Spatial distribution, abundance and habitat use of the endemic Mediterranean fan mussel Pinna nobilis in the Gera Gulf, Lesvos (Greece): comparison of design-based and model-based approaches

Abstract

An important population of the endemic Mediterranean fan mussel Pinna nobilis thrives in the marine protected area of the Gera Gulf (Lesvos island, northeastern Aegean Sea, Greece), and was assessed for the first time. To estimate the abundance, spatial distribution and habitat use of fan mussels in the Gera Gulf, a distance sampling underwater survey was conducted. Detectability was properly modelled to secure unbiased estimates of population density. Two approaches were applied to analyze survey data, a design-based and a model-based approach using generalized additive models. The first approach was based on stratified random sampling on two strata, one assumed ‘preferable’ zone close to the coastline, and an assumed unsuitable habitat with predominantly muddy sediments, in which low sampling effort was put. For the needs of the model-based approach, a dedicated cruise was conducted to collect bathymetric data with a single-beam echo-sounder and map the bathymetry of the study area. A very high-resolution image from the Worldview-3 satellite was processed based on an object-based image analysis for mapping all main habitat types in the study area. The estimated abundance by the design-based approach was biased low as the stratum of pre-assumed unsuitable habitat proved to include patches of suitable habitats with high population densities that were missed by sampling. The model-based approach provided an abundance estimate of 213300 individuals (95% confidence interval between 97600–466000 individuals), which renders the fan mussel population of the Gera Gulf the largest recorded population in Greece. Population density peaked between 1.5-8 m depth and became practically zero at depths >15 m. There was a bathymetric segregation of fan mussel size classes, with the density of small individuals peaking at shallow waters, while that of large individuals peaked deeper. The highest population densities were observed in Posidonia oceanica meadows, followed by mixed bottoms (with reefs, rocks and sandy patches), while densities were very low on sandy and zero on muddy sediments. The current assessment provided a baseline for the future monitoring of the fan mussel population in the Gera Gulf. In view of the currently (2017-2018) ongoing mass mortality of the species in the western Mediterranean, continuing monitoring of the main fan mussel populations, such as the one in Gera Gulf, is of utmost importance.

Full text

Medit. Mar. Sci., 19/3, 2018, 642-655
642
Mediterranean Marine Science
Indexed in WoS (Web of Science, ISI Thomson) and SCOPUS
The journal is available on line at http://www.medit-mar-sc.net
DOI: http://dx.doi.org/10.12681/mms.14156
Spatial distribution, abundance and habitat use of the endemic Mediterranean fan mussel
Pinna nobilis in Gera Gulf, Lesvos (Greece): comparison of design-based
and model-based approaches
ALEXANDROS TSATIRIS, VASILEIOS PAPADOPOULOS, DESPINA MAKRI, KONSTANTINOS
TOPOUZELIS, EVA MANOUTSOGLOU, THOMAS HASIOTIS and STELIOS KATSANEVAKIS
Department of Marine Sciences, University of the Aegean, 81100 Mytilene, Greece
Corresponding author: stelios@katsanevakis.com
Handling Editor: Emma Cebrian
Received: 5 July 2017; Accepted: 6 August 2018; Published on line: 31 December 2018
Abstract
An important population of the endemic Mediterranean fan mussel Pinna nobilis thrives in the marine protected area of Gera
Gulf (Lesvos island, north-eastern Aegean Sea, Greece), and was assessed for the first time. To estimate the abundance, spatial
distribution and habitat use of fan mussels in Gera Gulf, a distance sampling underwater survey was conducted. Detectability was
modelled to secure unbiased estimates of population density. Two approaches were applied to analyze survey data, a design-based
and a model-based approach using generalized additive models. The first approach was based on stratified random sampling on
two strata, an assumed ‘preferable’ zone close to the coastline and an assumed unsuitable habitat, with predominantly muddy sed-
iments, in which low sampling effort was applied. For the needs of the model-based approach, a dedicated cruise was conducted
to collect bathymetric data with a single-beam echo-sounder and map the bathymetry of the study area. A very high-resolution
image from the Worldview-3 satellite was processed, based on an object-based image analysis, for mapping all main habitat types
in the study area. The estimated abundance using the design-based approach was low-biased as the stratum of pre-assumed un-
suitable habitat proved to include patches of suitable habitats with high population densities that were missed by sampling. The
model-based approach provided an abundance estimate of 213300 individuals (95% confidence interval between 97600-466000
individuals), which renders the fan mussel population of Gera Gulf the largest recorded population in Greece. Population density
peaked between 1.5-8 m depth and became practically zero at depths >15 m. A bathymetric segregation of fan mussel size classes
was noted, with the density of small individuals peaking in shallow waters, while that of large individuals peaked deeper. The
highest population densities were observed in Posidonia oceanica meadows, followed by mixed bottoms (with reefs, rocks and
sandy patches), while densities were very low on sandy and zero on muddy sediments. The current assessment provides a baseline
for future monitoring of the fan mussel population in Gera Gulf. In view of the current (2017-2018) ongoing mass mortality of the
species in the western Mediterranean, continuous monitoring of the main fan mussel populations, such as the one in Gera Gulf, is
of utmost importance.
