Identification of Mangrove Forests Using Multispectral Satellite Imageries

International Journal of Remote Sensing and Earth Sciences Vol. 16 No.
1 June 2019: 63-86
IDENTIFICATION OF MANGROVE FORESTS USING

MULTISPECTRAL SATELLITE IMAGERIES
Anang Dwi Purwanto* and Wikanti Asriningrum
Remote Sensing Applications Center,
National Institute of Aeronautics and Space
*E-mail: anang.dwi@lapan.go.id
Received: 28 January 2019; Revised: 21 August 2019 Approved: 22 August 2019
Abstract. The visual identification of mangrove forests is greatly constrained by combinations of RGB
composite. This research aims to determine the best combination of RGB composite for identifying
mangrove forest in Segara Anakan, Cilacap using the Optimum Index Factor (OIF) method. The OIF
method uses the standard deviation value and correlation coefficient from a combination of three image
bands. The image data comprise Landsat 8 imagery acquired on 30 May 2013, Sentinel 2A imagery
acquired on 18 March 2018 and images from SPOT 6 acquired on 10 January 2015. The results show
that the band composites of 564 (NIR+SWIR+Red) from Landsat 8 and 8a114 (Vegetation Red
Edge+SWIR+Red) from Sentinel 2A are the best RGB composites for identifying mangrove forest, in
addition to those of 341 (Red+NIR+Blue) from SPOT 6. The near-infrared (NIR) and short-wave infrared
(SWIR) bands play an important role in determining mangrove forests. The properties of vegetation are
reflected strongly at the NIR wavelength and the SWIR band is very sensitive to evaporation and the
identification of wetlands.
Keywords: Mangrove, OIF, Landsat 8, Sentinel 2A, SPOT 6, Combinations
1 INTRODUCTION intrusion, storm surges, maintenance of

The existence of mangrove forests in the natural habitat, as areas for the
coastal areas is very important because of spawning and growth of various species
the many benefits they provide for other of fish, shrimp, shellfish, birds and other
ecosystems. Mangrove forests are a fauna, and land-forming. Mangrove forest
common form of coastal vegetation in can also be utilised for conservation,
tropical and sub-tropical coastal areas. education and ecotourism (Setyawan et
They are influenced by sea tides and are al., 2006). Mangrove forest grows well
capable of adapting to brackish around the Segara Anakan lagoon,
environments. Nybakken (1982) stated Cilacap. The calm water supports this
that mangrove forests is a general term location as one of the areas with the
used to describe a variety of tropical greatest mangrove potential in Java
beach communities dominated by several (Ardli, 2008).
species of distinctive trees or shrubs that Remote sensing technology has been
are able to grow in saltwater. In addition, used to identify the distribution of
mangroves grow on sheltered or flat mangrove forests since their presence
beaches, usually alongside islands in inland and in sea transition areas
areas protected from the wind or behind produces a distinctive spectral reflection
coral reefs off sheltered shorelines. Some compared to other objects (Faizal &
of the functions and benefits of mangrove Amran, 2005). The spectral values
forests include their ability to protect derived from remote sensing satellite
coastal stability from abrasion, seawater images can be extracted to provide
63
http://dx.doi.org/10.30536/j.ijereses.2019.v16.a3097 @National Institute of Aeronautics and Space of Indonesia (LAPAN)
Anang Dwi Purwanto and Wikanti Asriningrum
mangrove type object information in the mission generates imagery with a

range of visible and near-infrared (NIR) spectral resolution consisting of 13
spectra (Suwargana, 2008). channels that include visible sensors,
Landsat 8 imagery can be used to NIR and short-wave infrared (SWIR). Its
detect mangrove forests. This type of spatial resolution is regarded as quite
satellite imagery is medium-resolution high, at 10 metres for the red, blue, green
and is an evolution of the mission of the and NIR bands, in addition to 6 bands
previous Landsat 7 (ETM +) satellite. The with a resolution of 20 m, and 3 bands
advantage of Landsat imagery is the with a spatial resolution of 60 m. It has a
existence of an NIR band that can help in field of view of 290 km (Kawamuna et al.,
the identification of mangrove forests. 2017). Sentinel 2A imagery has been
Purwanto et al. (2014) used Landsat 8 widely used in a range of different
imagery to map the mangrove forests in applications, including analysis of land
Segara Anakan, Cilacap, achieving a cover and land-use change, disaster
result accuracy rate of 82.05%. Setiawan mitigation, agriculture, forestry,
et al. (2015), meanwhile, used Landsat 8 environmental monitoring and urban
imagery when mapping the mangrove planning. Wachid et al. (2017) mapped
forest in Muara Gembong, Bekasi. mangrove canopy density using Sentinel
Sentinel-2 involves two remote 2A imagery in Teluk Jor, Lombok Timur
sensing satellites with passive sensors Regency, Indonesia using regression
that were launched in 2015 and 2017 as analysis between field data hemispherical
part of the Copernicus European Space photography and the subjective density
Agency (ESA) programme. The Sentinel method.
2A satellite launched as part of this
Figure 1-1: Map of the research area (source: Spatial Planning Map of
Cilacap District 2011-2031).
64 International Journal of Remote Sensing and Earth Sciences Vol. 16 No. 1 June 2019
Identification Of Mangrove Forest...
The SPOT 6 satellite was launched in bands (band 2 to band 7) on Small Subi
India on 9 September 2012. SPOT 6 is a Island.
continuation of the previous mission This study aims to determine the best
undertaken by the SPOT satellite series band combination for identifying
and is also the first SPOT-generation mangrove forests in Segara Anakan,
satellite to feature a blue spectral Cilacap from multispectral satellite
channel, meaning it can more clearly imageries (Landsat 8, Sentinel 2A and
identify coastlines, sedimentation and SPOT 6) using the OIF method.
coral reefs (Pustekdata, 2015). SPOT 6
imagery has very good potential for 2 MATERIALS AND METHODOLOGY
mangrove mapping. Rudiastuti et al. 2.1 Location and Data
(2018) mapped mangrove canopy density This research was conducted in the
using SPOT 6 imagery in East Lombok, Segara Anakan area, Cilacap Regency,
Indonesia, achieving in excess of 85% Central Java province with the coordinate
accuracy. limitations of 7°37'22"-7°47'37" LS and
According to Marini et al. (2015), 108°45'11"-109°2'54" BT. The satellite
visual interpretation of mangrove using imageries used in this study were
remote sensing satellite imagery can be Landsat 8 Path 121 Row 065 image
achieved through the proper selection of acquisition dated 30 May 2013, Sentinel
RGB composites; however, the process 2A image acquisition dated 18 March
requires a considerable amount of time 2018 and SPOT 6 image acquisition dated
due to the many potential combinations 10 January 2015. Table 2-1 contains a
of the 3 (three) composite bands. Thus, comparison of the band specifications of
the need exists for a fast and easy method Landsat 8, Sentinel 2A and SPOT 6. The
of identifying mangrove forest. One such mangrove distribution data used in this
method is the Optimum Index Factor study were obtained from field
(OIF) which uses statistical calculations measurements taken between 2013 and
to determine the appropriate combination 2015 in Segara Anakan, Cilacap. Segara
of 3 (three) bands in the image that will Anakan itself is a lagoon located on the
yield the optimal and most informative southern coast of Java Island in an area
colour display (Chavez et al., 1982). The that forms part of Cilacap Regency,
higher the OIF value of the selected band Central Java. Nusakambangan Island
combination, the more information (high stands in front of this area and stretches
standard deviation) with less ‘duplication’ for approximately 30 kilometres in an
(low correlation values between bands) it east-west direction, protecting the bay
conveys. To this end, each of the selected from the waves of the Indian Ocean. Most
three-band colour composites can be of the Segara Anakan area comprises of
evaluated for their effectiveness. The OIF lowlands dominated by mud and
method of determining the best RGB waterways such as the estuary of the
composites with which to identify Citandui river, the Donan river and the
mangrove forests has been widely used by Cibereum river, with the coastal areas
researchers. For example, Manoppo et al. overgrown with mangrove forests. The
(2015) employed OIF in their research to calm water gives this area some of the
identify mangrove forest using SPOT 6 highest mangrove potential in Java.
and Landsat 8 imagery in Lingayan Figure 1-1 contains a map of the
Island, while Marini et al. (2015) used 6 mangrove distribution in Segara Anakan,
Cilacap as the research area
International Journal of Remote Sensing and Earth Sciences Vol. 16 No. 1 June 2019 65
Table 2-1: Band specifications of Landsat 8 OLI/TIRS

