03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

Boletim de Poluição Marinha

Volume 62, Edição 3 ,Março de 2011, Páginas 564-572

Mapeamento do habitat de macrobentos em uma planície

de maré usando dados de sensoriamento remoto e um
modelo probabilístico baseado em SIG
https://doi-org.ez278.periodicos.capes.gov.br/10.1016/j.marpolbul.2010.11.028
Obtenha direitos e conteúdo

Acesso ao texto completo

Resumo
Este artigo propõe e testa um método para produzir mapas de potencial de habitat de macrofauna com base em um
modelo de pesos de evidência (uma abordagem probabilística) para a planície de maré de Hwangdo, Coreia.
Amostras de macrobentos foram coletadas durante o trabalho de campo, e consideramos cinco espécies de moluscos
para o mapeamento do habitat. Um modelo de pesos de evidência foi usado para calcular os pesos relativos de 10
fatores de controle que afetam o habitat do macrobentos. Os fatores de controle foram compilados como um banco
de dados espacial a partir de dados de sensoriamento remoto combinados com análise de SIG. O peso relativo de
cada fator foi integrado como um índice de potencial de espécie (SPI), que produziu mapas de potencial de habitat.
Os mapas foram comparados com os locais de habitat pesquisados, revelando uma forte correlação entre os mapas
de potencial e os locais das espécies. A combinação de um modelo de pesos de evidência com base em SIG e técnicas
de sensoriamento remoto é um método eficaz para determinar áreas de potencial de habitat de macrobentos em um
ambiente de planície de maré.

Destaques da pesquisa
► O sensoriamento remoto é eficaz para mapear o ambiente bentônico em planícies de maré. ► Ruditapes
philippinarum é dominante e ocorre principalmente em planícies de areia em baixas elevações. ► O habitat de
macrobentos na região costeira é influenciado por vários ambientes bentônicos. ► O modelo de pesos de evidência
baseado em SIG é ideal para o mapeamento de habitats de macrofauna.

Previous Next

Palavras-chave
Mapeamento de habitats; Planície de maré; Moluscos; Pesos de evidência; Sistema de informação geográfica (SIG);
Sensoriamento remoto

1. Introduction
Macrofaunal invertebrates are known to be a crucial barometer of the ecological values of coastal regions, and their
habitat is strongly influenced by the benthic environment (Lee and Park, 1998, Yap et al., 2003). The tidal flats that
are widely distributed along the west coast of the Korean Peninsula have undergone continuous change in

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 1/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

sedimentary environment during recent decades, due to both natural causes and human activities such as land
reclamation (Ryu et al., 2004, Woo and Je, 2002). In such a setting, identification of changes in the macrobenthos
distribution is an important task for the conservation or rehabilitation of the economical and ecological properties of
the tidal flats.

Many studies have examined the relations between the macrobenthos distribution and relevant controlling factors
in coastal regions. For example, Koo et al. (2005) investigated the burrow architecture of macrofauna in a tidal flat
with the aim of understanding the influences of temperature on the species distribution, and Yoo et al. (2007)
discussed the critical environmental factors that control the patterns of species distribution in a tidal flat. Van der
Wal et al. (2008) used remote sensing techniques to examine the variables that influence the spatial distribution of
macrobenthos in a tidal flat. Mapping of the habitat in coastal areas has also been attempted using models based on
a geographic information system (GIS). Jorgensen and Kollmann (2009) employed GIS for mapping the distribution
of an invasive shrub that has a negative effect on biodiversity in dune ecosystems, based on airborne remote sensing.
Hatten and Parsley (2009) established a spatially explicit GIS model for the habitat of the white sturgeon Acipenser
transmontanus, based on water depth, riverbed slope and roughness, and fish locations determined in the field.
Fulton et al. (2010) sought to predict the location of a mussel habitat by establishing correlations between mussel
counts and hydraulics, using a GIS-based numerical model.

Few studies on macrobenthos distribution have attempted quantitative estimations of the relationships among
benthic environments that affect the macrobenthos distribution in a tidal flat. The mapping of biological
distributions in a tidal flat, based on the quantitative relations between the habitats and control factors, is important
for the effective management of ecosystems and for the evaluation of ecological properties. In this study, we
investigate the applicability of a GIS-based weights-of-evidence model, combined with remote sensing data and
field observations, to the potential habitat mapping of macrofauna in a tidal flat. The weights-of-evidence, a
probabilistic model, has been widely used for event potential mapping based on the spatial relationships between
event occurrence and controlling factors (Dahal et al., 2008, Lee and Pradhan, 2006, Oh et al., 2009, Regmi et al.,
2010, Yalcin, 2008). We estimate quantitatively the spatial relationships between controlling factors that influence
the mollusca habitat and the actual existence of the species, using the weights-of-evidence model. The relationships
are then integrated to generate mollusca habitat potential maps.

The study was performed in the Hwangdo tidal flat, Cheonsu Bay, located on the central western coast of Korea.
Cheonsu Bay is characterized by a shallow water depth of less than 25 m and is bordered by the Anmyeondo District
of Taean County and the Kanweoldo District of Seosan City (Fig. 1a). The total area of the bay covered by tidewater
was originally 380 km2, but this was reduced to 180 km2 following the construction of embankments during the
Seosan A and B Land Reclamation Exercise in 1984, the Hongseong–Boryeong reclamation in 1991, and the
development of the freshwater Bunam and Kanweol lakes (So et al., 1998). The water depth within Cheonsu Bay has
been maintained at over 10 m since construction of the embankments. A large intertidal area remains around the
coastline.

