soil erosion (Lugo and
Snedaker 1974), protecting storms and landward sea ingression (Nabahungu and
Visser 2011; Ha, Dijk, and Visser 2014; Wood et al. 2013).
Sundarban, the largest deltaic settings of the world (covers approximately 10,000 km2
area; 40% in India, 60% in Bangladesh) is a heaven for nurturing magnificent biodiversity
with rich heritage of mangrove flora and thus be designated as world heritage site by
UNESCO in 1987. Indian Sundarban has recently been designated as Ramsar Site (https://
rsis.ramsar.org/ris/2370).
In spite of the significant ecological and economic services, mangroves belong to the
most threatened and vulnerable ecosystem worldwide. Mangrove ecosystem has experi-
enced a dramatic negative impact during the last half century (Kuenzer et al. 2011; Ha,
Dijk, and Visser 2014; Islam and Gnauck 2008). The most highlighted factors concerning
mangrove forest loss are shrimp farming (Gilman et al. 2008; Pramanick 2014), indus-
trialization, agriculture (Farnsworth and Ellison 1997), tsunami, cyclone (Chan and Baba
2009) and global warming (Alongi 2008) and other anthropogenic activities (Blasco and
Aizpuru 2002; Pramanick 2014). Moreover, sea level rise has caused coastal erosion and
submergence of many islands of Sundarban. Therefore, mangrove dominated land
masses have been extensively reduced (Mondal and Saha 2018). As a consequence,
mangrove species have become seriously threatened, and some are extinct too from
many regions. Thus, there is an urgent need to develop plans and procedures for
mangrove protection and management as well as up to date information of mangrove
composition, species distribution and pattern of species zonation. However, one of the
greatest limitations to mangrove protection is the lack of proper inventory and monitor-
ing due to difficulties in collecting information through field survey in a repetitive
manner.
Although cloud-free satellite data availability is an issue, remote sensing technology
has been playing an important role to map and monitor mangroves in harsh tropical
coastal environment like Sundarban. Field-based data collection in such environment is
a difficult task due to inaccessibility and natural habitat of several carnivorous animals
like Royal Bengal Tiger, Crocodiles, etc. Generally, the conventional remote sensing
instruments and techniques are not sufficient for species-level mapping of mangroves
(Benfield et al. 2007). The spatial and spectral information provided by these conven-
tional data and methods are not even sufficient for analysis of mangrove diversity
(Heumann 2011; Held et al. 2003). Therefore, for classification and analysis of mangroves
at species/community level, a combination of new generation high spatial and spectral
data are sought.
Hyperspectral data is being used as a reliable ancillary data to draw spectra (Held et al.
2003; Kamal and Phinn 2011; Giri et al. 2014; Kumar et al. 2017) of various plant species. It
provides ample spectral information to identify and distinguish spectrally similar features
or objects. The EO-1 Hyperion hyperspectral images comprised of both visible – near-
infrared (VNIR: 1–70) and short wave infrared (SWIR: 71–242) bands are freely available
through United States Geological Survey (USGS) to perform such analysis. The spectral
features of mangroves depend on number of factors including biochemical composition,
 INTERNATIONAL JOURNAL OF REMOTE SENSING 3

leaf structure and liquid water content of leaves in micro-scale (Fourty et al. 1996) and also
depend on canopy structure, soil background reflectance (Schmidt and Skidmore 2003).
However, the hyperspectral data is limited up to 30 m spatial resolution, too coarse for
differentiation of spatial distribution of mangroves. In the recent time, LISS-IV remote
sensing data from IRS Resourcesat-2 mission has increased the potential of mangrove
mapping due to its high spatial resolution (5 m), although the spectral resolution is limited
to green, red and NIR bands only. Commercially available very high resolution satellite data,
such as, QuickBird, IKONOS, WorldView series can also be exploited for such purpose
(Wang, Sousa, and Gong 2004a; Wang et al. 2004b; Wang, Silvan-Cardenas, and Sousa
2008; Wang et al. 2016), but in the present context they are not cost effective. Therefore, in
this research work, an attempt has been made to integrate the use of EO-1 Hyperion and
IRS Resourcesat-2 LISS-IV data for species-level mangrove mapping in some easily acces-
sible part of the Indian Sundarban.

2. Study area
The largest active delta and world’s largest halophytic formation of Sundarban mangrove
are situated on the estuary formed by mainly three great rivers Ganges, Brahmaputra and
Meghna. According to Hazra et al. (2002) the Sundarban delta was initiated at the end of
Miocene, and the present scenario of Sundarban delta have reached not more than 10,000
years back (Pleistocene to recent). Sundarban is situated in a humid region and mainly with
tropical climate. The annual maximum temperature is 35°C, whereas minimum tempera-
ture drops at 9°C during winter season. Average annual rainfall is 144 cm, out of which 70%
is received during June to September through monsoon phenomena.
The people of Sundarban are depending mainly on primary activities for their liveli-
hood. Fishing has remained major livelihood activities in this region, spreading over
deep sea water, river and commercial fishing in small/big ponds and in wetlands.
Agriculture is another important source of livelihood and about 50% of people are
depending on agriculture and allied activities for their livelihood. Besides these activities,
forest products like wood, honey, wax, medical plants collection are important sources
of livelihood. Recently, through the constant effort of the Government of West Bengal,
Sundarban is being developed as an important tourist destination, which also acts as
prosperous livelihood activities for the local people.
Identification of some of the mangrove species and mapping were carried out in Lothian
Island and Saptamukhi Reserve Forest comprising of Henry Islands and Bakkhali coast
(Figure 1) due to abundance of almost all species of mangroves of Sundarban and easy
accessibility to conduct field survey. Although Saptamukhi Reserve Forest and its surround-
ings are influenced by anthropogenic activities, Lothian Island is free from human influ-
ence. The study area is geographically located between 21°33ʹ33.00”N and 21°42ʹ27.78”N
latitude and between 88°20ʹ57.51”E and 88°15ʹ44.02”E longitude (Figure 1).