(https://landsat.usgs.gov/sites/default/files/documents/Landsat8DataUsersHandbook.pdf),
Sentinel 2A (Zhang et al., 2018) and SPOT 6 MS (Jhonnerie, 2015).
LDCM OLI/TIRS SENTINEL 2A SPOT 6 MULTISPECTRAL
Central Central Wavelength Central

Band Band Band Wavelength
Wavelength
Band 1 Coastal/Aerosol, Band 1 Coastal Aerosol, (0.443 Band 1 Blue, (0.485
(0.443 µm), 30 m µm), 60 m µm), 6 m
Band 2 Blue, (0.482 µm), Band 2 Blue, (0.490 µm), 10 m Band 2 Green, (0.560
30 m µm), 6 m
Band 3 Green, (0.561 µm), Band 3 Green, (0.560 µm), 10 m Band 3 Red, (0.660
30 m µm), 6 m
Band 4 Red, (0.655 µm), Band 4 Red, (0.665 µm), 10 m Band 4 NIR, (0.825
30 m µm), 6 m
Band 5 Near-Infrared, Band 5 Vegetation Red Edge,

(0.865 µm), 30 m (0.705 µm), 20 m
Band 6 SWIR 1, (1.609 Band 6 Vegetation Red Edge,

µm), 30 m (0.740 µm), 20 m
Band 7 SWIR 2, (2.201 Band 7 Vegetation Red Edge,

µm), 30 m (0.783 µm), 20 m
Band 8 Pan, (0.590 µm), Band 8 NIR, (0.842), 10 m

15 m
Band 9 Cirrus, (1.373 Band 8A Vegetation Red Edge,

µm), 30 m (0.865 µm), 20 m
Band 10 LWIR 1, (10.895 Band 9 Water Vapour, (0.945

µm), 100 m µm), 60 m
Band 11 LWIR 2, (12.005 Band 10 SWIR-Cirrus, (1.375 µm),

µm), 100 m 60 m
Band 11 SWIR, (1.610 µm), 20 m
Band 12 SWIR, (2.190 µm), 20 m
2.2 Methods the vegetation elements (Pusfatja, 2015).