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 2/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

Download: Download full-size image

Fig. 1. (a) Landsat ETM + image of Cheonsu Bay and the Hwangdo tidal flat, acquired on 23 October 2006. (b) IKONOS
RGB (432 bands) composite image of the Hwangdo tidal flat acquired on 26 February 2001, overlain with 28 sample
sites at which the macrofauna was acquired in May 2009.

The tides in the study area are semi-diurnal, with a mean tidal range of 4.59 m (spring tide = 6.33 m, neap
tide = 2.86 m). The maximum tidal-current velocities in the main tidal channel are approximately 1.00 ms–1 during
flood tide and approximately 0.70 ms–1 during ebb tide (Kim and Kim, 1996). Sand dunes located in the
northwestern part of the bay are extensively exposed during ebb tide (Lee and Park, 1998). The central Hwangdo
tidal flat is characterized by complex channels. The tidal flat is 1.65 km wide and 5.15 km long (Fig. 1b), and consists
of the following sedimentary facies (from the high tide waterline to the low tide waterline): mud flats, mixed flats,
and sand flats. The mud flat facies is dominant in a relatively elevated area with a steep slope, whereas the sand flat
facies is located in a lower area with a gentle slope and upwardly concave relief (Choi et al., 2010).

2. Materials
It is important to determine the main factors that control the macrofauna distribution in a tidal flat. The
macrobenthos distribution is closely related to the distribution of sedimentary facies (Lee and Park, 1998, Yap et al.,
2003). According to Koo et al., 2005, Koo et al., 2007, the benthic fauna habitat is strongly dependent on the
temperature conditions, which are mainly influenced by exposure duration, and in turn, intertidal elevation. The
temperature in a tidal flat is also influenced by the water content of the surface sediment; thus, the tidal channel
network is a major factor in controlling the macrobenthos distribution. Ten control factors are considered in this
study: surface sediment facies (generated from IKONOS and KOMPSAT-2), an intertidal digital elevation model
(DEM), slope, aspect, distance from a tidal channel, tidal channel density, exposure time, and the spectral reflectance
of near-infrared (NIR) bands in high-spatial-resolution satellite data (IKONOS and KOMPSAT-2). Slope and aspect
were included as topographical components along with the intertidal DEM, and NIR bands were employed to
account for relative water content and thus the surface temperature in the tidal flat. Macrofauna samples were
collected in the field, among which five dominant species of mollusca were selected for habitat mapping. The
macrofauna samples were collected in 2009, while the map of surface sediments and the tidal channel network was

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 3/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

compiled from IKONOS satellite data collected in the early 2000s, which was the most recently available data. Apart
from the IKONOS, KOMPSAT-2 satellite image was used to generate the sediment distribution and the NIR band to
assess the applicability in this study of a high-resolution satellite image from the Korean remote sensing system.
Table 1 lists the spatial database used in this analysis.

Table 1. Spatial database of the controlling factors that influence the macrobenthos distribution in a tidal flat.

Category Extracted factors Data type

Benthic Macrofauna Each species Point

DEM Intertidal DEM Grid

Slope

Aspect

Exposure time

IKONOS Distance from channel Grid

Density of channel

Sediment distribution

Band 4

KOMPSAT Sediment distribution Grid

Band 4

The surface sediment facies in the Hwangdo tidal flat were mapped using grain-size data (collected in March 2004)
based on high-resolution IKONOS satellite data. The IKONOS image was acquired at 11:20 AM local time on 26
February 2001, which was 0.5 h before low tide, when the tide was at 0.78 m (as measured at Boryeong Station). The
sediment distribution was mapped using an object-based classification method. IKONOS bands 2, 3, and 4 were used
for the mapping of surface sediment, taking into account the high sensitivity of soil and water content to the tidal
channel distribution and topography in coastal regions. Fig. 2a shows the distribution of five classes of surface
sediment: mud flat (I), mud flat (II), mixed flat, sand flat, and sand shoal. Mud flats were differentiated into two
classes based on a minor difference in their reflectance. A KOMPSAT-2 image (acquired on 8 April 2008) was also
used for surface sediment mapping in combination with 28 grain samples obtained in May 2009. The map of surface
sediment distribution compiled from the KOMPSAT-2 image consists of four classes: mud flat, mixed flat, sand flat,
and sand shoal.

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 4/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

Download: Download full-size image

Fig. 2. Factors constructed in this study. (a) Sedimentary facies (from IKONOS). (b) Intertidal DEM. (c) Density of the
tidal channel. (d) DN values of KOMPSAT-2 band 4.

The intertidal DEM of the Hwangdo tidal flat was constructed from Landsat ETM + satellite images by the waterline
method, which is a useful approach in the practical application of satellite remote sensing to topographic mapping
in a tidal flat (Hoja et al., 2000, Ryu et al., 2008). The waterline method exploits the different tide conditions that are
seen in each image as a topographic contour line. The tidal flat DEM can be generated by stacking all the waterlines
acquired over a given short period. Five Landsat ETM + images obtained from December 2006 to June 2009 were used
as data in this study. Geometric rectification for Landsat ETM + data was conducted using a rectified IKONOS image,
employing an image-to-image method. The method also required a reference elevation for the extracted waterline
(Mason et al., 1997, Chen and Rau, 1998). Leveling data obtained on September 2009 within the central part of the
study area at ebb tide were used as the reference elevation. We extracted waterlines by applying the procedures of
Ryu et al. (2002). Fig. 2b shows the intertidal DEM of the study area constructed by the waterline method. The
overall trend shows a high elevation in the western part, with relatively high topographic gradients, and low relief in
the east. The slope gradient and aspect were derived from the intertidal DEM.