3. Database
The database used in the study comprises of EO-1 Hyperion and IRS Resourcesat-2
LISS-IV satellite data, topographical base map and field samples of different man-
grove species. The Hyperion data were used in this study due to its open access
4 B. MONDAL ET AL.

Figure 1. Location map of Lothian Island and Saptamukhi Reserve Forest. Resourcesat-2 LISS-IV
images are shown in the background.

(through https://earthexplorer.usgs.gov/) and availability of selected cloud-free

scenes over the desired study area. The data consist of 242 narrow spectral bands
(10 nm width) with spatial resolution of 30 m. The bands 1–70 are in VNIR, and the
rest of 71–242 bands are in SWIR (Table 1). The LISS-IV data consists of three spectral
bands in visible and near-infrared region (Table 1). The spatial coverage of LISS-IV is
5.8 m, however, the ortho-rectified and resampled images are usually supplied at
5 m spatial resolution (https://nrsc.gov.in/).
 INTERNATIONAL JOURNAL OF REMOTE SENSING 5

Table 1. Specifications of remote sensing data used.

Path/ Product Wavelength Spatial resolution
Data Date of acquisition Row type Bands (nm) (m)
EO-1 Hyperion 23 November 2014 138/45 L1R 16 bit 242 355.59–2577.08 30
IRS Resourcesat- 2 25 January 2014 108/57 L1A 10 bit Band 2 520–590 5
LISS-IV (Green)
Band 3 620–680
(Red)
Band 4 (NIR) 770–860

4. Methodology
The broad methodology adopted in this study is shown in Figure 2.

4.1. Pre-processing of satellite images

Image pre-processing is necessary to minimize the effects of atmospheric absorption
and scattering, sensor calibration, standardization and geometric integrity. Both the
satellite data sets were pre-processed for atmospheric and radiometric corrections and
co-registered with half a pixel accuracy for further analysis. At first, the Digital Number
(DN) value of LISS-IV image was converted to top-of-atmosphere radiance, followed by
surface reflectance using standard procedure (Chavez Jr 1996). However, for EO-1
Hyperion data, removal of bad bands, water absorbed bands, abnormal vertical strips
and low signal-to-noise ratio bands (Pervez, Khan, and Valiuddin 2015; Datt et al. 2003)
were carried out in ENVI software. After removal of 84 bad bands, remaining 158 bands
were converted to radiance image. Further, Fast Line-of-sight Atmospheric Analysis of
Hypercubes (FLAASH) atmospheric correction model was used to convert radiance
image to reflectance image using input parameter listed in Table 2.

4.2. Field sampling

A field survey was conducted during winter (2015–2016) considering mangrove
phenological cycle same as remote sensing data acquizition in the study area. The
information consists of ground distribution of mangrove species and other associated
landuse and landcover classes (Table 3). As mentioned earlier, conducting field survey
in Sundarban remains a difficult task due to remote inaccessible locations and harsh
environment. A handheld Germin Montana 650 GPS along with Nikon GPS (GP-1A)
assisted geocoded digital photography was used to collect approximately five ground
control points from pure populations of each species spanning over a sufficiently
larger area in both the study areas, i.e. Lothian Island and Saptamukhi Reserve Forest
(Figure 3).

4.3. Spectral signatures of mangrove species

Due to unavailability of field spectroradiometer, the spectral signature curve of various
species was generated through overlaying the ground control points (or seed points of
each pure population) on Hyperion image (Figure 4). Image-based endmember
6 B. MONDAL ET AL.

Figure 2. Methodological framework of mangrove mapping.

Table 2. Parameters used for FLAASH atmospheric correction.

Latitude 21°38ʹ0.25” N Water retrieval No
Longitude 88°18ʹ21.19” E Water column multiplier 1
Date of acquisition 23 November 2014 Aerosol model Rural
Time (start) 03:32:31 Aerosol retrieval 2-Band (K-T)
Ground elevation 0.006 km Initial visibility 40 km
Sensor altitude 705 km Spectral polishing Yes
Sensor type Hyperion Width (No. of bands) 9
Atmospheric model Tropical Wavelength recalibration No
Source: ENVI 5.0 manual and EO-1 Hyperion product guide

reflectance spectra were ideal because they were drawn from the population of the data
points to enhance the accuracy of landcover estimation (Kamal and Phinn 2011).
Seven mostly dominant mangrove species/community along with five other associated
land-cover types were distinguished from unique pattern of spectral signature curves. To
Table 3. Characteristics of mangroves and associated landuse/landcover.
Mangroves/Landuse landcover Important characteristics
Aegialitis Rotundifolia ● Local name: Tora
● Soft-wooded shrubs and with a club-shaped stem
Prefer to grow in more saline mudflat habitat
Aegiceras corniculatum ● Local name: Khalsi
● Small trees, up to 4 m in height.
● Flowers generally white, fruit capsule cylindrical, one seeded.
● Bark is smooth, dark brown or grey.
● True mangrove species found in riverine areas.
Avicennia community Avicennia alba ● Local name: Baen
● It has smooth light grey bark made up of thin, stiff, brittle flakes.
● Grows in high saline tidal flow.
Avicennia marina ● Local name: Jat baen
● Shrubs or small bushy trees, and up to 4 m in height
● Bark is whitish-brown and leaves are elliptic and flowers are yellow in colour.
Avicennia officinalis ● Local name: Piara baen
● Trees, up to 20 m in height
● It has pencil-like pneumatophores, grey-brown bark elliptic leaves, flowers yellow and fruit capsule, flat, four seeded.
Ceriops community Ceriops decandra ● Local name: Garan
● Small trees or shrubs
● Prefer to grow in estuarine condition.
Ceriops tagal ● Local name: Garan
● Small trees, evergreen and up to 5 m high.
● Flowers generally white and grow during September–January months.
Excoecaria agallocha ● Local name: Gengwa
● Generally, 12 m in height and grey in colour.
● Grows mainly in comparatively less saline environment mostly away from creeks.
Lumnitzera racemosa ● Local name: Kripa
● Restricted in landward fringe of mangrove community in low-lying mudflat with regular tidal inundation.
Phoenix paludosa ● Local name: Hental
● Not more than 5 m in height
● Prefer to grow in high-elevated saline soil in river course.
Mixed mangrove ● More than three or four mangrove species mostly under different genera are dominated in a place.
Salt marsh ● Salt marshes are growing in high saline incrusted soil where no other mangrove establishes themselves, only salt marshes do develop.
● Salt tolerant plants grow in that place like herbs, grasses and shrubs.
INTERNATIONAL JOURNAL OF REMOTE SENSING