The number of band combinations We used the OIF method (Chaves et al.,
was determined using the following 1982) to determine the best three-band
equation: composite as follows:
𝑁 𝑁! 3 3
( )=
3 (3! ∗ (𝑁 − 3)!) (2-1) 𝑂𝐼𝐹 = ∑ 𝑆𝐷𝑖 ⁄∑ 𝐴𝐵𝑆(𝐶𝐶𝑗)
𝑖=1 𝑗=1
(2-2)
where:
N = The total number of bands in the map where:
list SDi = Standard deviation of band i
Interpretation of the mangrove ABS =Absolute value of 3-band
forest can be made using an image colour correlation coefficient
composite to sharpen the appearance of
Figure 2-1: Distribution of training samples of mangrove
The identification of mangrove The OIF method uses the total value
forests using a spatial approach was of the standard deviations divided by the
performed by observing the appearance of correlation coefficient of the three
the canopy texture and location selected bands. Before calculating the
(Danoedoro, 2009). It is also based on two OIF value, it is necessary to determine the
important properties, namely the number of three-band combinations from
mangrove forest growing in coastal areas each image using the existing equations.
and the chlorophyll content of leaves that The number of combinations (N) resulting
absorbs the red spectrum and strongly from Landsat 8 imagery stands at 35,
reflects the infrared spectrum (Susilo, Sentinel 2A has 20 combinations and the
2000). Leaves have green characteristics, SPOT 6 imagery has 4 combinations.
where chlorophyll absorbs the spectrum
of red and blue and reflects the green 3 RESULTS AND DISCUSSION
spectrum (Lo, 1996). The intensity of the 3.1 OIF Calculation of Landsat 8 OLI
reflection depends on the wavelength OIF calculations were performed for
used and the three vegetation each of the three-band combinations. The
components, i.e. leaves, substrate and combination of RGB composites
shadow. Leaves reflect weakly at the blue with the highest OIF values represents
and red wavelengths but strongly at the the best combination results. The results
NIR wavelength. The characteristic green of the 35 band combinations for Landsat
colour of leaves means the chlorophyll 8 are shown in Attachment 1. Table 3-1
absorbs the red and blue electromagnetic shows that the correlation coefficient
spectrum and reflects the spectrum of value between band 4 and band 5 is
green radiation. The highest OIF value is 0.494, the
the result between the highest standard correlation coefficient value between
deviation divided by the lowest band 4 and band 6 is 0.571 and the
correlation coefficient of the band correlation coefficient between band 4
combination used. The standard and band 7 is 0.59. The total of these
deviation and correlation coefficients are three correlation coefficient values is
obtained by delineating some regions as relatively small compared to the totals of
areas of mangrove forest (Figure 2-1). the other correlations.
Table 3-1: Correlations between Landsat 8 bands.
Band Band 1 Band 2 Band 3 Band 4 Band 5 Band 6 Band 7
Band 1 1
Band 2 0.983 1
Band 3 0.909 0.922 1
Band 4 0.888 0.917 0.83 1
Band 5 0.715 0.7 0.866 0.494 1
Band 6 0.683 0.676 0.821 0.571 0.884 1
Band 7 0.67 0.665 0.795 0.59 0.829 0.993 1
The highest OIF value resulted from research by Suwargana (2008) which
a combination of bands 4, 5 and 6. That identified mangrove forests with satellite
value is only slightly higher than the imagery using band channels in the
band combinations with the second- range of the visible and NIR spectra. In
highest (bands 2, 5 and 6) and third- addition, the vegetation properties reflect
highest OIF values (bands 1, 5 and 6). strongly at the NIR wavelength and
The 35 band combinations from Landsat absorb the red and blue electromagnetic
8 are shown in Table 3-2. Based on Table spectrum and reflect the green
3-2, the combination of band 4 (Red), electromagnetic spectrum (Lo, 1996).
band 5 (NIR) and band 6 (SWIR) has the According to Winarso and Purwanto
highest OIF value (OIF=2722.284). (2017), remote sensing technology plays
Meanwhile, the combination of band 1 an important role in the context of
(coastal/aerosol), band 2 (blue) and band mangrove forest habitat where wetland
4 (red) has the lowest OIF value conditions absorb electromagnetic waves
(OIF=146.513). The combination of bands in the SWIR spectrum, meaning that the
4, 5 and 6 has a high standard deviation use of the band 6 (SWIR) electromagnetic
value and the correlation coefficient value spectrum is optimal for mangrove
is very low. Thus, the combination of identification. It was also confirmed in
these bands has the highest OIF value. this study that all three of the highest OIF
On the other hand, the combination of values for Landsat 8 use band 5 (NIR) and
bands 1, 2 and 4 has the lowest standard band 6 (SWIR). The RGB composite for
deviation; also, the correlation coefficient natural colour in Landsat 8 imagery is
value is relatively high, which means this 432 where all the bands are in the range
band combination has the lowest OIF of the visible electromagnetic spectrum.
value. OIF values are generated from the Based on the OIF calculations, the
band combinations that are in the range combination of natural colour bands from
of visible, visible + NIR, visible + SWIR, Landsat 8 imagery is ranked 32 with an
visible + NIR + SWIR, and NIR + SWIR OIF value 214.903, meaning it is not
spectrums. The three best OIF values for optimal for use in identifying mangrove
identifying mangrove forest are forest because in principle the natural
dominated by the visible band, NIR band colour composite produces the actual
and SWIR band. This is in line with colour of the object on the earth surface.
Table 3-2: OIF values of Landsat 8.
Sum of
Sum of
Band correlation
standard OIF Rank
combinations coefficient
deviations
of bands
456 5,305.731 1.949 2722.283 1
256 5,270.418 2.26 2332.043 2
156 5,257.317 2.282 2303.819 3
457 4,389.251 1.913 2294.433 4
356 5,422.416 2.571 2109.068 5
567 5,621.112 2.706 2077.277 6
257 4,353.938 2.194 1984.474 7
157 4,340.837 2.214 1960.630 8
145 4,025.456 2.097 1919.626 9
345 4,190.555 2.19 1913.495 10
245 4,038.557 2.111 1913.101 11
357 4,505.936 2.49 1809.612 12
235 4,155.242 2.488 1670.113 13
125 3,990.143 2.398 1663.946 14
135 4,142.141 2.49 1663.510 15
246 1,688.752 1.488 1134.913 16
467 2,039.446 2.154 946.818 17
267 2,004.133 2.334 858.668 18
167 1,991.032 2.346 848.692 19
346 1,840.750 2.222 828.420 20
367 2,156.131 2.609 826.420 21
146 1,675.651 2.142 782.283 22
236 1,805.437 2.419 746.356 23
136 1,792.336 2.413 742.783 24
126 1640.338 2.342 700.400 25
347 924.270 2.215 417.277 26
237 888.957 2.382 373.197 27
137 875.856 2.374 368.936 28
247 772.272 2.172 355.558 29
147 759.171 2.148 353.431 30
127 723.858 2.318 312.276 31
234 573.576 2.669 214.902 32
134 560.475 2.627 213.351 33
123 525.162 2.814 186.624 34
124 408.477 2.788 146.512 35
For example, vegetation appears as bands, the display will be dominated by

green, sea is blue and so on. Natural magenta colours. Meanwhile, if the NIR
colour composite makes it more difficult and SWIR bands are placed in the green
to distinguish between mangrove forest and blue channels, the colours will be
and non-mangrove objects due to the dominated by cyan. The best position is
limitations of the electromagnetic for the NIR and SWIR bands to be placed
spectrum channel used. The NIR and on the red and green channels, which
SWIR bands play an important role in gives the mangrove a reddish colour.
distinguishing mangroves from the However, if the NIR and SWIR bands are
surrounding objects. Both bands affect placed on the green and red channels,
colour visualisation. If the red and blue then the NIR band provides natural
channels are placed by the NIR and SWIR colour information that is green.
Rank 1 - Rank 1 - Rank 1 -