The exposure time at each macrofauna sampling site was calculated by comparing the tidal height acquired from the
Boryeang tidal station during 2008 with in situ observations of sea level at each site. A database of exposure duration
was generated using the intertidal DEM and the total duration for which a location at a certain sea level was exposed
above that level during the measurement time interval. A FORTRAN program was employed to calculate the
exposure time at each specific sea level.

The tidal channel network in the study area was extracted from the IKONOS image acquired on 26 February 2001. A
multi-spectral image (spatial resolution, 1 m) was generated from the panchromatic band image with a spatial
resolution of 1 m, combined with the visible-to-NIR band image (spatial resolution, 4 m) using an image sharpening
technique. The tidal channels were extracted and digitized from the spectrally enhanced image. A map of the tidal

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 5/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

channel density for the study area was constructed from the tidal channel database via surface analysis (Fig. 2c). A
map showing distance from tidal channels was also generated from the tidal channel database, using a buffering
method. Fig. 2c shows that the tidal channels are more densely distributed in the area of the mud flat facies, which is
dominated by relatively fine-grained particles, rather than in the area of coarse-grained sand flats.

Each NIR band reflectance of the IKONOS and KOMPSAT-2 images was constructed as a raster database. The NIR band
is an indicator of the sedimentary environment in the tidal flat, in terms of water content and remnant surface
water, and it shows a close relation with the surface sedimentary facies and surface temperature (Koo et al., 2005,
Ryu et al., 2004). We used the digital number (DN) value instead of spectral reflectance, because we could not
acquire the calibration coefficient for the conversion of radiance (DN value) into optical reflectance for the satellite
images. The value of each cell in the grid file was a DN value of IKONOS band 4 and KOMPSAT-2 band 4 (Fig. 2d).

In tidal flats, the spatial distribution of macrobenthic species shows continuous change due to spatial variations in
benthic environmental factors. A GIS can be effectively applied as a basic analysis tool in an assessment of the spatial
relationships between the occurrence of macrobenthos species and related environmental factors. In the present
study, maps were prepared as a spatial database using ArcGIS software (Environmental Systems Research Institute),
and these data were used in applying the weights-of-evidence model. For data processing, we used ArcGIS 9.3 with
the optional Spatial Analyst and 3D Analyst application. The spatial database included the intertidal DEM, slope,
aspect, distance from a channel, channel density, sedimentary facies (from IKONOS and KOMSAT-2 satellite data),
NIR bands (from IKONOS and KOMSAT-2 satellite data), and exposure time. The five species of mollusca considered
for habitat mapping were as follows: Cerithideopsilla cingulata, Cingulina cingulata, Laternula anatina, Musculista
senhousia, and Ruditapes philippinarum (Table 2). Each calculated and extracted factor was converted into a 4 × 4 m
grid file. The dimensions of the study area grid were 1238 rows by 503 columns (384,932 cells in total).

Table 2. Abundance of the dominant mollusca at each site (unit: individuals per 0.1 m2).

Name of H1 H2 H3 H4 H5 H6 H7 H8 H9 H10 H11 H12 H13 H14 H15 H16 H17 H18 H19 H20 H21 H22 H
species

Cerithideopsilla 2 10 6 7 0 0 0 4 13 0 3 0 0 8 26 0 0 0 0 9 32 0 0
cingulata

Cingulina 0 0 0 0 0 13 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
cingulata

Cyclina 3 0 0 3 0 0 0 2 3 0 5 0 0 0 1 0 0 0 0 1 1 0 0
sinensis

Laternula 0 3 1 0 0 0 0 1 4 0 8 0 0 2 2 2 0 0 1 10 3 1 0
anatina

Musculista 0 0 0 0 0 2 104 0 1 0 0 12 1 6 11 0 0 1 0 8 36 2 0
senhousia

Ruditapes 0 0 0 0 0 160 6 0 0 1 2 73 35 0 0 2 7 0 2 0 0 1 1
philippinarum

3. Methodology

3.1. In situ sampling

Samples of macrobenthos and surface sediment were collected from the study area (Fig. 3a). Macrobenthos samples
from 28 sites were collected on the Hwangdo tidal flat in May 2009 during a low spring tide. The sites were selected
by considering the tidal height and sediment facies known to be important factors in controlling the spatial
distribution of macrobenthos. Sediments were sampled four times at each site using can cores (sampling area,

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 6/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

0.025 m2; depth, 0.3 m; 4 replications) for analyses of grain size and macrofaunal mollusca. A differential GPS
(Pathfinder Pro XR Trimble Co., CA, USA) with 1 m horizontal accuracy was used for accurate positioning of the
sample sites. For mollusca analyses, the sediment samples were sieved through a 1.0 mm mesh screen by washing
with seawater; the residue on the screen was preserved in a 10% neutralized formalin solution. In the laboratory, the
residue specimens were sorted into major faunal groups, of which mollusca were identified to the species level if
possible and counted. Among the dominant species of identified mollusca, in terms of abundance and biomass
(Table 2, Table 3), five were determined and employed in mapping the potential mollusca habitats by remote sensing
and GIS. Fig. 3b–f shows photographs of the five species: Cerithideopsilla cingulata, Cingulina cingulata, Laternula
anatina, Musculista senhousia, and Ruditapes philippinarum. For grain size analysis, the sand and mud fractions were
separated by wet sieving through a 63 μm stainless-steel sieve after removing organic material and carbonate by
immersion in solutions of 10% H2O2 and 0.1 N HCl. Grain size distributions were determined by standard sieving
(Folk, 1980) and a Sedigraph-5100 for the sand and mud fractions, respectively. An inclusive graphic method was
used to determine the sediment type, mean grain size, and sorting (Folk and Ward, 1957).