Sand ● Sand deposits in coastal area.

Water ● Deep sea water or deep, stagnate water in ponds.
Silted water ● Water with suspended sediments.
7
8 B. MONDAL ET AL.

Figure 3. Field photographs of dominant mangroves in the study area.

understand the separability of each species, the whole EM spectra covered by Hyperion data
were divided into eight spectral groups popularly used by remote sensing satellites (Table
4). Analysis of variance (ANOVA) test was performed to assess the statistical significance
among the group means (Wang and Sousa 2009) of the seven mangrove species considered
in this study. It was observed that mangrove species are highly separable (at 99% con-
fidence level) using visible – near-infrared bands in comparison to SWIR bands (Table 4). The
above observation also favours the potential of LISS-IV data for mangrove discrimination.

4.4. Extrapolation of field samples

As the number of ground control points collected was small, a region growing approach
was applied to increase the sample sizes of each class. The reflectance value of each
endmember sample generated through region growing at each of 158 bands was
correlated with corresponding ground control points or seed points of a particular
class. On an average, 15 randomly selected endmember samples per class that have
coefficient of determination (R2) value more than 0.95 were considered, and the rest of
the extrapolated points were rejected. The sample locations were further validated
through the ecological positioning of each species under expert supervision of trained
plant taxonomist. All such-validated samples were used as training samples in an Object-
based Image Analysis Approach (OBIA).
 INTERNATIONAL JOURNAL OF REMOTE SENSING 9

Figure 4. Mean spectral profile of mangrove species derived from Hyperion data.

Table 4. ANOVA test on eight spectral groups to assess discrimination potential of mangrove
species.
ANOVA test
Landsat 8 OLI IRS R-2 Significant at
equivalent LISS-IV equivalent 99%
Spectral Wavelength Band Hyperion band (Wavelength band (Wavelength confidence
group (nm) name bands (nm)) (nm)) p-Value level?
1 457–508 Blue 11–16 Blue -NA- 0 Significant
(452–512)
2 528–599 Green 18–25 Green Green 0 Significant
(533–590) (520–590)
3 630–681 Red 28–33 Red Red 0 Significant
(636–673) (620–680)
4 691–762 Red 34–41 -NA- -NA- 0.002 Significant
edge
5 772–884 NIR 42–53 NIR NIR 0.001 Significant
(851–879) (770–860)
6 1366–1548 SWIR-A 122–140 Cirrus (IR) -NA- 0.086 Insignificant
(1363–1384)
7 1558–1650 SWIR-B 141–150 SWIR-1 -NA- 0.103 Insignificant
(1566–1651)
8 2080–2345 SWIR-C 193–219 SWIR-2 -NA- 0.012 Insignificant
(2107–2294)
10 B. MONDAL ET AL.

4.5. Object-based image analysis (OBIA)

As the accurate mapping of mangroves is the main focus in this research for effective
conservation and management, traditional pixel-based classification scheme considering
only spectral properties is not sufficient to deal with the complexity within (Wang,
Sousa, and Gong 2004a; Wang et al. 2004b). In contrast, OBIA classification technique
can exploit spatial, spectral, textural and ancillary information in an integrated manner.
The technique was used widely in different fields and the result achieved has remained
far better than the pixel base image classification (Wang, Sousa, and Gong 2004a; Wang
et al. 2004b; Conchedda, Durieux, and Mayaux 2008; Wang, Silvan-Cardenas, and Sousa
2008; Huang and Jia 2012). It is well established that there are large numbers of OBIA
classification methods but none of the methods is linked with field-sampling scheme for
natural landscapes where visual interpretation is not as straightforward as human land-
scape (Heumann 2011). Another reason behind using OBIA as a tool of classification is
that it can incorporate neighbourhood objects with same properties through region
growing process which produces a more accurate representation of surface patterns
compared to pixel-based methods (Blaschke and Strobl 2001).
(i) Multi-resolution segmentation:
To initiate the classification process, the LISS-IV image was first segmented using
most widely used multiresolution segmentation method (Blaschke and Strobl 2001;
Navulur 2007) with scale factor 1, shape factor 0.1 and compactness factor 0.5 derived
through trial and error in eCognition software for better segregation of mangrove
species at 5 m spatial resolution level.
(ii) Segregation of non-vegetative land-cover types:
In the present study area, five non-vegetation and non-mangrove land-cover types,
i.e. water, silted water, sand, salt marsh and Casuarina equisetifolia were identified
through field studies along with mangroves. These non-mangrove classes were partially
discriminated using a rule-based classification approach in OBIA. At first, Normalized
Difference Vegetation Index (NDVI) (Curran 1980) and Normalized Difference Water
Index (NDWI) (Mcfeeters 1995) were calculated for each object or segments of the
image using the following standard formula:

NDVI ¼ ½Rmean-NIR  Rmean-RED =½Rmean-NIR þ Rmean-RED  Eq:1

NDWI ¼ ½Rmean-GREEN  Rmean-NIR =½Rmean-GREEN þ Rmean-RED  Eq:2

where Rmean stands for mean spectral reflectance of each object or segments of
corresponding spectral band (e.g. RED, GREEN, NIR, etc.).
In the Level-1 of rule-set, vegetations were separated from the non-vegetations using
a NDVI threshold value determined through trial and error (Table 5). NDVI value less
than equal to 0.09 is considered as non-vegetations. In Level-2, NDWI threshold values
were used to extract waterbodies, silted water and sand from non-vegetation objects.
The objects with NDWI value greater than equal to 0.12 were classified as water, objects
with NDWI value within 0.12 and 0 were classified as silted water and objects with NDWI
value less than 0 were classified as sand. These water and silted water classes were
mainly aquaculture ponds found within Henry Island of Saptamukhi Reserve Forest. The
remaining unclassified vegetation comprised of non-mangrove and mangroves were
 INTERNATIONAL JOURNAL OF REMOTE SENSING 11

Table 5. Process tree of the OBIA classification.

Levels Image objects Classification method
Level-0 – Multiresolution segmentation
(Scale factor: 1, Shape factor: 0.1, Compactness
factor: 0.5)
Level-1 Vegetation Non-vegetation NDVI threshold
Level-2 – Sand NDWI threshold
Water
Silted water
Level-3 Mangrove Non-mangrove Spectral distance threshold
Level-4 – Casuarina Nearest neighbour
equisetifolia OBIA classification
Salt marsh
Level-5 Aegialitis – Nearest neighbour
rotundifolia OBIA classification
Aegiceras
corniculatum
Avicennia
community
Ceriops community
Excoecaria
agallocha
Lumnitzera
racemosa
Phoenix paludosa
Mixed mangroves

separated by spectral reflectance distance calculation (Kamal, Phinn, and Kasper 2015) in
Level-3 (Eq. 3).

Spectral reflector distance ¼ ½Rmean-NIR  Rmean-RED =½Rmean-GREEN  Rmean-RED  Eq:3

Non-mangroves, i.e. coastal Casuarina equisetifolia and salt marsh were classified using
nearest neighbour sampling approach in Level- 4 and mangrove classes in Level-5.
(iii) Feature selection, optimization and nearest neighbour classification:
In Level-4 and 5 processing, at first, a total of 33 features (Navulur 2007) were
considered consisting of spectral statistics, spectral indices, extent, shape factors of the
objects as well as textural elements (Table 6) based on critical analysis of feature space.
Training objects were collected for each class through overlaying training sample set
generated through the procedure described earlier. Further, a feature space optimiza-
tion routine was applied based on selected samples with nearest neighbour rules to
select best features to discriminate all the classes more efficiently. Six features, i.e. GLCM
homogeneity (all directions), NDWI, Shape index, Maximum spectral difference, NDVI,
Mean NIR were shortlisted in feature space optimization routine and used for classifying
the remaining objects to produce a mangrove classified map separately for two desig-
nated test areas.
(iv) Accuracy assessment:
Accuracy assessment is an integral part of image classification to validate the classi-
fication results. To remove bias in accuracy assessment, an equalized random sampling
technique was used to collect a total 360 test sample points, i.e. 30 pixels in each class
over the study area. Purity of these samples were verified using the correlation coeffi-
cient method similar to one used during training sample selection (Figure 2). These
 12

Table 6. Object features considered for OBIA.

B. MONDAL ET AL.

Customized feature Layer values Feature selection Geometry Feature selection Texture Feature selection
NDVI Mean Brightness Extent Area GLCM homogeneity GLCM homogeneity (all directions)
Maximum spectral difference Border index
NDWI Green Length
NIR Length/thickness
Spectral distance Red Length/width
Standard Green Number of pixels
deviation NIR Relative border to Image border
Thickness
Volume
Width
Shape Asymmetry
Border index
Compactness
Red Density
Elliptic fit
Main direction
Radius of largest enclosed ellipse
Radius of smallest enclosing ellipse
Rectangular fit
Roundness
Shape index
 INTERNATIONAL JOURNAL OF REMOTE SENSING 13

samples were also validated through the ecological positioning of each species. The
object-based classified image was reconverted to pixels and error matrix was computed.
The overall accuracy was calculated by dividing the total-corrected pixels by the total
pixels in the error matrix.
To evaluate the improvement of classification accuracy over traditional pixel-based
spectral classification (e.g. Giri et al. 2014), LISS-IV bands were classified by Maximum
Likelihood Classifier (MLC) using same set of training samples and evaluated with same
set of test samples.
To understand accuracy of individual classes, producer’s and user’s accuracy (Congalton
1991) were also analysed. Generally, the total number of correct pixels is divided by the total
number of reference pixels (column total), called producer’s accuracy which indicates how
well a certain area can be classified. Whereas if the total number of correct pixels in
a category is divided by the total number of pixels that are classified (row total), called
user’s accuracy. It indicates the probability of a pixel classified on the map actually repre-
sents that category on the ground (Story and Congalton 1986).

5. Results and discussion

The methodology described in the previous section was helpful to generate
a detailed map of mangroves and its associated land-cover types. In this section,
the mangrove distribution of the island is described with linkage to its ecological
positioning.