Keywords: Abundance estimation; spatial distribution; distance sampling; line transects; SCUBA; endangered species.
Introduction
Pinna nobilis Linnaeus, 1758 is a marine bivalve mol-
lusc of the Pinnidae family, commonly known as noble
pen shell or fan mussel. It is one of the largest bivalves
in the world and the largest in the Mediterranean Sea,
where it is endemic. The average anterio-posterior length
of adult individuals is 30-50 cm but it can reach the size
of 120 cm (Zavodnik et al., 1991). Its lifespan commonly
exceeds 20 years and can even reach 45 years (Rouanet
et al., 2015). P. nobilis occurs at depths ranging between
0.5 and 60 m. It usually inhabits seagrass meadows such
as Posidonia oceanica, Zostera marina, Z. noltii and Cy-
modocea nodosa (Zavodnik et al., 1991), but it can also
be abundant in macroalgal beds (Katsanevakis &Thessa-
lou-Legaki, 2009) and unvegetated soft bottoms (Katsa-
nevakis, 2006; Addis et al., 2009).
P. nobilis was formerly targeted for its meat and bys-
sus from which sea silk was produced. It is currently
strictly protected under the EU Habitats Directive (92/43/
EEC, Annex IV), the Protocol for Specially Protected
Areas and Biological Diversity in the Mediterranean of
the Barcelona Convention (Annex II), and the national
legislation of most Mediterranean countries. Neverthe-
less, it is still illegally exploited and marketed in many
countries (Katsanevakis et al., 2011). Despite protection,
in the last decades its populations have been declining
(Basso et al., 2015), due to direct threats such as trawling
Research Article

643
Medit. Mar. Sci., 19/3, 2018, 642-655
and anchoring (Vázquez-Luis et al., 2015), illegal collec-
tion by divers for food, decorative purposes, and for its
byssus (Zavodnik et al., 1991; Katsanevakis, 2007a), and
indirect threats such as habitat loss or degradation. Since
autumn 2016, a mass mortality event, caused by the par-
asite Haplosporidium pinnae (Catanese et al., 2018) has
caused, so far, an estimated loss of ~90% of the Spanish
P. nobilis populations (Vázquez-Luis et al., 2017) and has
raised concerns about the status of the species in the en-
tire Mediterranean basin.
In addition to the general prohibition on its exploita-
tion and marketing, the NATURA 2000 network can
contribute to the protection of important populations of
P. nobilis and its important habitats, such as Posidonia
oceanica meadows. NATURA 2000 is one of the world’s
most extensive networks of conservation areas, which
currently consists of more than 27,200 sites, of which ap-
proximately 15% include marine areas (Mazaris et al.,
2018). Nevertheless, many of the marine sites of the NA-
TURA 2000 network are poorly monitored and managed,
and proper assessments of the population status of pro-
tected species within their boundaries are often lacking.
The aim of this study was to estimate the population
status of an important Pinna nobilis population in Gera
Gulf (Lesvos island, Greece), which is part of the NATU-
RA 2000 network (site codes: GR4110013, GR4110005).
Two approaches were followed for abundance estimation,
a design-based and a model-based approach. The latter
also allowed assessment of the spatial distribution of the
species in the gulf and its habitat use, i.e. variability in
its abundance in different habitat types, which is actually
a combination of preferential settlement and differential
mortality. Despite Gera Gulf being part of the NATURA
2000 network, there have been no previous assessments
of its P. nobilis population, and thus this study serves as
a baseline for assessment of future population trends.
In view of the ongoing mass mortality of the species in
the western Mediterranean (Vázquez-Luis et al., 2017),
monitoring all important populations of the species is of
utmost importance.
Methods
Study Area
Gera Gulf is an enclosed elongated embayment, locat-
ed in the south-eastern part of Lesvos Island, north-east-
ern Aegean Sea, Greece (Fig. 1). It receives discharges
from seasonal streams and small rivers, and is connect-
ed to the open sea through a narrow channel of ~6.5 km
length and 300 – 800 m width. For the purposes of this
study, a detailed large scale map of Gera Gulf was created
using a SENTINEL-2 satellite image. The total surface
of the gulf is 4009.9 ha (calculated using ArcGIS 10.2.2
‘calculate geometry function’). Nearshore the substrate is
dominated by sand mixed with gravel, cobbles or rocks,
followed further offshore by sandy and muddy mixtures.
In the south and western part of the gulf, there are patchy
Posidonia oceanica meadows.