Download: Download full-size image

Fig. 3. (a) Field sampling for macrofauna at Hwangdo tidal flat and five mollusca species: (b) Cerithideopsilla
cingulata, (c) Cingulina cingulata, (d) Laternula anatina, (e) Musculista senhousia, and (f) Ruditapes philippinarum.

Table 3. Dominant mollusca (in terms of abundance and biomass) in the study area. The occurrence frequency
represents the number of sites at which each species occurred among the 28 sites in total.

Species Total % Mean density Total % Mean biomass Occurring

abundance (ind./m2) biomass (gW Wt./m2) frequency

Ruditapes Bi 295 28.2 105 166.8 17.7 59.6 14

philippinarum

Musculista senhousia Bi 252 24.1 90 3.7 0.4 1.3 13

Cerithideopsilla Ga 134 12.8 48 108.8 11.5 38.9 12

cingulata

Laternula anatina Bi 42 4.0 15 3.7 0.4 1.3 15

Cyclina sinensis Bi 21 2.0 8 320.1 33.9 114.3 9

Cingulina cingulata Ga 14 1.3 5 0.2 0.0 0.1 2

In addition, 43 samples collected in March 2004 for grain size analysis were used to map the surface sediment
distribution in the study area at that time. Twenty-nine samples were obtained by a shipboard grab sampler during
flood tide, and 14 samples were obtained by hand at ebb tide. At least three samples were collected within a 10-m-
radius circle at each site, within the uppermost 5 mm of the surface sediment. Based on the percentage of grains
larger than very fine sand (0.0625 mm), grain-size data were classified into three facies types: sand flat (more than
70% of grains larger than very fine sand), mixed flat (30–70%), and mud flat (0–30%), following Folk (1968).

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 7/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

3.2. Weights-of-evidence model

The following formulation of the Bayesian probability model, known as the weights-of-evidence model, was applied
to mollusca habitat analysis, as synthesized from Bonham-Carter et al. (1989) and Bonham-Carter (1994). For a given
D, representing a pixel in the study area where a certain mollusca species occurred, N{D}, the total number of D, and
N{T}, the total number of the pixels in the study area, the prior probability of species occurrence (P{D}) is expressed
as

(1)

The prior probability is a constant for all pixels of every factor in the study area, because it does not relate to any of
the factors. For plural map data (in the present case, 10 control factors), let us suppose that B is a pixel that belongs
to the k class of the m factor, a predictor pattern, and N{B} is the total number of B in the study area and that a
number of individuals of the species occur preferentially within the class; i.e., N{D ∩ B}. Then, the probability of
species occurrence in areas occupied by the k class of the m factor is the probability of species occurrence inside the
surveyed area occupied by the k class of the m factor (P{D ∩ B}) to the probability that the surveyed area belongs to
the k class of the m factor (P{B}), as follows:

(2)

Similarly, the probability of species occurrence in the areas not occupied by the k class of the m factor is expressed
by Eq. (3):

(3)

Eqs. (2), (3) represent the conditional probability of species occurrence given the presence and absence of the
predictor pattern B, denoted by P{D|B} and , respectively; P{B|D} and are the conditional
probabilities of being inside and outside the predictor pattern B, respectively, given the presence of an occurrence D.
In addition, P{B} and are the prior probabilities of being inside and outside the predictor pattern, B,
respectively. The same model can be expressed as an odds-type formulation, where the odds, O, are defined as
O = P/(1 – P). Eqs. (2), (3) can be expressed as the odds, as follows:

(4)

(5)

where O{D|B} and are the conditional odds of species occurrence, given the presence and absence of a
binary predictor pattern B, respectively, and O{D} represents the prior odds of species occurrence. If the natural
logarithms are taken for Eqs. (4), (5), the weights for the binary predictor pattern B are obtained as follows:

(6)

(7)

where W+ and W− are the weights of evidence when a binary predictor pattern is present and absent, respectively.
Ultimately, W+ is obtained by taking the natural logarithm to the ratio of ((number of species in the k class of the m
factor)/(total number of he species in the study area)) to ((number of pixels belonging to the k class of the m factor
in which the species did not occur)/(total number of pixels in the study area in which the species did not occur)). W−
represents the natural logarithm of the ratio of ((number of species in classes other than the k class of the m
factor)/(total number of species in the study area)) to ((number of pixels belonging to classes other than the k class
of the m factor in which the species did not occur)/(total number of pixels in the study area in which the species did
not occur)).

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 8/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

To generate the binary predictor patterns of each factor, it was necessary to determine the rating or range that
optimized the spatial association between mollusca occurrence and each factor. The optimum cutoff for the binary
pattern was determined by calculating C/S(C), the studentized value of contrast, which is estimated as the ratio of
the contrast to its standard deviation. The magnitude of the contrast, C, is determined from the difference between
W+ and W−, and provides a measure of the spatial association between a set of mollusca and the predictor patterns.
The binary predictor patterns were also assigned weights with the maximum C/S(C) and were calculated as follows:

(8)

where SPI = species potential index, WOE = W+, and W− is the binary pattern for the range of each factor value or factor
class at Max. C/S(C).