5.1. Mangrove mapping in Lothian Island

As the Lothian Island is free from human influence and around 80% of the land is
dominated by different species mangroves (Table 7). The results of accuracy assessment
in Lothian Island show 75.28% overall accuracy (Table 8) which is 8.89% better than the
traditional pixel-based classification (66.39%). Particularly Aegiceras corniculatum, Ceriops
community and mixed mangroves have achieved better producer’s accuracy (above
72%) which suggest these classes have high identification potential through this
method. The kappa coefficient (κhat) for the classification is also found to be 0.73
which falls under substantial classification category (Landis and Koch 1977; Munoz and
Bangdiwala 1997). Spatial distribution of mangrove species in Lothian Island shows that
Avicennia community and Phoenix paludosa are the dominant mangroves which cover
around 26% and 16% of the area, respectively. Aegialitis rotundifolia (15%) is the third
dominant mangrove species in Lothian Island. In accordance with, a large portion of
Lothian Island is covered by mixed type of mangrove vegetation (about 8%). Other
mangrove communities cover 12% of the area (Figure 5).
Though Aegialitis rotundifolia prefers to grow in more saline habitat, due to its short
stretch mostly avoid to grow on the bank side and thus here growing in the inner part of
the islands having saline mudflats. Three species of Avicennia (A. alba, A. marina,
A. officinalis) were seen growing in the outer estuarine regions of mangrove habitat
with high saline tidal flow. In the present investigation, the image shows that the
population of Avicennia community distributed along the riverside and the creeks inside
the island.
14 B. MONDAL ET AL.

Table 7. Spatial distribution of mangrove species and associated landuse/landcover in Lothian Island
and Saptamukhi Reserve Forest.
Lothian Island Saptamukhi Reserve Forest
Area Area
Types Classes Area (ha) Area (%) (ha) (%)
Mangrove species Aegialitis rotundifolia 574.58 14.96 30.01 3.98
Aegiceras corniculatum 95.76 2.49 – –
Avicennia community 984.85 25.64 145.49 19.28
Ceriops community 175.17 4.56 32.91 4.36
Excoecaria agallocha 162.28 4.22 47.50 6.30
Lumnitzera racemosa – – 27.87 3.69
Phoenix paludosa 599.80 15.61 56.77 7.52
Mixed mangrove 300.80 7.83 74.15 9.83
Non-mangroves Casuarina equisetifolia 34.73 0.90 29.26 3.88
Salt marsh 431.76 11.24 71.79 9.51
Non-vegetation Sand 112.22 2.92 69.28 9.18
Water 141.62 3.69 86.94 11.52
Silted water 227.95 5.93 82.59 10.95

Table 8. Error matrix of the OBIA classification of Lothian Island.

User’s
Classified Producer’s accuracy
AG AV CP EX AC PH MM CS SM SD WT SW pixel total accuracy (%) (%)
Aegialitis 19 3 0 3 0 1 2 0 2 0 0 0 30 68 63
rotundifolia
(AG)
Avicennia 1 22 1 1 0 2 3 0 0 0 0 0 30 65 73
community
(AV)
Ceriops 1 3 18 4 0 2 2 0 0 0 0 0 30 72 60
community
(CP)
Excoecaria 1 2 3 19 2 2 1 0 0 0 0 0 30 59 63
agallocha
(EX)
Aegiceras 2 0 0 1 20 7 0 0 0 0 0 0 30 83 67
corniculatum
(AC)
Phoenix 2 2 1 2 1 21 0 0 1 0 0 0 30 55 70
paludosa
(PH)
Mixed 1 1 1 1 0 3 23 0 0 0 0 0 30 72 77
mangrove
(MM)
Casuarina 0 0 0 1 0 0 0 26 3 0 0 0 30 100 87
equisetifolia
(CS)
Salt marsh (SM) 1 1 1 0 1 0 1 0 24 1 0 0 30 77 80
Sand (SD) 0 0 0 0 0 0 0 0 1 27 0 2 30 93 90
Water (WT) 0 0 0 0 0 0 0 0 0 0 25 5 30 93 83
Silted water 0 0 0 0 0 0 0 0 0 1 2 27 30 79 90
(SW)
Reference pixel 28 34 25 32 24 38 32 26 31 29 27 34 271/360
total
Overall accuracy: 75.28%, Kappa coefficient (κhat): 0.73
 INTERNATIONAL JOURNAL OF REMOTE SENSING 15

Figure 5. Classified map of mangroves and other land-cover types of Lothian Island.
16 B. MONDAL ET AL.

As Phoenix paludosa prefers to grow in the high-elevated saline soil either in the bank
side or upland mudflats or mostly in the penetrable groves. In the present investigation,
the image shows that concentration of Phoenix paludosa either in the southern part of
elevated bank of the island or inside upland mudflat.
Field investigation shows that the Ceriops tagal prefers to grow in high saline tidal
flow of the upper estuarine conditions, whereas, Ceriops decandra are growing compara-
tively less saline conditions having more creeks and canals. The classified map at this
resolution could not differentiate the two species, rather clubbed together indicating
a mosaic population dominating in the entire island. The Excoecaria agallocha has very
sporadic distribution in the island and mostly growing away from the creeks and canals
where salinity is comparatively lesser.
Salt marshes are growing in high saline incrusted soil or in the salt pan. In some low-
elevated regions of the island where tidal water and saline waterlogged regions trans-
form into saline-incrusted soil after due evaporation. Therefore, no other mangrove
plant species establishes and only salt marshes do develop. In the present investigation,
only a very few pockets with such physical conditions facilitate the growth of salt marsh
like Suaeda maritima and Tamarix troupii (Figure 3). There are few non-mangrove
Casuarina equisetifolia patches which are planted mainly in south-east corner of
Lothian Island.
Beside individual mangrove population having the preference of specific physico-
chemical environment, there are some regions of island found to be intermingled with
more than one mangrove population. The same has also been depicted in the classified
map (Figure 5).