Bathymetry
Bathymetric data were collected during a cruise on-
board R/V Amfitriti, using a Simrad CA44 single-beam
echo-sounder operating at 200 kHz, along a ~270 km sur-
vey grid of crossing lines. Vessel speed was maintained at
about 4 knots. The depth was corrected for sound velocity
(1500 m/s) and transducer depth. ArcGIS 10.2 was used
to produce the bathymetry of the gulf through interpola-
tion. However, it is well-known that different interpola-
tion techniques produce different values at the same grid
points thus introducing a degree of uncertainty (Chiles
& Delfiner, 1999). Therefore, to adopt the most reliable
results, 4 interpolation methods were examined: Topo to
raster, Kriging (Ordinary and Universal), Inverse Dis-
tance Weighted (with topical and spherical parameters)
and Spline with barriers. Errors quantification was man-
aged by the Mean Absolute Error (MAE) and the Root
Mean Square Error (RMSE). For the validation proce-
Fig. 1: Gera Gulf and its location in the Aegean Sea. The eigh-
teen sampling stations are indicated.

Medit. Mar. Sci., 19/3, 2018, 642-655
644
dure, a subgroup of 13830 points was pre-selected (25%
of the total points) to compare the results with the initial
dataset and estimate the MAE and RMSE. The compari-
son showed that the Spline method was the best, having
the lowest MAE (0.02) and RMSE (0.14). Finally, a ras-
ter file with 2-m pixel size was created from the point
data set.
Habitat Map
A habitat map of Gera Gulf was created by classify-
ing a very high spatial resolution image from the Worl-
dview-3 satellite, acquired on 18-11-2015. The spatial
resolution of the five multispectral bands (coastal, blue,
green, red, infrared) was 1.5 m.The image was pre-pro-
cessed by applying a land mask derived from the infrared
band. An Object Based Image Analysis (OBIA) approach
was followed, using the other four bands and eCognition
5.4. software. The analysis involved image segmentation
into small objects (segments), which are groups of pixels
with similar characteristics, used as the main processing
element (Blaschke et al., 2010). Segmentation was ap-
plied using a scale factor of 50, and a homogeneity cri-
terion (with shape value of 0.1 and compactness value
of 0.5). Finally, supervised classification took place in
the following four classes: (a) Posidonia oceanica mead-
ows, (b) mixed sea bed (cobbles, rocky reefs and sandy
patches), (c) sandy sediment and (d) muddy sediment.
There were no extensive rocky areas in Gera Gulf and all
hard substrates were patchily distributed among soft sub-
strates, which was the reason for not including a separate
hard substrate habitat in our classification.
Line transect sampling - Field Work
The single observer line transect distance sampling
method by SCUBA diving was applied for abundance es-
timations (Katsanevakis, 2007b). This approach has been
used extensively for surveying Pinna nobilis populations
(Katsanevakis, 2006, 2007b; Katsanevakis &Thessalou
–Legaki, 2009) and is better compared to strip transect
sampling, as detectability is properly accounted for. The
critical assumption of strip transects is that all individuals
present within the transect surface are detected. However,
this assumption can easily be violated in the marine envi-
ronment leading to substantial underestimation of popu-
lation density and abundance (Katsanevakis et al., 2012).
The imperfect detectability issue is overcome in line
transect sampling, where a standardized survey is con-
ducted along a series of lines searching for the animals of
interest. For each animal detected, the distance, y, from
the line or point is recorded. A detection function, g(y),
is fitted from the set of recorded distances (Buckland et
al., 2001, 2004), which is used to estimate the proportion
of animals missed by the survey and, hence, accurately
estimate abundance.
Eighteen transect locations were randomly placed
in the study area, 15 close to the shore and three in the
central part of the gulf. Sampling was conducted in the
summer of 2016. Sampling effort was focused in the
shallow coastal areas, as preliminary surveys indicated
the absence of fan mussels in the deeper muddy seabed.
Nearshore transects were defined vertically to the coast,
and oriented towards the centre of the gulf using a diving
compass. Transect length varied between 100 and 200
m, depending on the depth, diving conditions and diving
limitations. Each transect length (Lj) was defined with
a nylon line deployed using a diving reel. The line was
segmented at five meter intervals (hereafter called seg-
ments) with water resistant labels, and was marked with
water resistant paint at one-meter intervals. Depth mea-
surements were taken at the mid-point of each segment
with a dive computer. The habitat type was classified into
four basic categories (sandy, muddy, mixed, and Posido-
nia oceanica meadows) and the dominant category of
each segment was recorded. For each fan mussel obser-
vation, the following data were noted on diving slates:
the longitudinal distance from the start of the transect (lx),
the perpendicular distance from the line (ly) and shell size
(Si), defined as the maximum dorso-ventral length of the
shell. The perpendicular distances were measured with a
measuring tape (0.5 cm accuracy) and shell size with ver-
nier callipers (for widths >15 cm with an accuracy of 0.5
cm and for widths <15 cm with an accuracy of 0.05 cm).
For each transect, a visibility index was estimated empir-
ically: one of the two divers stood still while holding a
white board and the start of a measuring tape, while the
other receded slowly. When the board was barely visible,
the corresponding distance was considered as an index of
average visibility.
Detection function modelling
Two candidate models for the detection function,
g(y), were fitted, the one-parameter half-normal model
, and the two-parameter hazard-rate mod-
el , where σ is a scale param-
eter and b a shape parameter (Buckland et al., 2001). It
is possible to include covariates vj in these models, i.e.
variables that may affect detectability, through the scale
parameter σ, according to the equation:
where βi are estimable parameters (Marques & Buckland,
2004).