RGB Composite 456 RGB Composite 465 RGB Composite 546

Figure 3-1: Composite results for Bands 4, 5 and 6 with the highest OIF values.
The composite results for bands 4, 5 and shown in a green colour and rice fields
6 are shown in Figure 3-1. are indicated in dark green. Mangrove
The visual interpretations of forest is also identified based on its
mangrove forest in Figure 3-2 reveal that texture and location on the beach
the RGB 564 composite has the best (Susanto & Asriningrum, 2011). The RGB
appearance and is more informative 564 composite was used for
compared to the other RGB composites, interpretation of the mangrove forest,
while RGB 546 shows all vegetation in a which included the three bands in the
magenta colour. The reflectance of NIR is range of the visible and NIR spectrum.
the highest compared to the red band, Figure 3-2 contains a comparison of
thus indicating that the position of the the False Colour and True Colour RGB
NIR band greatly determines the display composites derived from the Landsat 8
of colour. imagery. The false colour is a
The visual appearance of the RGB combination of several bands on the
564 composite is similar to that of the satellite imagery that allows us to obtain
RGB 567 composite because the red and spatial information on the satellite image.
green channels of the two composites use The natural colour, meanwhile, is the
the NIR band and the SWIR 1 band. The actual colour found on the surface of the
blue channel of the RGB 564 composite earth. The natural colour consists of
uses the red band but the RGB 567 band 1 shown in blue, band 2 shown in
composite uses the SWIR 2 band, which green and band 3 shown in red. The RGB
means that water appears differently in 564 composite appears more optimal in
the two composites (water in the RGB 564 terms of distinguishing between
composite is blue, while in the RGB 567 mangrove forest and other objects,
composite it is black). compared to the RGB 432 composite. The
Mangrove forest is displayed with a mangrove forest contrasts with other
reddish-brown colour, deep water is vegetation and land and waters objects.
black, shallow water is blue, vegetation is The reddish colour indicates a relatively
higher intensity of water conditions research conducted by Marini et al.

compared to the lighter colours. The RGB (2015) in Subi Kecil Island, Riau revealed
564 composite, as one of the band RGB composite 573 of a Landsat 8 image
combinations with the highest OIF to have the highest OIF value (Figure 3-
values, indicates the appropriate colour 4). The difference in the RGB composite
for identifying mangrove and non- result from Subi Kecil Island is due to the
mangrove objects. This shows that the difference in land cover.
SWIR 1 band is important for identifying Subi Kecil Island is a small island
mangrove both visually and digitally. The surrounded by shallow water and a
RGB 432 composite shows vegetation in relatively small mangrove area. This
a relatively homogeneous colour, which research area (Segara Anakan, Cilacap)
makes it difficult to distinguish the meanwhile contains a river estuary and is
mangrove forest from other vegetation. also affected by the tidal flow from the
Figure 3-3 indicates the differences in sea. The left side of the research area
appearance between the two RGB contains association mangrove (derris
composites, notably showing the trifoliata and acanthus ilicifolius), while
composition of existing mangrove where the right side is true mangrove (genus of
the RGB 564 composite (false colour) is rhizopora, bruguiera and sonneratia).
able to distinguish between major The central section contains Nypa mixed
mangrove and minor mangrove with Ceriops Tagal.
distribution. In contrast, it is very
difficult to distinguish between major and 3.2 OIF Calculation of Sentinel 2A
minor mangrove distribution in the RGB Identification of mangrove forests
432 composite (natural colour). was also carried out using Sentinel 2A
The results have similarities with imagery. One of the advantages of
the OIF research conducted by Manoppo Sentinel 2A imagery is the relatively large
et al. (2015) where RGB composite 564 number of bands (multispectral). There
from a Landsat 8 image had the highest are a total of 13 bands divided into 3
OIF value for identifying mangrove forests types of spatial resolution (10 m, 20 m
in Lingayan Island, Central Sulawesi and 60 m).
(Figure 3-3). The results of other OIF
1 1 2
2
(a) (b)
Figure 3-2: Comparison of RGB composites, (a) RGB 564 (False Colour),
(b) RGB 432 (Natural Colour).
bands used. The initial experiment was

conducted by selecting bands with a
spatial resolution of 20 m, as previously
carried out by Kawamuna et al. (2017)
who used RGB composite 8a115 to
identify mangrove using Sentinel 2A
images in Pangpang Bay, Banyuwangi.
The other experiment involved choosing
certain bands (up to 6), including band 2
(blue), band 3 (green), band 4 (red), band
5 (vegetation red edge), band 8a
(vegetation red edge) and band 11 (SWIR).
Bands for spectral vegetation are only
Figure 3-3: Identification of mangroves in used in bands 5 and 8a, whereas SWIR
Lingayan Island with RGB 564 Landsat 8
bands used only band 11. The use of 6
(Manoppo et al., 2015).
bands, namely band 2 (blue), band 3
(green), band 4 (red), band 5 (vegetation
red edge), band 8a (vegetation red edge)
and band 11 (SWIR) showed optimal and
effective results and proved capable of
clearly distinguishing mangrove forest
objects.
The amount of duplication between
band combinations can be identified from
the correlation coefficient values of each
band. Table 3-3 shows the correlation
Figure 3-4: Identification of mangroves in Subi between bands for the Sentinel 2A
Kecil Island with RGB 573 Landsat 8 (Marini et imagery. From the calculation of the total
al., 2015). correlation between band 4, band 8a and
OIF calculations for identifying band 11, we can see there is a relatively
mangrove forests using Sentinel 2A low correlation value. A low correlation
images were carried out through two indicates a low duplication of information
experiments related to the selection of from the combination of the bands.
Table 3-3: Correlation between Sentinel 2A bands