5.2. Mangrove mapping in Saptamukhi Reserve Forest

Due to its direct connection with human-populated areas, the Saptamukhi Reserve
Forest (Figure 6) has only 55% area covered by mangroves (Table 7). The accuracy
assessment results indicate that Ceriops community, Lumnitzera racemosa, Phoenix palu-
dosa and Aegialitis rotundifolia, have highest identification potential through this
method as they show above 78% of producer’s accuracy in Saptamukhi Reserve
Forest. The results of accuracy assessment in Saptamukhi Reserve Forest shows 79.72%
overall accuracy (Table 9) which is 8.33% better than the traditional pixel-based classi-
fication (71.39%). The kappa coefficient (κhat) for the classification is also found to be
0.78 which falls also under substantial classification category (Landis and Koch 1977;
Munoz and Bangdiwala 1997).
Spatial coverage and dominant mangrove species in Saptamukhi Reserve Forest
was a bit different from the Lothian Island (Table 7). Avicennia community was most
dominant in Saptamukhi Reserve Forest which covered about 20% of the area.
The second dominant mangroves were Phoenix paludosa consisting of about 8% of
the area, besides these two species, Excoecaria agallocha is the third dominant
species which was comprised of 6.30% area. Mixed type mangroves covered the
reserve forest of about 10% of the area. A large number of artificial ponds used for
aquaculture purpose were seen in this reserve forest. The salt marshes were mostly
distributed everywhere.
 INTERNATIONAL JOURNAL OF REMOTE SENSING 17

Figure 6. Classified map of mangroves and other land-cover types in Saptamukhi Reserve Forest.

In the Henry Island (northeastern part of the reserve forest) along with Saptamukhi
estuary, the vegetation was mainly mosaic of mangrove species (Figure 6). Avicennia
community was mostly distributed in northern part and along the edges of the creeks,
sometimes extended to the mudflats. Phoenix paludosa is colonized on the higher
elevated mudflats, predominantly in the Saptamukhi estuarine regions than the north-
ern and north-eastern part of Henry Island.
Ceriops community were more concentrated in the northern part of the Island
than the Saptamukhi estuarine part. Excoecaria agallocha which are mostly growing
in the lesser saline conditions and thus only having very scanty population in the
extreme northern part of the island. Aegialitis rotundifolia and Lumnitzera racemosa
had very discrete population mostly with low-lying mudflats having regular tidal
inundation. A few exposed non-shady regions of the island with high evaporation
rate creating salt incrusted conditions for salt marshes especially Suaeda nudiflora.
Some planted Casuarina equisetifolia species are found in the southern part of the
reserve forest.
18 B. MONDAL ET AL.

Table 9. Error matrix of the OBIA classification of Saptamukhi Reserve Forest.

User’s
Classified Producer’s accuracy
AG AV CP EX LM PH MM CS SM SD WT SW pixel total accuracy (%) (%)
Aegialitis 21 1 0 2 2 1 3 0 0 0 0 0 30 78 70
rotundifolia
(AG)
Avicennia 1 26 0 2 0 0 1 0 0 0 0 0 30 62 87
community
(AV)
Ceriops 1 2 20 3 1 0 3 0 0 0 0 0 30 87 67
community
(CP)
Excoecaria 1 3 0 20 0 2 4 0 0 0 0 0 30 65 67
agallocha
(EX)
Lumnitzera 1 3 1 0 23 0 2 0 0 0 0 0 30 87 77
racemosa
(LM)
Phoenix 0 4 0 1 0 22 3 0 0 0 0 0 30 85 73
paludosa
(PH)
Mixed 2 1 2 1 1 0 23 0 0 0 0 0 30 56 77
mangrove
(MM)
Casuarina 0 0 0 1 0 1 2 25 1 0 0 0 30 100 83
equisetifolia
(CS)
Salt marsh (SM) 0 2 0 1 0 0 0 0 26 1 0 0 30 87 87
Sand (SD) 0 0 0 0 0 0 0 0 3 27 0 0 30 90 90
Water (WT) 0 0 0 0 0 0 0 0 0 0 27 3 30 96 90
Silted water 0 0 0 0 0 0 0 0 0 2 1 27 30 90 90
(SW)
Reference pixel 27 42 23 31 27 26 41 25 30 30 28 30 287/360
total
Overall accuracy: 79.72%, Kappa coefficient (κhat): 0.78

6. Conclusion
In this study, seven different types of mangrove species along with associated landuse/
landcover could be mapped in Lothian Island and Saptamukhi Reserve Forest in the
Indian Sundarban with substantial accuracy and cost efficiency. The methodology
adopted in this study has overcome the limited availability of ground control points
and absence of field spectroradiometer by using a region growing tool applied on
hyperspectral data. This method would be highly beneficial to map difficult terrain like
Sundarban with greater accuracy. The study also highlights the potential of 5 m spatial
resolution LISS-IV data for mangrove mapping with OBIA approach as high classification
accuracy could be achieved. The present methodology could be replicated for the whole
region to prepare a detailed mangrove map in a repetitive manner which would be
helpful for sustainable management of this fragile ecosystem.

Acknowledgements
Research grant provided by Research Council, University of Delhi is duly acknowledged. Prof.
Elmar Csaplovics, TU-Dresden, Germany through DAAD reinvitation program for providing access
to ENVI software. The first author is thankful to University Grants Commission for research
 INTERNATIONAL JOURNAL OF REMOTE SENSING 19

fellowship. Authors are also thankful to Directorate of Forest, Government of West Bengal for
necessary field permission and Dr Sumit Manna, Assistant Professor, Moyna College, West Bengal
for assistance during field survey. Thanks are due to Dr Gyan Prakash Sharma and Dr Praveen
Kumar Pathak, University of Delhi for their assistance in statistical analyses in SPSS.

Disclosure statement
No potential conflict of interest was reported by the authors.