In this study, three covariates were considered as po-
tentially affecting detectability, namely, the size of fan
mussel individuals, habitat type and water visibility. The
hazard-rate and half-normal models were used with no,
one, two or three covariates. Thus, sixteen candidate
models mi (i = 1 to 16) were included in the set of candi-
date models for the detection function (Table 1). In mod-
els with an odd index, the half normal function was used,

645
Medit. Mar. Sci., 19/3, 2018, 642-655
whereas the hazard rate function was used in those with
an even index. In models m1 and m2, the σ parameter was
constant, while in the rest at least one covariate was in-
cluded.
The best model was selected using Akaike’s Infor-
mation Criterion (AIC; Akaike, 1973). Goodness-of-fit
of the best model was assessed with Q-Q plots and the
Cramér-von-Mises test, weighted to give higher weight
to distances near zero (Burnham et al., 2004). The Mul-
tiple Covariates Distance Sampling (MCDS) engine in
DISTANCE v7.0 (Thomas et al., 2010) was used for de-
tection function modelling.
Design-based approach for abundance estimation
In the design-based approach, inference was based
on the design characteristics of the survey, i.e. stratified
random sampling, and each transect was treated as a sam-
pling unit. Two strata were defined based on preliminary
observations that fan mussels were mostly restricted to
the nearshore zone. Towards the deeper part of the gulf,
model function covariate No.
of parameters PaΔi
population
density abundance 95% CI of
abundance
m1Half-normal - 1 0.580
± 0.070 15.19 0.0028 112800 64200-198400
m2Hazard-rate - 2 0.740
± 0.060 17.98 0.0022 88900 50900-155300
m3Half-normal size 2 0.585
± 0.055 16.98 0.0029 112900 64500-197500
m4Hazard-rate size 3 0.725
± 0.050 20.34 0.0024 90500 51800-158000
m5Half-normal visibility 2 0.575
± 0.055 12.67 0.0029 114300 65300-199900
m6Hazard-rate visibility 3 0.690
± 0.055 16.99 0.0024 95400 54600-166600
m7Half-normal habitat 3 0.560
± 0.060 2.31 0.0030 118000 67300-206700
m8Hazard-rate habitat 4 0.695
± 0.055 6.76 0.0025 94500 54100-165200
m9Half-normal visibility
& size 30.575
± 0.055 14.38 0.0029 114300 65300-200000
m10 Hazard-rate visibility
& size 40.690
± 0.055 18.89 0.0024 95600 54700-167100
m11 Half-normal habitat
& size 40.550
± 0.050 2.17 0.0030 119100 68000-208700
m12 Hazard-rate habitat
& size 50.700
± 0.055 9.34 0.0025 93800 53700-163800
m13 Half-normal habitat
& visibility 40.550
± 0.050 0.00 0.0030 120100 68500-210400
m14 Hazard-rate habitat
& visibility 50.720
± 0.060 10.29 0.0025 91500 52400-159800
m15 Half-normal
Habitat
& size
& visibility
50.545
± 0.055 0.73 0.0031 120900 69000-211900
m16 Hazard-rate
Habitat
& size
& visibility
60.695
± 0.060 6.86 0.0026 94800 54200-165800
Table 1. Parametrization of the 16 candidate models mi for the detection function, average probability of detection Pa (± SE),
Akaike differences Δi, estimated population density and abundance of P. nobilis in the study area, and 95% confidence intervals of
abundance (based on bootstrapping; 999 resamples). The best model is given in bold.

Medit. Mar. Sci., 19/3, 2018, 642-655
646
muddy sediments prevail and the substrate is unsuitable
for the survival of fan mussels. It has been extensively
documented that fan mussels are absent in muddy sedi-
ments (e.g. Katsanevakis, 2006). Hence, the first stratum
was defined as a 200-m buffer zone along the coastline,
while the second stratum included all the rest of the gulf
(Fig. 2). As zero densities were anticipated in the sec-
ond stratum, assuming that only muddy sediments occur,
the survey effort was relatively low, also due to logis-
tical constraints (need for support vessel, costs). In to-
tal, 15 transects were randomly defined in the first stra-
tum and three transects in the second stratum. ArcMAP
v10.2.2was used to define the two strata and estimate
their areas (A1, A2).
The total number of fan mussels within the covered
transects was estimated through the Horvitz-Thomp-
son-like estimator (Borchers, 1996) , where
is the probability of detecting individual j, and was ob-
tained from the estimated best model of the detection
function. Hence, the population density at each stratum h
(h = 1, 2) was estimated as , where Act,h is the
surface of the covered transects in stratum h. The stan-
dard error of the population density at each stratum was
estimated as , where Di is the es-
timated density at each transect. The overall population
density in the entire study area was estimated as
, where A = A1 + A2 is the total
study area. Total abundance was estimated as
, and the corresponding standard error was obtained from
the equation , where Wi
= Ai / A (Krebs, 1999).