Band Band
Band 2 Band 3 Band 4 Band 5
8a 11
Band 2 1
Band 3 0.856 1
Band 4 0.851 0.632 1
Band 5 0.677 0.899 0.473 1
Band 8a 0.607 0.779 0.271 0.755 1
Band 11 0.738 0.834 0.473 0.803 0.906 1
The results of the combination of mangrove forest objects from

the 20 bands from Sentinel 2A are shown surrounding objects, where the mangrove
in Table 3-4. Meanwhile, forests are displayed in a reddish-brown
the results of the RGB composite colour. Similar to the results of the OIF
appearance for each combination are calculations of Landsat 8, the Vegetation
shown in Attachment 2. Red Edge band (band 8a) of Sentinel 2A
Figure 3-5 shows the composite with a mid-wavelength of 0.865 µm
results for the combination of bands 4, 8a greatly affects the colour visualisation
and 11 that have the highest OIF values produced. The best position is band 8a
from Sentinel 2A imagery. The and the SWIR bands (band 11) placed on
combination of bands 4, 8a and 11 led to the red and green channels, which are
the formation of around 6 RGB able to give the mangrove a reddish
composites, namely RGB 48a11, RGB colour. If the red and green channels are
4118a, RGB 8a114, RGB 8a411, RGB placed by band 8a and the SWIR band,
118a4 and RGB 1148a. From the six RGB this produces a more uniform colour
composites, RGB 8a114 (vegetation red which makes it difficult to distinguish the
edge+swir+red) has an excellent display mangroves from their surroundings.
for identifying and distinguishing
Table 3-4: Highest OIF values and ranks of the combinations of Sentinel 2A bands.
Sum of
Sum of
Band correlation
standard OIF Rank
deviations
of bands
4, 8a, 11 1465.62 1.65 888.254 1

4, 5, 8a 1161.662 1.499 774.957 2
3, 4, 8a 1148.058 1.682 682.555 3
2, 8a, 11 1452.512 2.251 645.274 4
2, 4, 8a 1093.397 1.729 632.386 5
5, 8a, 11 1520.777 2.464 617.198 6
3, 8a, 11 1,507.173 2.519 598.321 7
2, 5, 8a 1,148.554 2.039 563.292 8
2, 3, 8a 1134.95 2.242 506.222 9
3, 5, 8a 1,203.215 2.433 494.539 10
3, 4, 11 540.906 1.939 278.961 11
2, 5, 11 541.402 2.218 244.094 12
2, 4, 11 486.245 2.062 235.812 13
3, 5, 11 596.063 2.536 235.040 14
4, 5, 11 554.51 2.536 218.655 15
2, 3, 11 527.798 2.428 217.379 16
3, 4, 5 236.948 2.004 118.237 17
2, 3, 5 223.84 2.432 92.039 18
2, 4, 5 182.287 2.001 91.097 19
2, 3, 4 168.683 2.339 72.117 20

RGB Composite 48a11 RGB Composite 4118a RGB Composite 8a114

RGB Composite 8a411 RGB Composite 118a4 RGB Composite 1148a
Figure 3-5: Composite results for Bands 4, 8a and 11 that have the highest OIF values.
Like the OIF calculation results for blue channel with the input from the
Landsat 8, the OIF calculation results SWIR 1 band produces a colour
from Sentinel 2A also show that the NIR dominated by magenta colours, thus
band dominates the band combination of making it difficult to identify the
RGB composites with the highest OIF mangrove forests.
value. The position of the NIR band from The best RGB composite results
the Sentinel 2A image is in band 8a, obtained by this study (RGB 8a114) differ
where the mean wavelength is 0.86 µm. from the RGB composite 8a115 identified
The NIR band is highly optimal for by Kawamuna et al. (2017), where the
detecting mangrove forest if it is placed difference was found to be in the position
on the Red channel and is supported by of the blue channel placed by the
SWIR band 1 placed on the green Vegetation Red Edge band. In this study,
channel. The results of RGB 8a114 the blue channel is placed by the Red
composite show a clear distinction band. RGB composite 8a114 produces
between the mangrove forest and colours that tend to be reddish, while
surrounding objects, while RGB 8a411 RGB composite 8a115 produces a darker
composite is less optimal because the colour (Figure 3-6).
(a) (b)
Figure 3-6: Comparison of (a) RGB Composite 8a114 and (b) RGB Composite 8a115.
3.2 OIF Calculations for SPOT 6 3 and band 4 is -0.115. So, there will be
In addition to using Landsat 8 and a relatively lower total correlation if the
Sentinel 2A imagery, this research uses combinations used are band 1 and band
imagery from SPOT 6 to identify 4 or band 3 and band 4. The OIF
mangrove forests. Compared to the calculation results are shown in Table 3-
previous images (Landsat 8 and Sentinel 6 and the result of the 4 band
2A), SPOT 6 imagery has the highest combinations for SPOT 6 are shown in
spatial resolution. The images from SPOT Attachment 3. The OIF calculations used
6 have spatial resolutions of 1.5 metres 4 bands: band 1 (blue), band 2 (green),
for panchromatic, 6 metres for band 3 (red) and band 4 (NIR). The OIF
multispectral and can be used for calculation results reveal the
applications in defence, agriculture, combination of band 1 (blue), band 3 (red)
forestry, land cover and land-use change, and band 4 (NIR) to have the highest OIF
coastal, engineering, and the oil, gas and values. Meanwhile, the lowest OIF value
mining industries (Pustekdata, 2015). is a combination of band 1 (blue), band 2
The OIF calculations for identifying (green) and band 3 (red).
mangrove forests used 4 multispectral The combination of SPOT 6’s RGB
bands for SPOT 6 images, namely the 134 image has the highest OIF value.
blue, green, red and NIR bands. A total of This combination has a high standard
4 RGB combinations were obtained from deviation and a low correlation
the four bands, namely combinations of coefficient, thus indicating high diversity
RGB 123, RGB 124, RGB 134 and RGB with low duplication of information. The
234. Table 3-5 shows that the correlation correlation values show the magnitude of
between band 1 and band 4 is 0.015. linear relationships between two
Meanwhile, the correlation between band variables.
Table 3-5: Correlation between SPOT 6 bands.