4. Results

4.1. Mollusca composition

According to an analysis of in situ data, 42 molluscs species comprising 1047 specimens were identified in the study
area, with a mean density and biomass of 374 inds./m2 and 337 gW Wt./m2, respectively. The dominant species in
terms of numerical abundance was a short-necked clam, Ruditapes philippinarum, with a mean density of
105 inds./m2, showing 14 times of occurrence frequency in the study area (Table 3). This species occurred mainly in
the lower flat, especially in areas of sandy sediment, and showed the highest density of 1600 inds./m2 at site H6
(Table 2). The second most dominant species was Musculista senhousia, a green bagmussel, which made up 24.1% of
the total abundance. The muddy middle flat was the main habitat of this species, located at a higher elevation than
the habitat of Ruditapes. The mud creeper, Cerithideopsilla cingulata, and two bivalves (Laternula anatina and Cyclina
sinensis) were the next main contributors to abundance; these species were distributed throughout the upper and
middle muddy flats. Cerithideopsilla extended its habitat into the narrow strip of salt marsh along the coast of the
island, where Suaeda japonica is the dominant vegetation. A lantern clam, L. anatina, showed the highest occurrence
frequency in the study area, and a venus clam, C. sinensis, was the main contributor to biomass (total biomass,
320.1 gW Wt.), accounting for 33.9% of the total mollusca biomass (Table 3). A pyramid shell, Cingulina cingulata, was
restricted to the sand shoal at site H6, where it occurred with a mean density of 130 inds./m2.

4.2. Habitat potential mapping

The spatial relationships between species occurrence and species-related factors were derived via a weights-of-
evidence model to produce a species habitat potential map for the study area. The weights-of-evidence is based on
the observed relations between each factor and the distribution of species, and the C/S(C) values were obtained for
each subgroup of each factor. In the case of Cerithideopsilla cingulata, for example, the factor ‘Intertidal DEM’ has a
maximum C/S(C) value of 7.79 in the third class of high elevations. The class of the highest elevations had also a high
C/S(C) value (7.07), whereas the first five classes of low elevations show no relation with the C. cingulata habitat in
the study area. This finding indicates that C. cingulata mainly inhabits elevated locations in the Hwangdo tidal flat.
The maximum C/S(C) values in the sediment distribution derived from IKONOS or KOMPSAT-2 satellite data were
obtained for the class ‘Mixed flat’; the class ‘Mud flat’ also showed a relatively high value, indicating the
predominance of the species in the mixed or mud flat facies. For C. cingulata, the first class of long exposure duration
in the ‘Exposure time’ group showed the highest C/S(C) value (13.02) among all the classes of all factors. This species
is therefore an indicator of scarce remnant surface water and interstitial water, as a consequence of the long
exposure time. The spatial relations between each mollusca species and each control factor can be identified
quantitatively from the C/S(C) values for Cingulina cingulata, Laternula anatina, Musculista senhousia, and Ruditapes
philippinarum, respectively.

The weights-of-evidence value (WOE) for each class of each factor was determined from the class with the
maximum C/S(C) value. In the case of Cerithideopsilla cingulata, for example, the W+ and W− values of the third class
of high elevations in the factor ‘Intertidal DEM’ are the WOEs. WOE of the third class is W+ (i.e., 1.03), and WOE of the
other classes is W− (i.e., −0.36). WOE represents the possibility of a species occurrence; therefore, it can be assigned a

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 9/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

relative weight for the species habitat potential of each cell in the factor grid file. In the case of Cerithideopsilla
cingulata, in the intertidal DEM grid file, the group of cells belonging to the third class of high elevations had a WOE
of 1.03; other cells had a WOE of −0.36. Ten grid files of the factor maps, with a WOE at each pixel, were multiplied
using Eq. (8). The integrated value indicates the relative degree of the species potential habitat, which can be
expressed as a species potential index (SPI). A high SPI value indicates a higher potential for occurrence of the
species. Fig. 4a–e shows habitat potential maps for each of the five mollusca species, with an SPI value at each pixel.
The SPI values of each habitat potential map were categorized into four classes for visual interpretation: “very high”
for those belonging to the top 5% of values, “high” for the second class of 85–95%, “medium” for the third class of
70–85%, and “low” for the last class of 0–70%.

Download: Download full-size image

Fig. 4. Habitat potential maps for the following Mollusca: (a) Cerithideopsilla cingulata, (b) Cingulina cingulata, (c)
Laternula anatina, (d) Musculista senhousia, and (e) Ruditapes philippinarum.

The Cerithideopsilla cingulata habitat potential map (Fig. 4a) shows high SPI values in areas covered by mixed flat
facies (Fig. 2a), at high elevations (Fig. 2b), near the tidal channel (Fig. 2c), and with a high DN value for KOMPSAT-2
band 4 (Fig. 2d). This finding agrees with the WOE patterns discussed above. Cingulina cingulata and Ruditapes
philippinarum were concentrated mainly in the sand shoal and in areas without tidal channels, with a short exposure
time, and high NIR band reflectance in terms of WOE, which is consistent with the habitat potential map in Fig. 4b
and e. This result indicates that these two species mainly inhabit well-drained coarse-grained sediments at
relatively low elevations. Laternula anatina showed a similar habitat pattern to that of Cerithideopsilla cingulata in
that the species was predominant in the mud flat facies, in areas with relatively high elevations and longer exposure
duration (Fig. 4c). However, this species showed no relationship with the tidal channel distribution or NIR band DN

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 10/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

value in terms of the weights-of-evidence value. The Musculista senhousia habitat was strongly influenced by a steep
slope with short exposure time, even though it showed a considerable spatial relationship with the mud or mixed
flat facies (Fig. 4d).