Model-based approach for estimating abundance, spa-
tial distribution and habitat use
The second method applied for abundance estimation
was a model-based approach, as described by Katsane-
vakis (2007b) and Katsanevakis & Thessalou-Legaki
(2009). Specifically, the count method of Hedley & Buck-
land (2004) was applied; according to this method, the
transect lines are divided into smaller discrete units called
segments (of 5-m length), and the estimated number of
individuals in each segment is modelled by a Generalized
Additive Model (GAM; Hastie & Tibshirani 1990) using
explanatory spatial covariates. The Density Surface Mod-
elling (DSM) engine in DISTANCE v7.0 (Thomas et al.,
2010) was used for the model-based analysis.
Specifically, the total number of individuals within
each segment i was estimated using the Horvitz-Thomp-
son-like estimator (Hedley et al., 2004),
where was obtained from the best model of the detec-
tion function. These estimated values of abundance in
each segment were related to spatial covariates using the
general GAM formulation
,
where f is the link function, c is the intercept, sm(.) is the
1-dimentional smooth function for the predictor variable
m, zmi is the value of predictor variable m for segment i, Fr
are the categorical predictors, and As is the covered area
of the segment.
For this study, two spatial covariates were used: hab-
itat type as a categorical variable and depth as a contin-
uous variable. Both are considered very important for
predicting P. nobilis population density (Katsanevakis,
2007b; Katsanevakis & Thessalou-Legaki, 2009). A qua-
si-poison distribution and logarithmic link were used.
The latter ensures positive values for the mean response.
The smooth function sm(.) was represented using cubic
regression splines, estimated by penalized iterative least
squares (Wood, 2006). Four different GAM models were
created; h1 with no predictor, h2 with habitat type as pre-
dictor, h3 with depth as predictor, and h4 with both hab-
Fig. 2: Map of the study area, which was stratified to apply
the design-based method. Stratum 1 corresponds to the 200m
buffer zone from the coastline and Stratum 2 to the central part
of the gulf.

647
Medit. Mar. Sci., 19/3, 2018, 642-655
itat type and depth as predictors. The best GAM model
was chosen according to the generalized cross validation
(GCV) score (Wood, 2006). For this analysis, the DSM
and MRDS engines in DISTANCE v7.0 (Thomas, et al.
2010) and the package ‘mgcv’ (Wood, 2000, 2006) in R
v3.3.3 (R Core Team, 2015) were used.
For abundance predictions, the study area was seg-
mented into 64152 cells, measuring 25 x 25 m. For each
cell, the average depth and dominant habitat type were
estimated, according to the bathymetric and habitat maps.
For each cell, the abundance of fan mussels was
predicted using the best GAM model. The total abun-
dance of P. nobilis in Gera Gulf was estimated as the sum
of the predictions for all cells, i.e. . These
predictions were imported and visualized in a density sur-
face map of Gera Gulf using ArcMap v10.2.2.
Total variance was estimated by applying the del-
ta method (Seber, 1982), according to the equation
, where is
the coefficient of variation of the estimator of detection
probability and is the coefficient of variation
related to DSM. The first component was estimated em-
pirically (Buckland et al., 2001), while for the second
one a nonparametric bootstrap approach was followed, as
described in Katsanevakis & Thessalou (2009). No auto-
correlation was evident in the variogram of the deviance
residuals and thus the 5-m segment was selected as the
sampling unit for the bootstrapping.
Results
Bathymetry and habitat mapping
The gulf has a maximum depth of 19 m (29 m at the
channel) and is characterized by a relatively smooth
morphology down to ~11-12 m water depth (Fig. 3).
The steeper slope inclinations are encountered towards
the southeast, whereas the smoother relief appears at the
NNW side of the gulf. Between the ~12 and 19 m iso-
baths a peculiar microrelief occupies the seafloor in the
form of small hummocks that are distributed almost uni-
formly around the gulf. Their maximum height reaches 2
m in the south, close to the channel connecting Gera Gulf
to the open sea.
The analyzed satellite image enabled mapping of the
benthic habitats in the entire study area (Fig. 3), thanks
to the shallow depth, relatively transparent waters on the
day of acquisition, and very high image resolution. Im-
age classification allowed identification of the areas with
Posidonia oceanica meadows (a total area of 1.21 km2),
with a large meadow in the south-western part of the gulf,
narrow zones (0.10 km2) of mixed bottoms at various lo-
cations along the coastline, and extensive areas of sandy
(13.65 km2) and muddy (24.83 km2) sediments, the latter
covering the central part of the gulf.
Detection function modelling – design-based approach
for abundance estimation
The total sampling effort (i.e. total length of the tran-
sects) was 2800 m. Overall, 194 fan mussel individuals
Fig. 3: Bathymetry (left panel) and habitat map (right panel) of Gera Gulf.

