See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/308466536

Study of mangrove communities in Marine National Park and Sanctuary,

Jamnagar, Gujarat, India, by fusing RISAT-1 SAR and Resourcesat-2 LISS-IV
images

Article in International Journal of Image and Data Fusion · September 2016

DOI: 10.1080/19479832.2016.1232755

CITATIONS READS

3 856

4 authors, including:

Mohit Kumar Himanshi Chauhan

Integrated Research and Action for Development (IRADe) Sharda University
24 PUBLICATIONS 130 CITATIONS 12 PUBLICATIONS 134 CITATIONS

SEE PROFILE SEE PROFILE

Ajai Prof
Space Application Center
188 PUBLICATIONS 2,235 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mohit Kumar on 28 September 2017.

The user has requested enhancement of the downloaded file.

 International Journal of Image and Data Fusion

ISSN: 1947-9832 (Print) 1947-9824 (Online) Journal homepage: http://www.tandfonline.com/loi/tidf20

Study of mangrove communities in Marine

National Park and Sanctuary, Jamnagar, Gujarat,
India, by fusing RISAT-1 SAR and Resourcesat-2
LISS-IV images

Mohit Kumar, H. B. Chauhan, A. S. Rajawat & Ajai

To cite this article: Mohit Kumar, H. B. Chauhan, A. S. Rajawat & Ajai (2016): Study of mangrove
communities in Marine National Park and Sanctuary, Jamnagar, Gujarat, India, by fusing
RISAT-1 SAR and Resourcesat-2 LISS-IV images, International Journal of Image and Data Fusion,
DOI: 10.1080/19479832.2016.1232755

To link to this article: http://dx.doi.org/10.1080/19479832.2016.1232755

Published online: 22 Sep 2016.

Submit your article to this journal

View related articles

View Crossmark data

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

http://www.tandfonline.com/action/journalInformation?journalCode=tidf20

Download by: [14.96.200.19] Date: 22 September 2016, At: 10:05

INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION, 2016
http://dx.doi.org/10.1080/19479832.2016.1232755

RESEARCH ARTICLE

Study of mangrove communities in Marine National Park and

Sanctuary, Jamnagar, Gujarat, India, by fusing RISAT-1 SAR
and Resourcesat-2 LISS-IV images
Mohit Kumara,b,c*, H. B. Chauhanb, A. S. Rajawatb and Ajaid
a
Institute of Science, Nirma University, Ahmedabad, India; bGeo Sciences Division, Space Applications
Centre, Ahmedabad, India; cIntegrated Research and Action for Development (IRADe), New Delhi, India;
d
Space Applications Centre, Ahmedabad, India

ABSTRACT ARTICLE HISTORY

In this study, RISAT-1 (Radar Imaging Satellite-1) HH image has been Received 28 January 2016
fused with Resourcesat 2 LISS-IV (Linear Imaging Self Scanning-IV) Accepted 25 August 2016
image to study the mangrove communities of Jindra-Chhad island KEYWORDS
complex, in Marine National Park and Sanctuary (MNP&S), Jamnagar, Mangroves; community
Gujarat, India. Three different methods were used to fuse RISAT-1 and zonation; Marine National
LISS-IV images. In one case, the Synthetic Aperture Radar (SAR) data Park & Sanctuary; image
was simply integrated as an additional band to the three bands of fusion; class separability
LISS-IV data, whereas in the other, Intensity-Hue-Saturation method analysis
was used to merge the two data sets. In yet another exercise, the
vegetative and sedimentary parts were separated from the optical
data by computing normalised difference vegetation index (NDVI)
and by averaging the Red and Green bands, respectively. These two
layers, viz., NDVI and the average of Red and Green bands, were then
integrated with the SAR data. All the merged products were put to
supervised classification using maximum likelihood algorithm into
following seven classes: mangrove communities (Avicennia Dense,
Avicennia Sparse, Rhizophora-Ceriops Dense), Intertidal Mudflat,
Hightidal Mudflat, Sand and Sea. The class separability analysis indi-
cated that the separability obtained among the classes in the case of
fused products was higher than that obtained when both the data
sets were classified individually.

1. Introduction
Mangroves are salt-tolerant plants that occur along intertidal zones in the form of
narrow strips or as extensive patches in estuarine habitats and river deltas of tropical
and sub-tropical regions (Ajai et al. 2013). They are important ecologically as well as
economically. From the ecological point of view, they act as windbreaker, prevent the
hinterland from storms and heavy tides, provide stability to the coastline by reducing
erosion and serve as a nursery for a diversity of marine and coastal organisms. From the
economical perspective, they constitute an important source of fuel, wood, medicine
and fodder, especially for the human beings dwelling near the coast.

CONTACT Mohit Kumar [email protected] Nirma University, Ahmedabad, India

*Present Address: C-80, Shivalik, Malviya Nagar, New Delhi-110017, India
© 2016 Informa UK Limited, trading as Taylor & Francis Group
2 M. KUMAR ET AL.