4.3. Verification
The potential habitat analysis of macrobenthic species was validated using known species locations. Verification was
performed by comparing the known species data with the mollusca potential habitat maps (see Fig. 5). The success
rates shown in Fig. 5 indicate the degree to which the model and factors were able to predict the species habitat
potential. To obtain the relative ranks for each prediction pattern, the calculated SPI values of all the cells in the
study area were sorted in descending order. The ordered cell values were then divided into 100 classes, at 1%
intervals. This procedure was also adopted for the species occurrence cells by comparing the 100 classes with the
species distribution in the study area. For example, in the case of Cerithideopsilla cingulata, the 90–100% (10%) class
for the study area, where the species potential habitat index had a higher rank, explained the distribution of 61% of
all individuals of Cerithideopsilla cingulata. In addition, the 80–100% (20%) class for the study area, where the species
potential habitat index had a higher rank, explained the distribution of 71% of all individuals of Cerithideopsilla
cingulata. To quantitatively estimate the accuracy of the predicted potential habitat, we calculated the ratio of the
area under the curve to the total area of the graph. If 100% of the occurrences of Cerithideopsilla cingulata belonged
to the first class, indicating the highest 1% of SPI values, the area under the curve was equal to the total area of the
graph, corresponding to a prediction accuracy of 100%. Therefore, the area under the curve ratio can be used to
estimate the prediction accuracy. The area under the curve ratio (i.e., the prediction accuracy) for each of the five
species is shown in Fig. 5. The species habitat maps have accuracies of 88.68, 96.57, 80.88, 70.09, and 94.97 for
Cerithideopsilla cingulata, Cingulina cingulata, Laternula anatina, Musculista senhousia, and Ruditapes philippinarum,
respectively.

Download: Download full-size image

Fig. 5. Success rate curves showing the cumulative percentage of each species occurrence (y-axis) for the species
potential index (SPI) rank (x-axis) (in descending order) for Assiminea latericea, Cerithideopsilla cingulata,
Cerithideopsilla djadjariensis, Cingulina cingulata, Glauconome chinensis, Laternula anatina, and Musculista senhousia.

5. Conclusion and discussion

The occurrence potentials for five mollusca were quantitatively assessed for a tidal flat using a GIS-based weights-
of-evidence model. Ten controlling factors relating to the macrobenthos distribution were considered and their
relative weights were determined using the weights-of-evidence model. The weights-of-evidence values were then
integrated into a species potential index for habitat mapping of the five mollusca species.

Based on field observations, we identified 42 molluscs comprising 1047 specimens, among which Ruditapes
philippinarum and Musculista senhousia were the first and second most dominant species in the Hwangdo tidal flat.

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 11/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

The former occurred mainly in sandy sediments at relatively low elevations, and the latter was found in muddy or
mixed flat settings.

The WOE values were effective in estimating the spatial relationships between each mollusca species and control
factors that influence the species habitat, and they were successfully integrated as an SPI for mapping of habitat
potential. A mud creeper, Cerithideopsilla cingulata, and a bivalve, Laternula anatina, inhabited the mud flat area,
which occurred at relatively high elevations and had a long exposure duration. This finding indicates that these
species are indicators of fine-grained sediments with little surface or interstitial water, which show a relatively high
optical reflectance in satellite images of the Hwangdo tidal flat, as reported by Choi et al. (2010). The habitats of a
short-necked clam, Ruditapes philippinarum, and a pyramid shell, Cingulina cingulata, were limited to the sand shoal,
which has a short exposure time and high NIR band reflectance. This result demonstrates that the species are
indicators of well-drained coarse-grained sediments at low elevations. A green bagmussel, Musculista senhousia,
occurred in areas with a steep slope gradient and short exposure time.

The reliability of the habitat potential maps was evaluated in terms of prediction accuracy, using a success rate curve
method. The verification was carried out by comparing the produced maps with the sampling locations of the
species data. The accuracies all exceeded 88%, except for those for Laternula anatina and Musculista senhousia, which
were 80.88% and 70.09%, respectively. Therefore, the habitat potential maps were consistent with the actual
mollusca occurrences.

Os resultados atuais demonstram que um modelo de pesos de evidência baseado em SIG é eficaz para estimativas
quantitativas das relações espaciais entre a macrofauna e os ambientes bentônicos que influenciam o habitat.
Consequentemente, tal abordagem é ideal para o mapeamento de habitats em ambientes de planícies de maré. O
modelo seria facilmente empregado para decidir a ponderação de cada fator de controle e facilmente incorporado ao
software SIG para mapear a distribuição do macrobentos. No presente estudo, não consideramos variações sazonais
nos ambientes bentônicos. Koo et al. (2005) relataram que a temperatura é uma influência importante na
distribuição de espécies em planícies de maré. Assim, seria importante avaliar as diferenças sazonais em ambientes
bentônicos relacionados ao habitat da macrofauna, de uma perspectiva ecológica.

Este artigo considerou vários ambientes bentônicos em uma planície de maré como fatores relacionados ao
macrobentos. Os ambientes bentônicos são complexamente inter-relacionados. A distribuição de fácies
sedimentares superficiais está intimamente relacionada à topografia da maré, incluindo um DEM intermareal,
gradiente de declive do solo e a rede de canais, que são os principais fatores que influenciam a duração da exposição
e os conteúdos de água superficial e água intersticial. Em última análise, todos esses ambientes se combinam para
influenciar a reflectância espectral de dados de sensoriamento remoto. Em particular, no caso da planície de maré de
Hwangdo, as características da topografia da maré e das fácies sedimentares superficiais foram fortemente refletidas
na reflectância de sensoriamento remoto de alta resolução espacial ( Choi et al., 2010 ). Assim, espera-se que a
tendência aproximada do habitat do macrobentos possa ser deduzida a partir do processamento apropriado de
dados de sensoriamento remoto. Consequentemente, se uma série temporal de imagens de sensoriamento remoto
de alta resolução espacial pudesse ser adquirida, as variações sazonais no habitat poderiam ser determinadas na área
de estudo atual. Esse avanço seria de fundamental importância como um método científico de fácil aplicação para
avaliar e preservar as propriedades ecológicas das regiões costeiras.