Medit. Mar. Sci., 19/3, 2018, 642-655
648
were recorded at distances of up to 5.12 m from the tran-
sect line. No individual was found in stratum 2. Their size
(maximum width) varied between 3.97 and 19.65 cm and
had a bimodal distribution (Fig. 4). There was an appar-
ent segregation of size classes. Small individuals peaked
in shallow waters, while large individuals were less com-
mon in the shallow zone and peaked in the depth zone of
3.6-4.8 m (Fig. 4). Visibility varied between 1.5 and 7.0
m. Data were right-truncated at 4.2 m to avoid the effect
of outliers [and thus the covered area of each segment
was 5 m x (4.2 m x 2) = 42 m2].
Based on AIC, model m13 (half-normal with visibility
and habitat type as covariates) was the best amongst all
candidate models (Table 1). This model gave a good Q-Q
plot and provided a good absolute fit (Cramér-von Mis-
es test; p = 0.75). Model m15, which included visibility,
habitat type and size as covariates, was also substantially
supported by the data (Δ7 = 0.73) and produced a very
similar estimate of abundance (Table 1). The best model
(m13) is given by the equation:
where the distance from the line, y, is in cm, ‘visibili-
ty’ in m, and the two variables ‘mixed’ and ‘Posidonia’
are 1 (one), if the habitat is mixed or Posidonia oceanica
meadow respectively, and zero otherwise (Fig. 5).
Surprisingly, the detectability of fan mussels in Po-
sidonia oceanica meadows was better than the detectabil-
ity in mixed or sandy sediments (Fig. 5). This was largely
because of the within-transect variation of visibility. The
estimated index of visibility was an estimated average for
the wider area on the specific date. However, at different
habitats within each transect, visibility varied substantial-
ly. In Posidonia oceanica meadows, visibility was better
than in sandy areas where currents and the movement of
divers increased turbidity through resuspension of fine
sediment particles, and thus detectability was generally
low. Based on m13 and the design-based approach, Pin-
na nobilis abundance in Gera Gulf was estimated to be
, with a 95% confidence interval of 68500–
210400 individuals, exclusively in Stratum 1.
Density surface models – GAMs
According to the GSV score, h4 that included
both depth and habitat type as predictor variables,
was the best model for DSM (Table 2). The expres-
sion of h4 is , where
, , while the
smooth function for depth s(d) and the categorical pre-
dictor F(H) are given in Fig. 6. Population density was
Fig. 4: (A) Size distribution of all recorded Pinna nobilis individuals in Gera Gulf; (A–C) Bathymetric distributions of small (shell
width <9.9 cm), medium (shell width between 9.9–11.6 cm) and large (shell width >11.6 cm) fan mussels.

649
Medit. Mar. Sci., 19/3, 2018, 642-655
higher in shallow waters between 1.5 and 8 m depth, de-
clined in very shallow areas <1.5 m or at depths >8 m,
and became practically zero at depths >15 m. The highest
population densities were observed in Posidonia oceani-
ca meadows, followed by mixed bottoms, while densities
were very low on sandy and zero on muddy sediments
(Fig. 6).
Based on h4, Pinna nobilis abundance in Gera Gulf
was estimated to be , with a 95% confidence
interval between 97600–466000 individuals. Moreover, a
density surface map of Gera Gulf (Fig. 7) was produced
based on h4. In the south-eastern part of the study area,
fan mussels were restricted to a very narrow nearshore
zone, whereas in the western part of the study area fan
mussels were distributed throughout a much wider zone
(Fig. 7). The highest predicted densities coincided with
the P. oceanica patches.
Discussion
Comparison of the two approaches
The point estimate of abundance of P. nobilis in Gera
Gulf using the design-based approach was much lower
than the point estimate using the model-based approach.
The main reason for this discrepancy was that the initial
assumption of the design-based approach – that in the
entire area beyond the 200-m buffer zone muddy sedi-
ments prevail and thus the substrate is inappropriate for
fan mussels – is not true. In fact, as revealed by habitat
mapping, the non-muddy area used in the model-based
approach is almost double the size of the non-muddy
area assumed in the design-based approach (Stratum 1).
Hence, there were substantial areas suitable for P. nobilis
in Stratum 2, which were missed by the limited sampling
effort in that stratum, thus resulting in substantial under-
estimation of abundance.
Due to logistical constraints, habitat mapping was not
Fig. 5: Best detection function (model m13) of P. nobilis for the three habitat types in which individuals were detected (Posidonia
oceanica meadows, mixed bottoms, and sandy bottoms). For each habitat type, two curves are depicted, corresponding to the
minimum and maximum observed visibility at each habitat type.
Table 2. Evaluation of the 4 candidate GAMs for the population density of P. nobilis in the Gera Gulf, based on their generalized
cross validation (GCV) score. The percentage of deviance explained by each model and the abundance estimation in the study
area are provided.