ORCID
Biswajit Mondal http://orcid.org/0000-0002-7041-5207
Ashis Kumar Saha http://orcid.org/0000-0002-8184-0495
Anirban Roy http://orcid.org/0000-0003-3786-7380

References
Alongi, D. M. 2008. “Mangrove Forests: Resilience, Protection from Tsunamis, and Responses to
Global Climate Change.” Estuarine Coast Shelf Science 76: 1–13. doi:10.1016/j.ecss.2007.08.024.
Benfield, S. L., H. M. Guzman, J. M. Mair, and J. A. Young. 2007. “Mapping the Distribution of Coral
Reefs and Associated Sublittoral Habitats in Pacific Panama: A Comparison of Optical Satellite
Sensors and Classification Methodologies.” International Journal of Remote Sensing 28:
5047–5070. doi:10.1080/01431160701258062.
Blaschke, T., and J. Strobl. 2001. “What’s Wrong with Pixels? Some Recent Development Interfacing
Remote Sensing and GIS.” GeoBIT/GIS 6: 12–17.
Blasco, F., and M. Aizpuru. 2002. “Mangroves along the Coastal Stretch of Bay of Bengal: Present
Status.” Indian Journal of Marine Sciences 31: 9–20.
Chan, H. T., and S. Baba. 2009. “Manual on Guidelines for Rehabilitation of Coastal Forests
Damaged by Natural Hazards in the Asia-Pacific Region.” In International Society for Mangrove
Ecosystems (ISME) and International Tropical Timber Organization (ITTO), 66. Okinawa, Japan.
Chavez Jr, P. S. 1996. “Image-Based Atmospheric Corrections-Revisited and Improved.”
Photogrammetric Engineering & Remote Sensing 62: 1025–1036.
Conchedda, G., L. Durieux, and P. Mayaux. 2008. “An Object-Based Method for Mapping and
Change Analysis in Mangrove Ecosystem.” ISPRS Journal Photogrammetry and Remote Sensing
63: 578–589. doi:10.1016/j.isprsjprs.2008.04.002.
Congalton, R. G. 1991. “A Review of Assessing the Accuracy of Classifications of Remotely Sensed
Data.” Remote Sensing of Environment 37: 35–46. doi:10.1016/0034-4257(91)90048-B.
Curran, P. 1980. “Multispectral Remote Sensing of Vegetation Amount.” Progress in Physical
Geography: Earth and Environment 4: 315–341. doi:10.1177/030913338000400301.
Datt, B., T. R. McVicar, T. G. Van Niel, D. L. B. Jupp, and J. S. Pearlman. 2003. “Preprocessing EO-1
Hyperion Hyperspectral Data to Support the Application of Agricultural Indexes.” Geoscience
and Remote Sensing, IEEE Transactions 41: 1246–1259. doi:10.1109/TGRS.2003.813206.
Farnsworth, E. J., and A. M. Ellison. 1997. “The Global Conservation Status of Mangroves.” Ambio 26:
328–334.
Fourty, T., F. Baret, S. Jacquemoud, G. Schmuck, and J. Verdebout. 1996. “Leaf Optional Properties
with Explicit Description of Its Biochemical Composition: Direct and Inverse Problems.” Remote
Sensing of Environment 56: 104–117. doi:10.1016/0034-4257(95)00234-0.
Gilman, E. L., J. Ellison, N. C. Duke, and C. Field. 2008. “Threats to Mangroves from Climate Change and
Adaptation Options: A Review.” Aquatic Botany 89: 237–250. doi:10.1016/j.aquabot.2007.12.009.
Giri, S., A. Mukhopadhyay, S. Hazra, S. Mukherjee, D. Roy, S. Ghosh, T. Ghosh, and D. Mitra. 2014.
“A Study of Abundance and Distribution of Mangrove Species in Indian Sundarban Using
20 B. MONDAL ET AL.

Remote Sensing Technique.” Journal of Coastal Conservation 18: 359–367. doi:10.1007/s11852-