Agradecimentos
Esta pesquisa foi parte do projeto intitulado “Desenvolvimento do Sistema Integrado de Gestão Estuarina”,
financiado pelo Ministério de Terras, Transportes e Assuntos Marítimos da Coreia, e do Projeto de Pesquisa Básica do
Instituto Coreano de Geociências e Recursos Minerais (KIGAM), financiado pelo Ministério de Ciência e Tecnologia da
Coreia .

Artigos recomendados

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 12/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

References
Bonham-carter, 1994 G.F. Bonham-carter
Geographic Information Systems for Geoscientists: Modeling with GIS: Computer Methods
Pergamon, Netherlands (1994)
Google Scholar

Bonham-carter et al., 1989 Bonham-carter, G.F., Agterberg, F.P., Wright, D.F., 1989. Weights of evidence modeling: A
new approach to mapping mineral potential. In: Agterberg, F.P., Bonham-Carter, G.F. (Eds.), Statistical
Applications in the Earth Sciences, Geological Survey of Canada, vol. 98, pp. 171–183.
Google Scholar

Chen and Rau, 1998 L.C. Chen, J.Y. Rau

Detection of shoreline changes for tideland areas using multi-temporal satellite images
Int. J. Remote Sens., 19 (1998), pp. 3383-3397
View in Scopus Google Scholar

Choi et al., 2010 J.K. Choi, J.H. Ryu, Y.K. Lee, H.R. Yoo, H.J. Woo, C.H. Kim
Quantitative estimation of intertidal sediment characteristics using remote sensing and GIS
Estuar. Coast. Shelf Sci., 88 (2010), pp. 125-134
View PDF View article View in Scopus Google Scholar

Dahal et al., 2008 R.K. Dahal, S. Hasegawa, A. Nonomura, M. Yamanaka, M. Takuro, K. Nishino
GIS-based weights-of-evidence modelling of rainfall-induced landslides in small catchments for
landslide susceptibility mapping
Environ. Geol., 54 (2008), pp. 311-324
Crossref View in Scopus Google Scholar

Folk, 1968 R.L. Folk

A review of grain size parameters
Sedimentology, 6 (1968), pp. 73-93
Google Scholar

Folk, 1980 R.L. Folk

Petrology of Sedimentary Rocks
Hemphill, Austin, TX (1980)
Google Scholar

Folk and Ward, 1957 R.L. Folk, W.C. Ward

Brazos river bar: a study in the significance of grain size parameters
J. Sed. Petrology., 27 (1957), pp. 3-27
View in Scopus Google Scholar

Fulton et al., 2010 J.W. Fulton, C.R. Wagner, M.E. Rogers, G.F. Zimmerman
Hydraulic modeling of mussel habitat at a bridge-replacement site, Allegheny River, Pennsylvania,
USA
Ecol. Model., 221 (2010), pp. 540-554
View PDF View article View in Scopus Google Scholar

Hatten and Parsley, 2009 J.R. Hatten, M.J. Parsley

A spatial model of white sturgeon rearing habitat in the lower Columbia River, USA
Ecol. Model., 220 (2009), pp. 3638-3646
View PDF View article View in Scopus Google Scholar

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 13/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…
Hoja et al., 2000 Hoja, D., Lehner, S., Niedermeier, A., Romaneessen, E., 2000. DEM generation from ERS SAR
shorelines compared to airborne crosstrack InSAR DEMs in the German Bight. In: Proceedings of the
International Geoscience and Remote Sensing Symposium, IEEE, Hawaii, USA, pp. 1889–1891.
Google Scholar

Jorgensen and Kollmann, 2009 R.H. Jorgensen, J. Kollmann

Invasion of coastal dunes by the alien shrub Rosa rugosa is associated with roads, tracks and
houses
Flora, 204 (2009), pp. 289-297
View PDF View article View in Scopus Google Scholar

Kim and Kim, 1996 Y.S. Kim, J.N. Kim

Biogenic sedimentary structures of crustaceans at the intertidal flat of Whang Island, Cheonsu
Bay
J. Kor. Earth Sci. Soc., 17 (1996), pp. 357-364
(in Korean)
Crossref Google Scholar

Koo et al., 2005 B.J. Koo, K.K. Kwon, J.H. Hyun

The sediment-water interface increment due to the complex burrows of macrofauna in a tidal flat
Ocean Sci. J., 40 (2005), pp. 221-227
Google Scholar

Koo et al., 2007 B.J. Koo, K.K. Kwon, J.H. Hyun

Effect of environmental conditions on variation in the sediment-water interface created by
complex macrofaunal burrows on a tidal flat
J. Sea Res., 58 (2007), pp. 302-312
View PDF View article View in Scopus Google Scholar

Lee and Park, 1998 J.H. Lee, H.S. Park

Community structures of macrobenthos in Chonsu Bay, Korea
J. Kor. Soc. Oceanogr., 33 (1998), pp. 18-27
View in Scopus Google Scholar

Lee and Pradhan, 2006 S. Lee, B. Pradhan

Probabilistic landslide hazards and risk mapping on Penang Island, Malaysia
J. Earth Syst. Sci., 115 (2006), pp. 661-672
View in Scopus Google Scholar

Mason et al., 1997 Mason, D.C., Hill, D., Davenport, I., Flather, R., Robinson, G., 1997. Improving inter-tidal digital
elevation models constructed by the waterline technique. In: Proceedings of the third ERS Symposium,
Florence, Italy, pp. 1079-1082.
Google Scholar

Oh et al., 2009 H.J. Oh, S. Lee, W. Chotikasathien, C.H. Kim, J.H. Kwon
Predictive landslide susceptibility mapping using spatial information in the Pechabun area of
Thailand
Environ. Geol., 57 (2009), pp. 641-651
View in Scopus Google Scholar