Model Spatial Covariate CV Score Deviance explained
(%)
Abundance
estimation
h1- 555.5 0% 226600
h2habitat type 486.5 18.6% 277400
h3depth 507.5 16.7% 175700
h4habitat type + depth 460.8 26.2% 213300

Medit. Mar. Sci., 19/3, 2018, 642-655
650
available before survey design. Otherwise, the distribu-
tion of the main habitat types would have been used as
the basis for stratification. In the absence of habitat map-
ping, the initial observation that P. nobilis is restricted
nearshore was used as the basis for stratification. There
are many potential advantages of stratification of the
study area based on subjective information and previous
knowledge, such as reducing the survey cost and un-
certainty in the estimates (Krebs, 1999; Morrison et al.,
2001). In our case, this has been proved to be a problem-
atic approach as Stratum 2 was under-sampled, due to our
belief of zero abundance, and the related fan mussel pop-
ulation was underestimated. When dealing with sparsely
distributed individuals over large areas, it is not uncom-
mon to find a larger proportion of the population in the
“low-density” stratum than in the “high-density” strata or
“preferred” areas, as the low density is often multiplied
by a huge area (McDonald, 2004).
The model-based approach was advantageous not
only for making more accurate abundance estimates but
also because it provides additional information on the
spatial distribution of the species and its habitat use. The
precision of the abundance estimate by the model-based
approach could be greatly improved by stratifying the
study area according to the habitat types. In that respect,
this study can serve as a baseline for future monitoring of
the species and for improving sampling design.
Effect of depth and habitat type on the distribution of
fan mussels
When analyzing shelf assemblages, depth is the main
gradient along which faunal changes occur (e.g. Bianchi,
1992; Demestre et al., 2000; Katsanevakis et al., 2009).
This is less due to a direct effect of depth (because of the
increase of pressure) but mostly due to the correlation of
depth with many crucial environmental parameters such
as bottom substratum, hydrodynamics, light intensity,
temperature, primary and secondary productivity.
The pattern of bathymetric variation of fan mussel
density found in this study was that of high densities
at depths between 1.5 and 8m, lower densities < 1.5m
or > 8m, and zero densities below 15 m. Similar results
have been found in other studies, although density peak-
ed deeper and, overall, the density-depth curve shifted at
higher depths. Such data are available for two other ar-
eas in Greece, namely, Lake Vouliagmeni (Katsanevakis,
2007b) and Souda Bay (Katsanevakis &Thessalou-Lega-
ki, 2009). In Lake Vouliagmeni, there was a main peak of
population density at depths of 12–13 m, reduced density
in very shallow waters, and practically zero densities at
depths >22 m (Katsanevakis, 2007b). In Souda Bay, there
was a density peak at a depth of ~15 m and practically
zero densities in shallow areas (<4 m depth) and at depths
>30 m. In the Cabrera National Park (Balearic Islands,
Spain), the density peak was found at 9 m (Vázquez-Luis
et al., 2014) and although density declined with depth,
Fig. 6: Estimated smooth term s(d) (depth) and the categorical predictor H (habitat type), for model h4 of fan mussel abundance in
5x5 m plots in Gera Gulf. In the upper panels, the terms are given in the linear predictor scale and the respective 95% confidence
intervals are given with dotted lines. In the lower panels, the terms are given in the response scale (exp-transformed). At the bottom
of each graph there is a 1-dimentional scatter plot illustrating the distribution of available data.

651
Medit. Mar. Sci., 19/3, 2018, 642-655
fan mussels were found even at 46 m. In Tunisia, in a
study conducted at a depth range of 0 to 6 m, Rabaoui et
al. (2010) predicted a density of practically zero at 0.3
m depth, increasing with depth; in the absence of deep-
er transects, the depth of the peak was unknown. These
differences in the bathymetric distribution of the species
among studies are due to the local conditions of each area.
Two main factors seem to restrict fan mussel popula-
tions in very shallow waters, wave action (García-March
et al., 2007) and poaching by free divers (Katsanevakis
2007a). According to García-March et al. (2007), wave
action causes increased mortality and chronic levels of
hydrodynamic stress, which substantially decreases with
depth, and thus the selective pressure on the population
is the highest in very shallow waters. In addition, poach-
ing by free divers causes a selectively higher mortality
in shallow waters, especially for large individuals, which
may greatly affect fan mussel densities and the structure
of the population (Katsanevakis, 2007a). Poaching on fan
mussels can be severe, greatly affecting their population
dynamics and causing a size segregation of individuals,
with larger and older individuals restricted to deeper ar-
eas and smaller and younger individuals dominating in
shallow waters (Katsanevakis, 2009). These factors have
probably contributed to the size segregation of fan mus-
sels observed in Gera Gulf.
P. nobilis was absent from the deeper muddy bot-
tom of Gera Gulf; this is in agreement with the studies
in Lake Vouliagmeni (Katsanevakis, 2007b) and Souda
Bay (Katsanevakis &Thessalou-Legaki, 2009). The main
problem is that fan mussels cannot anchor adequately, in
a fixed vertical position, in muddy sediment as they can
easily sink into the sediment because of the movement
of their valves. Furthermore, high silt content may have
negative effects on respiration and feeding (Thorson,
1950; Cheung & Shin, 2005). Fan mussels lack siphons
but instead have an open pallial cavity, which offers them
a fairly high pumping rate, but at the cost of high vulner-
ability to the entry of sediments (Butler et al., 1993). This
explains the absence of P. nobilis from muddy areas and,
in general, areas of severe sediment disturbance, where
only siphonate infaunal bivalves may thrive (Butler et al.,
1993).