In spite of their ecological and economical significance, mangroves today are cate-
gorised as an endangered ecosystem. Major threats to them are transformation of
mangrove habitats into agricultural landscapes and industrial development zones,
human habitations, construction of salt pans, shrimp farming and reduced freshwater
availability due to construction of dams in the upland catchment areas (Valiela et al.
2001, Shah et al. 2007, Kumar et al. 2012, Ajai et al. 2013). Due to their unsustainable
utilisation, it has been felt necessary to implement protection and conservation mea-
sures, which in turn require a constant monitoring and surveillance plan. Visiting the
intertidal areas on foot to study mangroves, at regular intervals, is highly resource
demanding, in particular, in terms of time, money and labour. Satellite images, on the
other hand, offer synoptic, multi-spectral, multi-polarisation and frequent coverage of
such areas in a cost-effective manner.
A number of studies have been carried out using satellite data to study various
aspects of mangrove ecosystems throughout the globe (Bahuguna et al. 2007,
Panigrahi and Parihar 1986, Nayak et al. 1989, 2003, Roy et al. 1991, Gang and
Agatsiva 1992, Porwal et al. 1992, Spalding et al. 1997, Ramachandran et al. 1998,
Nayak and Bahuguna 2001, Proisy and Mougin 2001, Blasco and Aizpuru 2002, Shah
et al. 2005, Lucas et al. 2007, Giri et al. 2011, Kumar et al. 2012, Ajai et al. 2013,
Chakraborty et al. 2013, Chauhan et al. 2013, Cornforth et al. 2013). These studies
have highlighted the differentiation of mangroves from terrestrial vegetation, the
extent and density of mangroves, impact of coastal processes and anthropogenic
activities on mangroves, changes occurring within the mangrove ecosystems over a
period of time, etc. With the advancement in digital image processing techniques,
there have been attempts to understand the spatial relationship between mangroves
and their immediate environment through ‘community zonation of mangroves’
(Bahuguna and Nayak 1996, Blasco et al. 1998, Ajai et al. 2013). These communities
are made up of either one mangrove genus/species or an association of few man-
grove genera/species. Such mangrove community zones display distinct spectral
behaviour, and the potential of optical (Indian Remote Sensing Satellite) data in
discriminating the mangrove communities has been successfully demonstrated by
Nayak and Bahuguna (2001).
Surprisingly, there have been rather limited studies attempting community zonation
of mangroves, using microwave data alone or employing a combination of microwave
and optical data. In fact, there are very few studies characterising Indian mangrove
ecosystems using Synthetic Aperture Radar (SAR) data (Chakraborty et al. 2013). For
Indian researchers, lack of an indigenous radar sensor which could provide microwave
data at relatively lesser cost than what is available internationally might be one of the
reasons which restricted them from employing SAR data in an extensive manner.
However, launch of RISAT-1 (Radar Imaging Satellite-1) on 26 April 2012 by Indian
Space Research Organisation (ISRO) has offered new, exciting and cost-effective oppor-
tunities to conduct different experiments using SAR data to understand natural ecosys-
tems. In this study, the potential of C-band HH RISAT-1 MRS (Medium Resolution
ScanSAR mode) data has been explored, in conjunction with Resourcesat-2 LISS-IV
(Linear Imaging Self Scanning-IV) data, for identifying mangrove communities in Gulf
of Kachchh MarineNational Park and Sanctuary (MNP&S), situated in Jamnagar, Gujarat,
India.
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 3

2. Objectives
● To explore the potentials of RISAT-1 SAR data for mangrove community zonation.
● To exploit synergistic capabilities of RISAT-1 SAR and Resourcesat-2 LISS-IV data for
discrimination of mangrove communities.

3. Materials and methods

3.1. Study area
The MNP&S is located along the southern shore of Gulf of Kachchh in the Jamnagar district of
Gujarat, India (between 22° 15ʹN to 23° 00ʹN latitudes and 69° 00ʹ E to 70° 30ʹE longitudes).
MNP&S is endowed with ecologically sensitive habitats such as mangroves, mudflats, coral
reefs, seagrasses and sand dunes. It was established by a set of Government notifications
during the period 1980–1982. Presently, the entire notified protected area comprises of
457.92km2 of Marine Sanctuary and 162.89 km2 of Marine National Park. The MNP&S supports
215 species of molluscs including oysters, three species of sea turtles, three species of marine
mammals (Dolphin, Porpoise and Dugong), 144 different varieties of fishes, 27 species of
commercially important Prawns, 49 species of hard corals, 10 species of soft corals, 100 species
of algae and six species of seagrasses along with few mangrove species(Singh 2002, Singh et al.
2006, Satyanarayana and Ramakrishna 2009). Since 1991, mangroves and coral reefs have
been provided extra protection under the 1991 Coastal Regulation Zone notification of
Government of India.
The study was carried out in Jindra-Chhad island complex which is situated in one of
the core areas of MNP&S (Figure 1). Three major mangrove communities have been
identified here: Avicennia dense, Rhizophora-Ceriops dense and Avicennia sparse (Nayak
and Bahuguna 2001, Nayak et al. 2003). Avicennia dense is found in the seaward fringing
areas having frequent tidal influence, followed by Avicennia sparse towards the interior
of the island complex. Rhizophora-Ceriops community was restricted to relatively shel-
tered areas along the creek (Nayak et al. 2003).

3.2. Data used

RISAT-1 C-band MRS HH data of 24 September 2012 was used along with Resourcesat-2
LISS-IV data of 25 December 2011. The detailed specifications of satellite data used are
given in Tables 1 and 2.

4. Methodology
4.1 Pre-processing of RISAT-1 data
The backscattering received from different features being imaged in a microwave
remote sensing system is modified by a randomly distributed pattern of constructive
and destructive interference, known as speckle. For better analysis, the speckle content
in the microwave data should be reduced before doing any further processing. In the
present study, Gamma-MAP filter was used with 5 × 5 window size to reduce the speckle
4 M. KUMAR ET AL.