Band Band 1 Band 2 Band 3 Band 4
Band 1 1
Band 2 0.706 1
Band 3 0.95 0.651 1
Band 4 0.015 0.629 0.115 1
Table 3-6: Highest OIF values and ranks of the combinations for SPOT 6.
Sum of
Sum of
Band correlation
standard OIF Rank
deviations
of bands
1, 3, 4 346.814 0.85 408.016 1

2, 3, 4 354.838 1.165 304.581 2
1, 2, 4 354.823 1.35 262.831 3
1, 2, 3 68.837 2.307 29.838 4


Figure 3-7: Composite results for Bands 1, 3 and 4 that have the highest OIF values.
Figure 3-7 illustrates how mangrove mangrove substrates (Winarso &

forests are more easily identified using Purwanto, 2017).
the RGB 341 combination where the
mangrove forest appears as a dark green 3.3 Comparison of the OIF Values for
and can be distinguished from the Landsat 8, Sentinel 2A and SPOT 6
surrounding objects. The RGB 143 Imagery
combination produces greenish colours Figure 3-8 shows a comparison of
for mangrove forest objects but it is very the best RGB composites derived from
difficult to distinguish objects around the Landsat 8, Sentinel 2A and SPOT 6
mangroves, with one example being how imagery for identifying the distribution of
land and water objects appear in almost mangrove forests. The RGB 564
the same colour. The identification of composite from Landsat 8 and RGB
mangrove forests using SPOT 6 imagery 8a114 composite from Sentinel 2A are
still requires additional elements of relatively similar in terms of the
association interpretation, often defined appearance of mangrove forests. RGB
as the relationship between one object 341 composite from SPOT 6, meanwhile,
and another object, such that one object is less optimal for detecting mangrove
provides a clue as to the other objects. forest. This shows that the existence of
The association elements used to identify the NIR and SWIR 1 band is highly
mangrove forests include the presence of optimal in the detection of mangrove
marine waters, inland waters (rivers), forest. The SPOT 6 imagery does not have
river estuaries, settlements, land an SWIR band, which means it is less
vegetation and others. SPOT 6 imagery optimal in distinguishing mangrove
does not have a SWIR band, thus forests, especially in Segara Anakan,
meaning it is not optimal to use in Cilacap and requires additional key
identifying mangrove forests compared to interpretative associations in order to
Sentinel 2A and SPOT 6 images. SWIR identify objects around the mangrove
bands are very sensitive to terrestrial forest.
objects, which here are inundated
(a)
(b)
(c)
Figure 3-8: Results of the identification of mangrove forests: (a) Landsat 8 imagery (RGB 564), (b)
Sentinel 2A imagery (RGB 8a114), (c) SPOT 6 imagery (RGB 341).
Figure 3-9: Correlation and standard deviation between Landsat 8 bands.
Figure 3-9 shows the band between band 4, band 5 and band 7 is
correlation and total standard deviation 1,913. The total standard deviation of
of Landsat 8 images. From the Landsat 8 band 4, band 5 and band 6 is 5,305,731.
imagery, the correlation between band 4 In the same way, the total standard
and band 5, band 6 and band 7 is very deviation of band 4, band 5 and band 7 is
low. Likewise, the correlation between 4,389,251 (Figure 3-6a). The combination
band 6 and band 7 is relatively low. A list of band 4, band 5 and band 6 has an OIF
of combinations involving band 4 and value higher than the combination of
band 5 contains two combinations, band 4, band 5 and band 7. This is due
comprising one of band 4, band 5 and to the fact that the total standard
band 6, and the other of band 4, band 5 deviation of band 4, band 5 and band 6 is
and band 7. The total correlation between higher than the total of standard
band 4, band 5 and band 6 is 1,949. On deviation of band 4, band 5 and band 7.
the other hand, the total correlation
Figure 3-10: Correlation and standard deviation between bands of Sentinel 2A.
Figure 3-10 shows that the lowest deviation leads to a higher OIF value for
correlation for the Sentinel 2A imagery is the combination of band 4, band 5 and
between band 4 and band 8a. This band 8a than that of the combination of
combination can be formed using two band 4, band 5 and band 8.
band combinations (a combination of Figure 3-11 shows that the lowest
band 4, band 8a and band 11, plus a correlations for the SPOT 6 imagery are
combination of band 4, band 5 and band those between band 1 and band 4, and
8a). The total of the correlation between band 3 and band 4. This is attained using
band 4, band 8a and band 11 is 1.65. On the three-band combination of band 1,
the other hand, the total of the correlation band 3 and band 4, where the
between band 4, band 5 and band 8a is combination of the three bands produces
1.499. The total standard deviation for a total correlation of 0.85 and a total
band 4, band 8a and band 11 is 1465.62. standard deviation of 408.016. This also
Meanwhile, the total standard deviation has the highest OIF value compared to
for band 4, band 5 and band 8 is the other band combinations. Table 3-7
1161.662. The higher total standard shows a comparison between the RGB
composites for mangrove detection in this for identifying mangrove are the red
study with those of previous studies. It channel (0.865 µm), green channel (1,610
can be seen that the optimal wavelength µm) and blue channel (0.650 – 0.820 µm).
values that provide the best information
Figure 3-11: Correlation and standard deviation of SPOT 6 bands
Table 3-7: Comparison of RGB composites for mangrove detection

Studies Red Green Blue Location
Landsat 8
Manoppo et Band 5 Band 6 Band 4

Lingayan Island, Central Sulawesi
al. (2015) (~0.865 µm) (~1.610 µm) (~0.650 µm)
Marini et al. Band 5 Band 7 Band 3

Subi Kecil Island, Riau
(2015) (~0.865 µm) (~2.200 µm) (~0.565 µm)
This study Band 5 Band 6 Band 4

Segara Anakan, Central Java
(2019) (~0.865 µm) (~1.610 µm) (~0.650 µm)
Sentinel 2A
Kawamuna Band 8a Band 11 Band 5