014-0322-3.
Ha, T. T. P., H. V. Dijk, and L. Visser. 2014. “Impacts of Changes in Mangrove Forest Management
Practices on Forest Accessibility and Livelihood: A Case Study in Mangrove-Shrimp Farming
System in Ca Mau Province, Mekong Delta, Vietnam.” Land Use Policy 36: 89–101. doi:10.1016/j.
landusepol.2013.07.002.
Hazra, S., T. Ghosh., R. DasGupta, and G. Sen. 2002. “Sea Level and Associated Changes in The
Sundarbans.” Science and Culture 68: 309–321.
Held, A., C. Ticehurst, L. Lymburner, and N. Willams. 2003. “High Resolution Mapping of Tropical
Mangrove Ecosystem Using Hyperspectral and Radar Remote Sensing.” International Journal of
Remote Sensing 24: 2739–2759. doi:10.1080/0143116031000066323.
Heumann, B. W. 2011. “Satellite Remote Sensing of Mangrove Forests: Recent Advances and Future
Opportunities.” Progress in Physical Geography 35: 87–108. doi:10.1177/0309133310385371.
Huang, H., and X. Jia. 2012. “Integrating Remotely Sensed Data, GIS and Expert Knowledge to
Update Object-Based Land Use/Land Cover Information.” International Journal of Remote
Sensing 33: 905–921. doi:10.1080/01431161.2010.536182.
Islam, S. N., and A. Gnauck. 2008. “Mangrove Wetland Ecosystem in Ganges-Brahmaputra Delta in
Bangladesh.” Frontiers of Earth Science in China 2: 439–448. doi:10.1007/s11707-008-0049-2.
Kamal, M., and S. Phinn. 2011. “Hyperspectral Data for Mangrove Species Mapping: A Comparison of
Pixel-Based and Object –Based Approach.” Remote Sensing 3: 2222–2242. doi:10.3390/rs3102222.
Kamal, M., S. Phinn, and J. Kasper. 2015. “Object-Based Approach for Multi-Scale Mangrove
Composition Mapping Using Multi-Resolution Image Datasets.” Remote Sensing 7: 4753–4783.
doi:10.3390/rs70404753.
Kuenzer, C., A. Blumel, S. Gebbardt, T. Quoc, and S. Dech. 2011. “Remote Sensing of Mangrove
Ecosystems: A Review.” Remote Sensing 3: 878–928. doi:10.3390/rs3050878.
Kumar, T., A. Mandal, D. Dutta, R. Nagaraja, and V. K. Dadhwal. 2017. “Discrimination and
Classification of Mangrove Forests Using EO-1 Hyperion Data: A Case Study of Indian
Sundarbans.” Geocarto International. doi:10.1080/10106049.2017.1408699.
Landis, J. R., and G. G. Koch. 1977. “The Measurement of Observer Agreement for Categorical
Data.” Biometrics 33: 159–174. doi:10.2307/2529310.
Lugo, A. E., and S. C. Snedaker. 1974. “The Ecology of Mangroves.” Annual Review of Ecology and
Systematics 5: 39–64. doi:10.1146/annurev.es.05.110174.000351.
Mcfeeters, S. K. 1995. “The Use of the Normalized Difference Water Index (NDWI) in the Delineation
of Open Water Features.” International Journal of Remote Sensing 17: 1425–1432. doi:10.1080/
01431169608948714.
Mondal, B., and A. K. Saha. 2018. “Spatio-Temporal Analysis of Mangrove Loss in Vulnerable Islands
of Sundarban World Heritage Site, India.” In Geospatial Technologies for All. AGILE 2018. Lecture
Notes in Geoinformation and Cartography, edited by A. Mansourian, P. Pilesjö, L. Harrie, and
R. van Lammeren, 93–109. Springer, Cham. doi:10.1007/978-3-319-78208-9_5.
Munoz, S. R., and S. I. Bangdiwala. 1997. “Interpretation of Kappa and B Statistics Measures of
Agreement.” Journal of Applied Statistics 24: 105–111. doi:10.1080/02664769723918.
Nabahungu, N. L., and S. M. Visser. 2011. “Contribution of Wetland Agriculture to Farmers
Livelihood in Rwanda.” Ecological Indicators 71: 4–12. doi:10.1016/j.ecolecon.2011.07.028.
Naskar, K., and R. Mandal. 1999. Ecology and Biodiversity of Indian Mangroves. Delhi: Daya
Publishing House.
Navulur, K. 2007. Multispectral Image Analysis Using the Object-Oriented Paradigm. USA: CRC press
Taylor & Francis Group.
Pervez, W., S. A. Khan, and Valiuddin. 2015. “Hyperspectral Hyperion Imagery Analysis and Its
Application Using Spectral Analysis.” The International Archives of Photogrammetry, Remote
Sensing and Spatial Information Sciences. XL-3/W2, 169–175. doi: 10.5194/isprsarchives-XL
-3-W2-169-2015.
Pramanick, M. K. 2014. “Changes and Status of Mangrove Habitats in Ganges Delta: Case Study in
Indian Part of Sundarban.” International Journal of Geology, Earth, and Environmental Science 4:
109–118. doi:10.4172/2168-9776.1000153.
 INTERNATIONAL JOURNAL OF REMOTE SENSING 21

Schmidt, K. S., and A. K. Skidmore. 2003. “Spectral Discrimination of Vegetation Types in a Coastal
Wetland.” Remote Sensing of Environment 85: 92–108. doi:10.1016/S0034-4257(02)00196-7.
Snedaker, S. C. 1978. “Mangroves: Their Value and Perpetuation.” Nature and Resources 14: 6–13.
Story, M., and R. Congalton. 1986. “Accuracy Assessment: A User’s Perspective.” Photogrammetric
Engineering & Remote Sensing 52: 397–399.
Tomlinson, P. B. 1986. The Botany of Mangroves. Melbourne (Australia): Cambridge University Press.
Wang, L., J. L. Silvan-Cardenas, and W. P. Sousa. 2008. “Neural Network Classification of Mangrove
Species from Multi-Seasonal IKONOS Imagery.” Photogrametric Engineering and Remote Sensing
74: 921–927. doi:10.14358/PERS.74.7.921.
Wang, L., and P. Sousa. 2009. “Distinguishing Mangrove Species with Laboratory Measurements of
Hyperspectral Leaf Reflectance.” International Journal of Remote Sensing 30: 1267–1281.
doi:10.1080/01431160802474014.
Wang, L., W. P. Sousa, and P. Gong. 2004a. “Integration of Object Based and Pixel-Based
Classification for Mapping Mangroves with IKONOS Imagery.” International Journal of Remote
Sensing 25: 5655–5668. doi:10.1080/014311602331291215.
Wang, L., W. P. Sousa, P. Gong, and G. S. Biging. 2004b. “Comparison of IKONOS and QuickBird
Images for Mapping Mangrove Species on the Caribbean Coast of Panama.” Remote Sensing of
Environment 91: 432–440. doi:10.1016/j.rse.2004.04.005.
Wang, T., H. Zhang, H. Lin, and C. Fang. 2016. “Textural-Spectral Feature-Based Species
Classification of Mangroves in Mai Po Nature Reserve from Worldview-3 Imagery.” Remote
Sensing 8: 24. doi:10.3390/rs8010024.
Wood, A. L., J. R. A. Butler, M. Sheaves, and J. Wani. 2013. “Sport Fisheries: Opportunities and
Challenges for Diversifying Coastal Livelihoods in the Pacific.” Marine Policy 42: 305–314.
doi:10.1016/j.marpol.2013.03.005.