Figure 1. The study area.

Table 1. Specifications of RISAT-1 data.

Product type L2 Georeferenced
SAR band, polarization C-band, HH
Beam mode MRS (Medium Resolution ScanSAR mode)
Sampled pixel spacing 18 × 18
Number of looks (range × azimuth) 2×l
Incidence angle (degree) 36.548
Pass direction Descending
Map projection UTM

Table 2. Resourcesat-2 LISS-IV data specifications.

Spatial resolution (m) 5.8
Spectral resolution (nm) Band 2: 520–590
Band 3: 620–680
Band 4: 770–860
Swath (km) 70
Temporal resolution (days) 24
Map projection UTM

content in the RISAT-1 data. Then the DN values were converted into, more meaningful,
sigma nought values using the following equation (Padia and Mehra 2013):



σo ¼ 20log10 ðDNÞ  Kdb þ10log10 sin ip =sinðic Þ (1)
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 5

where
DN = digital number per pixel,
Kdb = callibration constant in dB,
ip = incidence angle for pixel position p,
ic = incidence angle at the scene centre.

4.2 Pre-processing of LISS-IV data

The DN values of LISS-IV image were converted into apparent reflectance values using
the method described in Nayak et al. (2003). Apparent reflectance is the ratio of
upwelling radiance recorded at the sensor to the solar irradiance and takes into
calculation the solar elevation at the time of acquiring the image. It is the reflectance
received at the top of the atmosphere and does not include the corrections for atmo-
spheric attenuation of radiation. Though atmospheric corrections are necessary for
studying the actual reflectance of target objects, it is not always possible to acquire
scene-specific atmospheric parameters and field-derived reflectance values, which are
important pre-requisites to do the atmospheric corrections of optical images. In such
cases, researchers have used apparent reflectance values to compare the behaviour of
target objects in different optical regions of electromagnetic spectrum (Nayak et al.
2003, Rahman et al. 2013).

4.3 Co-registration, subset extraction, image fusion and classification

The sigma nought and apparent reflectance images were then co-registered. The subset
of Jindra-Chhad island complex was extracted from both the data sets. These subsets,
one from RISAT-1 data and the other from LISS-IV data, were then separately classified
into following categories using the maximum likelihood algorithm (MLA): Avicennia
Dense (AD), Avicennia Sparse (AS), Rhizophora-Ceriops Dense (RCD), Intertidal Mudflat
(IM), Hightidal Mudflat (HM), Sand and Sea.
The subsets of the study area extracted from the two data sets, RISAT-1 and LISS-IV,
were also fused together to explore the synergistic potential of the SAR data with optical
data, for discriminating different mangrove communities in the study region. Three
different approaches were used to fuse the SAR and optical data:

(i) The integration approach: In this case, SAR data was integrated as an additional
band to the three bands of LISS-IV data. Then the three, 3-band combinations
employing SAR data, viz., SAR + Red + Green, SAR + NIR + Red and SAR + NIR +
Green, were put to supervised classification using the MLA into the same seven
categories, viz., AD, AS, RCD, IM, HM, Sand and Sea. In addition, the only four-
band combination, viz., Green + Red + NIR + SAR, was also classified in the
similar manner.
(ii) The IHS approach: Intensity-Hue-Saturation (IHS) is one of the most widely
used techniques to merge two different remote sensing images. The IHS
transformation is effective in separating the spatial (I) and spectral compo-
nents (HS) from an RGB (Red-Blue-Green) image (Pohl and Van Genderen
1998). After separating an RGB image into IHS components, one of the
6 M. KUMAR ET AL.

components, i.e. either I or H or S, is substituted with another image (which

is to be fused) and the resultant image is then again transformed into RGB
colour system. Mostly, the spatial component (Intensity, I) is replaced by
another image having better spatial resolution. In the present study, all the
three components, i.e. I, H and S, retrieved from LISS-IV data were system-
atically replaced by RISAT-1 data, and the resulting images were then
transformed into RGB colour space. Thus, the three possible combinations
generated were I + H+SAR, H + S+SAR and I+SAR+S. All these three
combinations were then classified using MLA into the same seven cate-
gories described earlier.
(iii) The band ratio fusion approach: Ratios of different bands of optical data
have been employed to study the vegetative as well as sedimentary envir-
onments over the years. In particular, for mangroves, this method has been
found to be more useful for separating different communities than other
approaches employing optical data (Shah et al. 2005). In our study, two
optical band ratios, viz., NDVI (Normalised Difference Vegetation Index) and
OB (Optical Brightness) were generated from LISS-IV data, and these ratios
were then integrated with RISAT-1 data. NDVI was computed by taking the
ratio of (NIR – Red) to (NIR + Red), whereas for obtaining OB, averaging of
Red and Green bands was done. Thus, a three-band integrated image was
generated comprising NDVI, OB and SAR, as individual bands. This combi-
nation was also classified to seven mangrove habitat categories described
earlier, using MLA.

A flow chart depicting the entire methodology is shown in Figure 2.

Figure 2. Methodology.
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 7

4.4 Evaluation
In the way described earlier, 10 images were classified into seven mangrove habitat
categories. Among the 10 images, eight were fusion products of SAR and optical images.
Remaining two images belong to LISS-IV and RISAT-1. A class separability analysis was
carried out, for each of the 10 classified images using Transformed Divergence (TD) matrix.
In the TD method, a covariance weighted distance between the mean values of two
categories is compared to assess the separability between them. The class separability values
may range from 0 to 2000. A separability value above 1700 indicates fairly good separability
between the two categories. On the other hand, any value below 1500 is an indication of
poor separability between the classes. Apart from the quantitative evaluation described
earlier, the images were also visually inspected with a view to get enhanced information
regarding the mangrove environment in general and mangrove communities in particular.