Pangpang Bay, Banyuwangi
et al. (2017) (~0.865 µm) (~1.610 µm) (~0.705 µm)
This study Band 8a Band 11 Band 4

Segara Anakan, Central Java
(2019) (~0.865 µm) (~1.610 µm) (~0.820 µm)
CONCLUSION Gaol and Mr Syarif Budhiman for their

The results have shown that the reviews of this paper and for providing
band composite of 564 (NIR+SWIR+Red) many suggestions for this research.
from Landsat 8 imagery and the band
composite of 8a114 (Vegetation Red REFERENCES
Edge+SWIR+Red) from Sentinel 2A Ardli, E. R. (2008). A trophic flow model of the
imagery are the best RGB composites for Segara Anakan lagoon. Cilacap,
identifying mangrove forest. In the same Indonesia. Dissertation, University Of
way, the band composite of 341 Bremen.
(Red+NIR+Blue) from SPOT 6 imagery is Chavez, P. S., Berlin, G. L., & Sowers, L. B.
also the best colour composite (R-G-B) for (1982). Statistical method for selecting
identifying mangrove forest in Segara Landsat MSS ratios. Journal of Applied
Anakan, Cilacap. Photographic Engineering, 8(1), 23-30.
The NIR band with a mean value of Danoedoro, P. (2009). Penginderaan Jauh
0.86 m plays an important role in the Untuk Inventarisasi Mangrove: Potensi,
determination of mangrove forests. The Keterbatasan Dan Kebutuhan Data.
RGB composite from the Landsat 8 and Proceedings Sinergi Survei dan Pemetaan
Sentinel 2A imagery shows that the red Nasional dalam Mendukung Pengelolaan
channel should be placed by the NIR Mangrove Berkelanjutan 2009, 98-113.
band (0.86 m), the green channel is Faizal, A., & Amran, M. A. (2005). Model
placed by the SWIR 1 band (1.6 m) and Transformasi Indeks Vegetasi yang Efektif
the blue channel is placed by the red untuk Prediksi Kerapatan Mangrove
band (0.65 m–0.8 m). The SWIR band Rhizophora Mucronata. Proceedings
is important for determining the RGB MAPIN XIV 2005.
composites to use in identifying Jhonnerie, R. (2015). Klasifikasi Mangrove
mangrove forests as it is very sensitive to Berbasis Objek Dan Piksel Menggunakan
evaporation and the identification of Citra Satelit Multispektral Di Sungai
wetlands. Kembung, Bengkalis, Provinsi Riau.
The RGB composites of images Dissertation, Bogor Agricultural
developed from Landsat 8 and Sentinel University.
2A imagery are capable of distinguishing Kawamuna, A., Suprayogi, A., & Wijaya, A. P.
mangrove forest from the surrounding (2017). Analisis Kesehatan Hutan
objects more clearly, although the Mangrove Berdasarkan Metode
composite images from SPOT 6 imagery Klasifikasi Ndvi Pada Citra Sentinel-2
still require additional association (Studi Kasus:Teluk Pangpang Kabupaten
elements in order to identify mangrove Banyuwangi). Journal of Geodesy, 6(1),
objects. 277-284.
Lo, C. P. (1996). Penginderaan Jauh Terapan.
ACKNOWLEDGEMENTS Jakarta.
We would like to thank the Remote Manoppo, A., Anggraini, N., & Marini, Y. (2015).
Sensing Application Center of LAPAN that Identifikasi Mangrove Dengan Metode
funded and facilitated this research. We Optimum Index Factor (OIF) Pada Data
are also grateful to the Remote Sensing SPOT 6 dan Landsat 8 di Pulau Lingayan.
Technology and Data Center of LAPAN for Proceedings MAPIN XX 2015.
providing access to SPOT 6 data, as well Marini, Y., Manoppo, A., & Anggraini N. (2015).
as to the European Space Agency (ESA) Teknik Penentuan Komposit Warna RGB
for the access to Sentinel 2A data. Special untuk Identifikasi Mangrove di Pulau
thanks also go to Mr Jonson Lumban
Subi Kecil Menggunakan Data Landsat 8. Susanto., & Asriningrum W. (2011).