Regmi et al., 2010 N.R. Regmi, J.R. Giardino, J.D. Vitek

Modeling susceptibility to landslides using the weight of evidence approach: Western Colorado,
USA
Geomorphology, 115 (2010), pp. 172-187

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 14/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…
View PDF View article View in Scopus Google Scholar
Ryu et al., 2008 J.H. Ryu, C.H. Kim, Y.K. Lee, J.S. Won, S.S. Chun, S. Lee
Detecting the intertidal morphologic change using satellite data
Estuar. Coast. Shelf Sci., 78 (2008), pp. 623-632
View PDF View article View in Scopus Google Scholar

Ryu et al., 2004 J.H. Ryu, Y.H. Na, J.S. Won, R. Doerffer
A critical grain size for Landsat ETM+ investigations into intertidal sediments: a case study of the
Gomso tidal flats, Korea
Estuar. Coast. Shelf Sci., 60 (2004), pp. 491-502
View PDF View article View in Scopus Google Scholar

Ryu et al., 2002 J.H. Ryu, J.S. Won, K.D. Min

Waterline extraction from Landsat TM data in a tidal flat – a case study in Gomso Bay, Korea
Remote Sens. Environ., 83 (2002), pp. 442-456
View PDF View article View in Scopus Google Scholar

So et al., 1998 J.G. So, G.T. Jeong, J.W. Chae

Numerical modeling of changes in tides and tidal currents caused by Embankment at Chonsu Bay
J. Kor. Soc. Coastal Ocean Eng., 10 (1998), pp. 151-164
(in Korean)
Google Scholar

Van der Wal et al., 2008 D. Van der Wal, P.M.J. Herman, R.M. Forster, T. Ysebaert, F. Rossi, E. Knaeps, Y.M.G. Plancke, S.J. Ides
Distribution and dynamics of intertidal macrobenthos predicted from remote sensing: response
to microphytobenthos and environment
Mar. Eco. Prog. Ser., 367 (2008), pp. 57-72
Crossref View in Scopus Google Scholar

Woo and Je, 2002 H.J. Woo, J.G. Je

Changes of sedimentary environments in the southern tidal flat of Kanghwa Island
Ocean Polar Res., 24 (2002), pp. 331-343
(in Korean)
Crossref View in Scopus Google Scholar

Yalcin, 2008 A. Yalcin

GIS-based landslide susceptibility mapping using analytical hierarchy process and bivariate
statistics in Ardesen (Turkey): comparisons of results and confirmations
Catena, 72 (2008), pp. 1-12
View PDF View article View in Scopus Google Scholar

Yap et al., 2003 C.K. Yap, A. Rahim Ismail, A. Ismail, S.G. Tan
Species diversity of macrobenthic invertebrates in the Semenyih River, Selangor, Peninsular
Malaysia
Pertanika J. Trop. Agric. Sci., 26 (2003), pp. 139-146
Google Scholar

Yoo et al., 2007 J.W. Yoo, I.S. Hwang, J.S. Hong

Inference models for tidal flat elevation and sediment grain size: a preliminary approach on tidal
flat macrobenthic community
Ocean Sci. J., 42 (2007), pp. 69-79
View in Scopus Google Scholar

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 15/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

Cited by (26)

Potential of remote sensing in management of tidal flats: A case study of thematic mapping in the
Korean tidal flats
2014, Ocean and Coastal Management

Citation Excerpt :
…Correlations between the distribution of macrozoobenthos and environmental factors were quantitatively calculated using a
frequency ratio model based on a geographic information system (GIS), logistic regression, and an artificial neural network.
Potential habitat mapping of macrozoobenthos was performed based on the weighted value for each factor as shown in Fig. 6
(Choi et al., 2011b,c,d; Lee et al., 2013). Herbaceous halophytes were mapped based on the backscattering coefficients in TerraSAR-
X data (Lee et al., 2012).…

Show abstract

Macrobenthos habitat potential mapping using GIS-based artificial neural network models
2013, Marine Pollution Bulletin

Citation Excerpt :
…Recently, attempts have been made to map the macrobenthos habitat in tidal flats based upon the relationships among the
controlling variables. Choi et al. (2011b,c) proposed and tested models of macrofauna habitat maps for mollusca and crustacean
species using a remote-sensing technique. The models were applied based upon a geographic information system (GIS).…

Show abstract

Spatial distribution characteristics of oil spills in the bohai sea based on satellite remote sensing and
GIS
2019, Journal of Coastal Research

High-accuracy tidal flat digital elevation model construction using tanDEM-X science phase data
2017, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

GOCI, the world's first geostationary ocean color observation satellite, for the monitoring of temporal
variability in coastal water turbidity
2012, Journal of Geophysical Research Oceans

Combining Multispectral and Radar Imagery with Machine Learning Techniques to Map Intertidal
Habitats for Migratory Shorebirds
2022, Remote Sensing

View all citing articles on Scopus

Ver Resumo

Copyright © 2010 Elsevier Ltd. Todos os direitos reservados.

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 16/17
03/09/2025, 10:12 Mapeamento do habitat de macrobentos em uma planície de maré usando dados de sensoriamento remoto e um modelo pro…

Todo o conteúdo deste site: Copyright © 2025 Elsevier BV, seus licenciadores e colaboradores. Todos os direitos reservados, incluindo os de mineração de texto e dados, treinamento em IA e
tecnologias similares. Para todo o conteúdo de acesso aberto, aplicam-se os termos de licenciamento relevantes.

https://www-sciencedirect-com.ez278.periodicos.capes.gov.br/science/article/pii/S0025326X10005175?via%3Dihub 17/17