5. Results and discussions

5.1. Classification
Nayak and Bahuguna (2001), Nayak et al. (2003), Shah et al. (2005) and Ajai et al. (2013)
have used IRS (Indian Remote Sensing Satellite) LISS-III data to classify mangroves at
community level. Nayak et al. (2003) compiled a set of signatures using satellite data for
various mangrove communities along the different stretches of Indian coast. These
signatures are extremely valuable for studying the spectral behaviour of mangrove
communities using visible/infra-red data. Different aspects of foliar components of
plants govern the spectral response received from a canopy for different regions of
electromagnetic spectrum. In visible region, the spectral response is because of pig-
ments present inside the cell; in NIR (Near Infra-Red) region, the spectral behaviour is
affected by the anatomy of leaves and how the leaves are stacked one above the other;
and in MIR/SWIR (Middle Infra-Red/Short Wave Infra-Red) the foliar moisture content is
responsible for the characteristic spectral response (Nayak et al. 2003). For sparse
mangroves (<40% canopy closure), the reflectance from the muddy substrata also
adds to the overall response received at the sensor.
Three major mangrove communities/species have been identified in the study area,
viz., Avicennia Dense, Avicennia Sparse and Rhizophora-Ceriops Dense (Nayak et al. 2003).
Avicennia zone (dense having >40% canopy closure and sparse having <40% canopy
closure) consists of three species, viz., A. marina, A. officinalis and A. alba (Nayak et al.
2003). Rhizophora-Ceriops zone comprises of Rhizophora mucronata and Ceriops tagal,
with the latter being the dominant species (Shah et al. 2005). For LISS-III, NIR and MIR
bands have been found to be extremely helpful in classifying mangroves at community
level (Nayak et al. 2003). But for LISS-IV, we have only three bands (Table 2), which
restricts the spectral response studies up to NIR band. Published literatures (Nayak et al.
2003, Shah et al. 2005) were used to select pixels for signature generation of three
dominant mangrove communities. It was observed that ‘Avicennia Dense’ [comprising A.
marina, A. officinalis and A.alba (Nayak et al. 2003)] has highest values in NIR region,
followed by ‘Rhizophora-Ceriops Dense’. ‘Avicennia sparse’ recorded highest values in
Green and Red channels, but lowest in NIR (Figure 3). After signature generation, the
LISS-IV data was classified into following categories using the MLA: Avicennia Dense
8 M. KUMAR ET AL.

0.25

0.2

0.15 Avicennia Dense

Avicennia Sparse
Rhizophora-Ceriops Dense
0.1

0.05

0
1 2 3

Figure 3. Spectral behaviour of major Mangrove Communities of Jindra-Chhad island complex,

observed in LISS-IV data [the values on the Y-axis (vertically) are the apparent reflectance values
and those on X-axis (horizontally) are the bands of LISS-IV (1 = Green, 2 = Red, and 3 = NIR).

(AD), Avicennia Sparse (AS), Rhizophora-Ceriops Dense (RCD), Intertidal Mudflat (IM),
Hightidal Mudflat (HM), Sand and Sea (Figure 4).
As the RISAT-1 data (specifications mentioned in Table 1) was georegistered with
LISS-IV data (Figure 2), the AOI (Area of Interest) polygons, used to generate the
signatures for different categories of classification on LISS-IV data, could be employed
to study the backscatter values of different communities on RISAT-1 data. Thus, the
same AOI polygons were used to obtain signatures from RISAT-1 data for various
categories under study, which were used to derive signatures from LISS-IV data. It was
observed that the three mangrove communities indeed have distinct behaviour in SAR
data as represented by the mean backscatter (σ°) plot (Figure 5). The classified image of
RISAT-1 data along with its sigma nought image is provided in Figure 6.
The two data sets, viz., LISS-IV and RISAT-1, were fused together using three different
approaches (Figures 7–9). The various products generated by fusing LISS-IV and RISAT-1
were also classified using MLA and the results have been evaluated in Section 5.2.

Avicennia Dense Avicennia Sparse Rhizophora Ceriops Dense Intertidal Mudflat

Hightidal Mudflat Sand Sea

Figure 4. Jindra-Chhad island complex, as observed in LISS-IV data [NIR (R), Red (G) and Green (B)]
(left) and its classified image (right) [R = Red, G = Green and B = Blue].
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 9

0
Mean Backscatter Value (db)

–2

–4

–6 Avicennia Dense
Avicennia Sparse
Rhizophora-Ceriops Dense
–8

–10

–12

–14

Figure 5. Backscatter response (σ°) (expressed in db and depicted on Y-axis) of major mangrove
communities of Jindra-Chhad island complex as observed in RISAT-1 data.

Avicennia Dense Avicennia Sparse Rhizophora Ceriops Dense Intertidal Mudflat

Hightidal Mudflat Sand Sea

Figure 6. Jindra-Chhad island complex, as observed in RISAT-1 (left) and its classified image (right).