Buku Buka Rampai Mangrove, Jakarta. Penginderaan Jauh Dengan Nilai Indeks
Nybakken, J. W. (1982). Marine biology: An Faktor Untuk Identifikasi Mangrove di
ecological approach. Jakarta. Batam (Studi Kasus Gugusan Pulau
Purwanto, A. D., Asriningrum, W., Winarso, G., Jandaberhias). J Berita Dirgantara, 12(3),
& Parwati, E. (2014). Analisis Sebaran 104-109.
dan Kerapatan Mangrove Menggunakan Susilo, S. B. (2000). Penginderaan Jauh
Citra Landsat 8 di Segara Anakan, Terapan. Bogor.
Cilacap. Proceedings Sinasja 2014. Suwargana, N. (2008). Analisis Perubahan
Pusfatja (2015). Pedoman Pengolahan Data Hutan Mangrove Menggunakan Data
Penginderaan Jauh Landsat 8 Untuk Penginderaan Jauh di Pantai Bahagia,
Mangrove, Jakarta. Muara Gembong, Bekasi. Journal of
Pustekdata (2015). Katalog Citra Satelit Resolusi Remote Sensing and Digital Image
Tinggi SPOT-6. Retrieved from Processing, 5, 64-74.
https://inderaja-catalog.lapan.go.id. Wachid, M. N., Hapsara, R. P., Cahyo R. D. et al.
Accessed 23 Apr 2018. (2017). Mangrove canopy density analysis
Rudiastuti, A. W., Yuwono, D. M., & Hartini, S. using Sentinel-2A imagery satellite data.
(2018). Mangrove mapping using SPOT 6 IOP Conf. Series: Earth and Environmental
at East Lombok Indonesia. IOP Conf. Science, 70(2017), 012020, doi:
Series: Earth and Environmental Science, 10.1088/1755-1315/70/1/012020
165(2018), 012005. doi: 10.1088/1755- Winarso, G., & Purwanto, A. D. (2017).
1315/165/1/012005 Evaluation of mangrove damage level
Setiawan, K. T., Purwanto A. D., & Kurniawan, based on Landsat 8 image. International
B. D. (2015). Inventarisasi Hutan Journal of Remote Sensing and Earth
Mangrove Menggunakan Data Landsat di Sciences, 11 (2), 105–116.
Pantai Utara Bekasi Jawa Barat. Zhang, T. X., Su, J. Y., Liu, C. J. et al. (2018).
Proceedings MAPIN-XX 2015. Potential bands of Sentinel-2A satellite for
Setyawan, A. D., & Winarno, K. (2006). classification problems in precision
Pemanfaatan Langsung Ekosistem agriculture. International Journal of
Mangrove di Jawa Tengah dan Automation and Computing, 16(1) doi:
Penggunaan Lahan di Sekitarnya; 10.1007/s11633-018-1143-x
Kerusakan dan Upaya Restorasinya.
Journal of Biodiversitas, 7(3), 282-291.
Attachment 1: RGB Composites of Landsat 8 Imagery
RGB 123 - Rank 34 RGB 124 - Rank 35 RGB 125 - Rank 14 RGB 126 - Rank 25
RGB 457 - Rank 4 RGB 467 - Rank 17 RGB 567 - Rank 6
Attachment 2: RGB Composites of Sentinel 2A Imagery
RGB 234 - Rank 20 RGB 235 - Rank 18 RGB 238a - Rank 9 RGB 2311 - Rank 16
RGB 245 - Rank 19 RGB 248a - Rank 5 RGB 2411 - Rank 13 RGB 258a - Rank 8
RGB 2511 - Rank 12 RGB 28a11 - Rank 4 RGB 345 - Rank 17 RGB 348a - Rank 3
RGB 3411 - Rank 11 RGB 358a - Rank 10 RGB 3511 - Rank 14 RGB 38a11 - Rank 7
RGB 458a - Rank 2 RGB 4511 - Rank 15 RGB 48a11 - Rank 1 RGB 58a11 - Rank 6
Attachment 3: RGB Composites of SPOT 6 Imagery
RGB 123 - Rank 4 RGB 124 - Rank 3
RGB 134 - Rank 1 RGB 234 - Rank 2

Identification of Mangrove Forests Using Multispectral Satellite Imageries

Uploaded by

Copyright:

Available Formats

Identification of Mangrove Forests Using Multispectral Satellite Imageries

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Identification of Mangrove Forests Using Multispectral Satellite Imageries

Uploaded by

Copyright:

Available Formats

International Journal of Remote Sensing and Earth Sciences Vol. 16 No.

1 June 2019: 63-86

IDENTIFICATION OF MANGROVE FORESTS USING

Keywords: Mangrove, OIF, Landsat 8, Sentinel 2A, SPOT 6, Combinations

1 INTRODUCTION intrusion, storm surges, maintenance of

mangrove type object information in the mission generates imagery with a

Table 2-1: Band specifications of Landsat 8 OLI/TIRS

LDCM OLI/TIRS SENTINEL 2A SPOT 6 MULTISPECTRAL

Central Central Wavelength Central

Band 5 Near-Infrared, Band 5 Vegetation Red Edge,

Band 6 SWIR 1, (1.609 Band 6 Vegetation Red Edge,

Band 7 SWIR 2, (2.201 Band 7 Vegetation Red Edge,

Band 8 Pan, (0.590 µm), Band 8 NIR, (0.842), 10 m

Band 9 Cirrus, (1.373 Band 8A Vegetation Red Edge,

Band 10 LWIR 1, (10.895 Band 9 Water Vapour, (0.945

Band 11 LWIR 2, (12.005 Band 10 SWIR-Cirrus, (1.375 µm),

Band 11 SWIR, (1.610 µm), 20 m

Band 12 SWIR, (2.190 µm), 20 m

2.2 Methods the vegetation elements (Pusfatja, 2015).

Figure 2-1: Distribution of training samples of mangrove

Band 3 0.909 0.922 1

Band 4 0.888 0.917 0.83 1

Band 5 0.715 0.7 0.866 0.494 1

Band 6 0.683 0.676 0.821 0.571 0.884 1

Band 7 0.67 0.665 0.795 0.59 0.829 0.993 1

For example, vegetation appears as bands, the display will be dominated by

Rank 1 - Rank 1 - Rank 1 -

Rank 1 - Rank 1 - Rank 1 -

higher intensity of water conditions research conducted by Marini et al.

bands used. The initial experiment was

Table 3-3: Correlation between Sentinel 2A bands

The results of the combination of mangrove forest objects from

4, 8a, 11 1465.62 1.65 888.254 1

Rank 1 - Rank 1 - Rank 1 -

Rank 1 - Rank 1 - Rank 1 -

Table 3-5: Correlation between SPOT 6 bands.

Band 3 0.95 0.651 1

Band 4 0.015 0.629 0.115 1

1, 3, 4 346.814 0.85 408.016 1

Rank 1 - Rank 1 - Rank 1 -

Rank 1 - Rank 1 - Rank 1 -

Figure 3-7 illustrates how mangrove mangrove substrates (Winarso &

Figure 3-9: Correlation and standard deviation between Landsat 8 bands.

Figure 3-11: Correlation and standard deviation of SPOT 6 bands

Table 3-7: Comparison of RGB composites for mangrove detection

Manoppo et Band 5 Band 6 Band 4

Marini et al. Band 5 Band 7 Band 3

This study Band 5 Band 6 Band 4

Kawamuna Band 8a Band 11 Band 5

This study Band 8a Band 11 Band 4

CONCLUSION Gaol and Mr Syarif Budhiman for their

Subi Kecil Menggunakan Data Landsat 8. Susanto., & Asriningrum W. (2011).

Attachment 1: RGB Composites of Landsat 8 Imagery

RGB 457 - Rank 4 RGB 467 - Rank 17 RGB 567 - Rank 6

Attachment 2: RGB Composites of Sentinel 2A Imagery

Attachment 3: RGB Composites of SPOT 6 Imagery

RGB 123 - Rank 4 RGB 124 - Rank 3

RGB 134 - Rank 1 RGB 234 - Rank 2