5.2. Evaluation
5.2.1. Qualitative evaluation
Qualitatively the images were visually inspected with a view to identify the different categories
of mangrove environment. Visual interpretation of LISS-IV and RISAT-1 data is described first,
which will help in comparative visual evaluation of fused images described later.
In the LISS-IV data, mangroves could be segregated into dense and sparse through visual
analysis. Dense mangroves (having closed canopies) appear bright red and sparse man-
groves (with open canopies) appear pale red in an FCC (NIR, Red and Green displayed as
RGB) shown in Figure 4. However, mangrove communities could not be identified through
visual inspection alone. Among the non-mangrove categories, intertidal mudflat can be
separated from the high tidal mudflat, as the former appears dark (due to high moisture
10 M. KUMAR ET AL.

content) relative to latter (Figure 4). Sand appears white whereas sea water appears blue
(Figure 4), thus making it quite easy to distinguish them from the rest.
The visual distinction among various categories becomes difficult while interpreting RISAT-
1 HH data. Often, there is a confusion between mangrove and mudflat categories (Figure 6).
The differentiation between sea, sand and mudflat is also very poor (Figure 6). Identification of
mangrove communities through visual inspection of RISAT-1 data alone seems not possible.
Integrating RISAT-1 as an additional band to LISS-IV bands offers exciting per-
spectives (Figure 7). In particular, an FCC of SAR, Red and Green displayed in RGB,
respectively, highlights the various mangrove communities effectively (Figure 8).
Dense mangroves can also be separated from sparse mangroves. High tidal mudflat
(bright blue) and intertidal mudflat (dark blue) can also be discriminated. However,
sand could not be separated from mudflat on the basis of colour only as both
appear blue in colour.
Integrating RISAT-1 with NIR and Red (Figure 7), and with NIR and Green (Figure 7)
does not help in discriminating mangrove communities visually, but dense and sparse
mangroves can be demarcated easily. Conclusively, integrating RISAT-1 with Red and
Green bands of LISS-IV provides an FCC which appears visually better for our purpose of
mangrove community zonation. This may be because C-band SAR data closely resem-
bles NIR band of optical data in terms of radiometry, image contrast and discernible
terrain features (Kushwaha et al. 2000). Therefore, using RISAT-1 with Red and Green
bands of LISS-IV data provides more contrast variation and therefore highlights the
different communities effectively.
Among the IHS-related combinations, merging RISAT-1 SAR with intensity (I) and satura-
tion (S) derived from LISS-IV data (Figure 10) offered better discrimination among mangrove
habitat categories than merging SAR with, intensity (I) and hue (H) or hue (H) and saturation

SAR(R)+NIR(G)+Red(B) SAR(R)+Red(G)+Green(B) SAR(R)+NIR(G)+Green(B)

Avicennia Dense Avicennia Sparse Rhizophora Ceriops Dense Intertidal Mudflat
Hightidal Mudflat Sand Sea

Figure 7. The three 3-band combinations of RISAT-1 and LISS-IV (top) and their respective classified
images (bottom) [R = Red, G = Green and B = Blue].
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 11

HM
AD

RCD
SW

RCD

IM
AS

Figure 8. Mangrove habitat of Jindra-Chhad island complex, as observed in an FCC prepared by

integrating RISAT-1 with Red and Green bands of LISS-IV [SAR(R), Red (G) and Green (B)]
(AD = Avicennia Dense, AS = Avicennia sparse, RCD = Rhizophora-Ceriops Dense, IM = Intertidal
Mudflat, HM = Hightidal Mudflat, SW = Sea Water).

I+H+SAR H+S+SAR I+S+SAR

Avicennia Dense Avicennia Sparse Rhizophora Ceriops Dense Intertidal Mudflat
Hightidal Mudflat Sand Sea

Figure 9. Combined products of LISS-IV and RISAT-1 using IHS approach (top), and their classified
images (bottom).

(S) (Figure 9). Though the image generated by merging RISAT-1 with I and S could not
provide visual discrimination among mangrove communities, it offered better differentia-
tion among rest of the mangrove habitat categories such as 'dense and sparse mangroves'
and 'sand and mudflat'.
12 M. KUMAR ET AL.

Sand
Mudflat

MD

MS

Figure 10. Mangrove habitat of Jindra-Chhad island complex, as observed in an image prepared by
merging RISAT-1 with intensity and saturation derived from LISS-IV data (MD = Mangrove Dense,
MS = Mangrove Sparse).

In the image prepared by layerstacking RISAT-1 with band ratios obtained from LISS-
IV, dense mangroves appear bright red whereas sparse mangroves appear in magenta/
bluish magenta colour (Figure 11). However, still it is difficult to visually discriminate
mangrove communities in this fused product. The visual differentiation among sedi-
mentary substrates is also not sharp.

5.2.2. Quantitative evaluation

As the objective of the entire exercise was to study separability/discrimination among
the seven categories of classification used (in particular the three mangrove commu-
nities), the quantitative evaluation of the images was done through a class separability

NDVI(R)+OB(G)+SAR(B)
Avicennia Dense Avicennia Sparse Rhizophora Ceriops Dense Intertidal Mudflat
Hightidal Mudflat Sand Sea

Figure 11. Combined product of LISS-IV and RISAT-1 using band ratio approach (left), and its
classified image (right) [R = Red, G = Green and B = Blue].
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 13

analysis. For this purpose, Transformed Divergence (TD) algorithm was employed. The
advantages associated with the TD algorithm relative to similar other algorithms are as
follows:

● It takes into account the mean, variance and covariance of the classes while
calculating the inter-class spectral distance whereas other methods (such as
Euclidean Distance) do not.
● It represents the inter-class spectral distance on a transformed scale to enable
comparisons from different such exercises.
● This method can work on multiple bands simultaneously, whereas other measures
such as M-Statistics can work with only one band at a time.

The idea was to study the performance of a single classifier, in this case MLA, when
different inputs are provided to it.
All the classified images obtained from the classification of RISAT-1, LISS-IV as well as
the different combinations of RISAT-1 and LISS-IV were put to class separability analysis
using TD measure. As mentioned previously, the TD analysis provides separability values
ranging from 0 to 2000. A separability value above 1700 indicates fairly good separ-
ability between the categories. On the other hand, any value below 1500 is an indication
of poor separability between the classes.
This evaluation exercise was performed on the samples of the seven classification
categories (parametric signatures) selected for training the classifier.

5.2.2.1 TD evaluation of parametric signatures. A parametric signature comprises a

set of statistical parameters (such as mean and co-variance matrix) of the pixels. This
information is then used to train the classifier for grouping the image pixels into desired
set of clusters/classes. The information set comprising the parametric signature includes
the following:

● the number of bands in the input image;

● the minimum and maximum data value for each band, for each sample;
● the mean data value for each band, for each sample;
● the covariance matrix for each sample;
● the number of pixels in the sample.

These statistics of parametric signatures were used to compute the TD matrix to

evaluate the separability among the various classification categories. A graph showing
the comparison of class separability values obtained for various images (individual as
well as merged) is shown in Figure 12.
By looking at the graph, it can be inferred that addition of RISAT-1 significantly
improves the class separability. All the combinations employing optical and SAR data
have shown increased class separability compared to using either LISS-IV or RISAT-1
alone. In addition, when all the four channels (two visible, one NIR and one microwave)
are integrated and used together, the separability obtained is highest. This inference
suggests that microwave data, in particular the C-band HH data, can be used
14 M. KUMAR ET AL.

2000 1902.61 1895.29 1884.3 1868.59 1868.15 1848.14 1835.45 1829.27 1777.11
1800
1600
1395.13
1400
1200
1000
800
600
400
200
0
AR

en

ed

AR

en

L4

1-
SA

SA

SA

AT
re

+R

re
+S

+S
+G

+G
B+

n+

n+

IS
IR
IR

ue

R
io

io
+O
IR

ed
+N
+N

+H
at

at
+N

+R
VI

ur

ur
ed

ity
SA
D
R

at

at

R
+R

ns
N
SA

SA
+S

+S
te
en

ty

ue
In
re

si

H
n
G

te
In

Figure 12. Class separability values (from 0 to 2000, depicted on vertical Y-axis) for different images
(represented on horizontal X-axis) used in the study for all the seven categories.

synergistically with optical data, in the form of an additional band, at least for discrimi-
nation of coastal features.
Now, as the prime objective was to evaluate the discrimination among the three
mangrove communities of the study area, the TD matrix was also computed for
studying the separability among the three mangrove communities. Figure 13 shows
the class separability values obtained employing different images for AD, AS
and RCD.
The graph shown has decreased separability values than that obtained when all the
seven classification categories were used for all the images studied. This is expected as
the separability is bound to reduce in case of mangrove communities which behave
spectrally similar than the other categories such as mudflat, sand and sea water.
However, the trend is identical to what was observed in Figure 12. The class discrimina-
tion among the three mangrove categories is significantly improved when RISAT-1 data
is used synergistically toLISS-IV data. The same trend is observed for AD and AS also
(Figure 14). However, the trend is little altered in case of AD and RCD (Figure 15) and for
AS and RCD (Figure 16). Though the separation has increased by employing both
microwave and optical data, in the case of AD and RCD (Figure 15) L4 data alone has
provided the second best separability value. In the case of AS and RCD (Figure 16),
RISAT-1 has scored slightly better than LISS-IV in terms of separability value. One
possible reason for this could be the more sensitive nature of RISAT-1 to the canopy
structure of mangroves relative to LISS-IV.
Among the IHS-related combinations, in all the TD analyses, combining intensity
and saturation obtained from LISS-IV data with RISAT-1 data provided more separ-
ability than combining either hue and saturation, or, intensity and hue with RISAT-1
data.
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 15

2000
1860.88 1850.41 1834.72 1812.27 1811.64
1800 1783.21 1765.43 1756.11
1681.59
1600
1400 1304.18
1200
1000
800
600
400
200
0
R

en

ed

AR

en

L4

-1
SA

SA
A

SA

AT
re

+R

re
+S
+S

+G

+G
B+

n+

n+

S
IR

ue
IR

RI
io

tio
+O
IR

ed
+N
+N

+H
at

ra
+N

+R
VI

ur
ed

ity

tu
SA
ND

at
R

R
Sa
ns
+R

SA

SA
+S

te

e+
en

ity

In

Hu
re

ns
G

te
In

Figure 13. Class separability values (from 0 to 2000, depicted on vertical Y-axis) for different images
(represented on horizontal X-axis) used in the study for the three mangrove categories, viz. AD, AS
and RCD.

2000
1807.16 1792.04 1774.35
1800 1745.1 1744.87 1725.53 1692.79
1654.59
1562.09
1600
1400

1200 1117.69

1000

800

600
400

200

0
en

ed

AR

AR

en

L4

-1
AR

SA

SA

AT
re

+R

re
+S

S
+S

+G

G
B+

n+

n+

IS
IR

ue

+
IR

R
io

io
+O
IR

ed
+N

+H
at

at
+N

+N

+R
VI

ur

ur
ty
ed

SA
D
R

at

at

R
si
N
SA

SA
+S

+S
en
+R

t
ity

ue
In
en

ns

H
re

te
G

In

Figure 14. Class separability values (from 0 to 2000, depicted on vertical Y-axis) for different
images (represented on horizontal X-axis) used in the study for the two mangrove categories,
viz. AD and AS.
16 M. KUMAR ET AL.

2000

1800 1739.66
1687.76 1684.12 1679.55 1676.08
1636.51 1634.49
1591.64
1600 1545.04

1400

1200 1145.85

1000

800

600
400

200

0
AR

L4

en

en

ed

AR

-1
SA

SA

SA

AT
re

re

+R

+S
+S

+G

+G
B+

n+

n+

IS
IR

ue
IR

R
io

io
+O
IR

ed

+N

+H
+N

at

at
+N

+R
VI

ur

ur
ity
ed

SA
D
R

at

at
R

ns
N
SA

SA
+R

+S

+S
te
ity

ue
en

In
ns

H
re

te
G

In

Figure 15. Class separability values (from 0 to 2000, depicted on Y-axis) for different images
(represented on X-axis) used in the study for the two mangrove categories, viz. AD and RCD.

2000
1848.13 1834.48 1829.43 1825.48 1824.26
1802.28 1797.57
1800 1693.19
1569.74 1539.15
1600

1400

1200

1000

800

600

400

200

0
AR

en

ed

AR

en

-1

L4
SA

SA

SA

AT
re

+R

re
+S

+S
+G

+G
B+

n+

n+

IS
IR
IR

ue

R
io

io
O
IR

ed
+N
+N

+H
at

at
+
+N

+R
VI

ur

ur
ed

ity
SA
D
R

at

at

R
+R

ns
N
SA

SA
+S

+S

te
en

ity

ue

In
re

ns

H
G

te
In

Figure 16. Class separability values (from 0 to 2000, depicted on Y-axis) for different images
(represented on X-axis) used in the study for the two mangrove categories, viz. AS and RCD.
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 17

Also worth mentioning is that the band ratio approach yielded better result than IHS-
associated approaches; however, simple integration of LISS-IV bands with RISAT-1
provided the highest class separability value.

6. Conclusions
The study demonstrated that the synergistic use of RISAT-1 C-band MRS and
Resourcesat-2 LISS-IV data improves mangrove community discrimination. The man-
grove communities discriminated are Avicennia Dense, Avicennia Sparse and
Rhizophora-Ceriops Dense in the MarineNational Park, Jamnagar, Gujarat. Among
the different approaches of merging SAR with optical data, maximum separability
among mangrove community classes could be obtained by integrating SAR data
with Red, Green and NIR bands of optical data. RISAT-1 SAR data due to its
sensitivity to the canopy geometry and heterogeneity provides detailed information
regarding the composition of mangrove communities. The approach adopted in this
study should be applied to test community zonation in other more diverse man-
grove areas.

Acknowledgments
Authors are grateful to Shri A.S. Kiran Kumar, former Director, Space Applications Centre (SAC),
Ahmedabad, and current chairman, Indian Space Research Organization (ISRO) and Dr J.S. Parihar,
former Deputy Director, Earth, Ocean, Atmosphere, Planetary Sciences and Applications Area
(EPSA), SAC, Ahmedabad for providing opportunity, institutional support and overall guidance
for this work. The first author (MK) also acknowledges Director, SAC, Ahmedabad, for providing the
Junior Research Fellowship (JRF) as well as satellite data sets.

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

Funding
The first author (MK) also acknowledges Director, SAC, Ahmedabad, for providing the Junior
Research Fellowship (JRF).

References
Ajai, et al., 2013. Mangrove inventory of India at community level. National Academy Science
Letters, 36 (1), 67–77. doi:10.1007/s40009-012-0087-x
Bahuguna, A. and Nayak, S., 1996. Mangrove community discrimination using IRS-1C data, In:
Proceedings of the National Symposiumon Remote Sensing for Natural Resources with specialem-
phasis to Water Management, Indian Society of Remote Sensing and NNRMS, Bangalore. Pune:
Microsoft Technoprint, 311–319.
Bahuguna, A., et al., 2007. Coastal habitat of selected Marine protected areas: Atlas of India.
Ahmedabad: Space Applications Centre, SAC/RESIPA/MESG/PR/59/2007.
Blasco, F. and Aizpuru, M., 2002. Mangroves alongthe coastal stretch of bay of Bengal: present-
status. Indian Journal of Marine Science, 31 (1), 9–20.
18 M. KUMAR ET AL.

Blasco, F., et al., 1998. Recent advances in mangrove studies using remote sensing data. Marine
and Freshwater Researcher, 49 (4), 287–296. doi:10.1071/MF97153
Chakraborty, M., et al. 2013. Initial results using RISAT-1 C-band SAR data. Current Science, 104 (4),
490–501.
Chauhan, H.B., et al., 2013. Synergistic application of radar and optical data for studying the coastal
habitats: A case study in Gulf of Kachchh, Gujarat, India. RISAT UTILISATION PROGRAMME: Status
Report, SAC/EPSA/ATGDG/ATD/RISAT-UP/03/2013, LU12-LU27. Ahmedabad: Space Applications
Centre.
Cornforth, W.A., et al., 2013. Advanced land observing satellite phased array type L-band SAR
(ALOS PALSAR) to inform the conservation of mangroves: Sunderbans as a case study.
RemoteSensing, 5 (1), 224–237.
Gang, P.O. and Agatsiva, J.L., 1992. The current status of mangroves along the Kenyan coast: a case
study of Mida creekmangroves based on remote sensing. Hydrobiologia, 247, 29–36.
doi:10.1007/BF00008202
Giri, C., et al., 2011. Status and distribution of mangrove forests of the world using earth observation
satellite data. Global Ecology and Biogeography, 20, 154–159. doi:10.1111/geb.2011.20.issue-1
Kumar, M., et al., 2012. Application of remote sensing and GIS techniques in understanding
changes in mangrove cover in parts of Indus delta around Kori Creek, Gujarat, India. Journal
of Environmental Research and Development, 7 (1A), 504–511.
Kushwaha, S.P.S., Dwivedi, R.S., and Rao, B.R.M., 2000. Evaluation of various digital image proces-
sing techniques for detection of coastal wetlands using ERS-1 SAR data. International Journal of
Remote Sensing, 21 (3), 565–579. doi:10.1080/014311600210759
Lucas, R.M., et al., 2007. The potential of L-band SAR for quantifying mangrove characteristics and
change: case studies from the tropics. Aquatic Conservation: Marine and Freshwater Ecosystems,
17, 245–264. doi:10.1002/(ISSN)1099-0755
Nayak, S. and Bahuguna, A., 2001. Application of remote sensing data to monitormangroves and
other coastal vegetation of India. Indian Journal of Marine Sciences, 30 (4), 195–213.
Nayak, S., et al., 2003. Community zonation of selected mangrove habitats of India using satellite
data. Ahmedabad: Space Applications Centre, Scientific Note, SAC/RESA/MWRG/MSCED/SN/17/
2003.
Nayak, S.R., et al., 1989. Application of satellite data for monitoring degradation of tidal wetlands
of the Gulf of Kachchh, Western India. Acta Astronautica, 20, 171–178. doi:10.1016/0094-5765
(89)90066-0
Padia, K. and Mehra, R., 2013. RISAT-1 data product format. Ahmedabad, Gujarat: Space
Applications Centre (SAC), Indian Space Research Organization (ISRO).
Panigrahi, S. and Parihar, J.S., 1986. Optimum spectral bands for identification of mangrove vegeta-
tion, results from joint indo-soviet remote sensing experiment. Bangalore: ISRO, TERRA, Special
Publication, ISRO-SP-17-86.
Pohl, C. and Van Genderen, J.L., 1998. Review article Multisensor image fusion in remote sensing:
concepts, methods and applications. International Journal of Remote Sensing, 19 (5), 823–854.
doi:10.1080/014311698215748
Porwal, M.C., et al., 1992. Importance of middle infrared band for classifying mangrove vegetation
and plantation in Middle Andaman Islands. In: Proceedings of National Symposium on Remote
Sensing for Sustainable Development, Lucknow, 30–36.
Proisy, C. and Mougin, E., 2001. Radar remote sensing of mangroves: results and perspectives. In:
IEEE 2001 International Geoscience and Remote Sensing Symposium, 2001. IGARSS '01, Vol. 7,
Sydney, NSW, 3056–3058. doi:10.1109/IGARSS.2001.978253
Rahman, M.M., et al., 2013. Comparison of Landsat image classification methods for detecting
mangrove forests in Sundarbans. International Journal of Remote Sensing, 34 (4), 1041–1056.
doi:10.1080/01431161.2012.717181
Ramachandran, S., et al., 1998. Application of remote sensing and GIS to coastal wetland ecology
of Tamil Nadu and Andaman and Nicobar group of islands with special reference to mangroves.
Current Science, 75, 236–244.
 INTERNATIONAL JOURNAL OF IMAGE AND DATA FUSION 19

Roy, P.S., et al., 1991. Tropical forest typo mapping and monitoring using remote sensing.
International Journal of Remote Sensing, 12, 2205–2225. doi:10.1080/01431169108955253
Satyanarayana, C. and Ramakrishna, 2009. Handbook on hard corals of Gulf of Kachchh. Kolkata:
Zoological Survey of India.
Shah, A.A., Kasawani, I., and Kamaruzaman, J., 2007. Degradation of Indus delta mangroves in
Pakistan. International Journal of Geology, 1 (3), 27–34.
Shah, D.G., et al., 2005. Zoning and monitoring dominant mangrove communities of a part of the
Marine National Park, Gulf of Kachchh. Journal of the Indian Society of RemoteSensing, 33 (1),
155–163. doi:10.1007/BF02990005
Singh, H.S., 2002. Protected areas in India: Status of coastal wetlands and their conservation in India.
Gandhinagar: GEER Foundation.
Singh, H.S., et al., 2006. An ecological and socio-economic study in Marine National Park and
Sanctuary in the Gulf of Kutch (A comprehensive study on biodiversity and management issues).
Gandhinagar: GEER Foundation.
Spalding, M., Blasco, F., and Field, C., 1997. World mangrove atlas. Okinawa, Japan: The
International Society for Mangrove Ecosystems.
Valiela, I., Bowen, J.L., and York, J.K., 2001. Mangrove forests: one of the world’s threatened major
tropical environments. Bioscience, 51, 807–518. doi:10.1641/0006-3568(2001)051[0807:
MFOOTW]2.0.CO;2

View publication stats