In Gera Gulf, Pinna nobilis reached its highest density
in Posidonia oceanica meadows. Lower densities were
observed in mixed and sandy habitats. This observation
concurs with the widely reported fidelity of P. nobilis for
P. oceanica seagrass meadows (e.g. Rabaoui et al., 2010;
Vázquez-Luis et al., 2014) and other vegetated habitats,
such as beds of the seagrasses Cymodocea nodosa and
Halophila stipulacea or the green alga Caulerpa cylind-
racea (Katsanevakis &Thessalou-Legaki, 2009). Nev-
ertheless, high densities are also found on unvegetated
bottoms, especially in areas of low hydrodynamism, such
as Lake Vouliagmeni (Katsanevakis 2006). The main fac-
tors of a “preference” for seagrass meadows seem to be
protection from intense hydrodynamism, good substrate
for anchoring, lower mortality caused by predators, and
limited poaching by free divers. Seagrass beds dissipate
wave energy and attenuate flow (Hendriks et al., 2007),
thus reducing the drag experienced by P. nobilis, which
thrives within seagrass canopies (Hendriks et al., 2011).
Hence, seagrass beds have a sheltering effect on fan mus-
sels, as hydrodynamic stress and mortality caused by
storms is reduced in comparison to unvegetated bottoms.
Furthermore, the robust network of rhizomes in seagrass
beds provides firm anchoring points for fan mussels
through their byssus threads (Basso et al., 2015). In addi-
tion, fan mussels, especially juveniles, are less vulnerable
to predation as they are well camouflaged in a seagrass
canopy. Similarly, it is more difficult for poachers to spot
fan mussels living on seagrass beds than individuals on
unvegetated bottoms, where poaching may lower fan
mussel populations (Katsanevakis, 2007a).
Significance of P. nobilis population in the Gera Gulf
The average population density estimated in Gera
Gulf is 5.3 individuals per 1000 m2, which is very similar
to the densities estimated for the other two assessed popu-
lations in Greece, namely those of Lake Vouliagmeni (5.7
individuals per 1000 m2) and Souda Bay (8.9 individuals
per 1000 m2). However, much higher average densities,
by 1 to 2 orders of magnitude, have been recorded in the
Mediterranean (Table 3). In Lake Vouliagmeni, there was
evidence of very high population densities in the past
(see Supplementary file of Katsanevakis, 2016), ~3 to 4
orders of magnitude higher than the current population
Fig. 7: Pinna nobilis population density map based on the mod-
el-based approach and on density model h4.

Medit. Mar. Sci., 19/3, 2018, 642-655
652
densities, i.e. thousands or tens of thousands of individ-
uals per 1000 m2. It has been indicated that poaching is
one of the main reasons for the low observed densities
in Lake Vouliagmeni (Katsanevakis, 2007a, 2009). An-
ecdotal information suggests that the level of poaching
in Gera Gulf is substantial, and fan mussels are contin-
uously illegally fished and even served in local seafood
restaurants. The fact that large individuals were scarce
and no individual with width > 19.65 cm was detected,
in contrast to Lake Vouliagmeni, where there were many
larger individuals (Katsanevakis, 2006), could be due to
higher mortality rates or lower growth rates in Gera Gulf
(but targeted investigation is needed to reach any solid
conclusions).
Nevertheless, the P. nobilis population of Gera Gulf
is the largest recorded population in Greece, followed
Table 3. Average population densities of Pinna nobilis in various Mediterranean sites (modified and updated from Rouanet et al.,
2015; Katsanevakis, 2016).
Location
Average population
density (individuals
/1000 m2)
Source
Port-Cros Island (Port-Cros National Park, MPA), Pro-
vence, France 10 Vicente et al., 1980; Combelles et al., 1986
Scandola marine reserve (MPA), Corsica 10 Combelles et al., 1986
Croatia, Adriatic Sea 90 Zavodnik et al., 1991
Chafarinas Islands, Spain, Northern Africa 32 Guallart, 2000
Scandola marine reserve (MPA, NTZ), Corsica 60 Charrier et al.
(mentioned in Rouanet et al.,2015)
Mljet National Park (MPA), Croatia, Adriatic Sea 20-200 Šiletić&Peharda, 2003
Murcia, Almeria and Balearic Islands,
Spain 100 García-March, 2003
Lake Vouliagmeni, Greece 5.7 Katsanevakis, 2006
Columbretes marine reserve (MPA), Castellón, Commu-
nitatvalenciana, Spain 15 García-March &Kersting, 2006
Mar Grande of Taranto, Ionian Sea, Italy 0-0.07 Centoducati et al., 2007
Souda Bay, Crete Island, Greece 8.9 Katsanevakis &Thessalou-Legaki, 2009
Port-Cros Island (Port-Cros National
Park, MPA), Provence, France 20-80 Vicente, 2009
Porquerolles Island, Provence, France 2-23 Vicente, 2009
Scandola marine reserve (MPA, NTZ), Corsica 140 Vicente, 2010
Tunisia (east and southeast coast) 15 Rabaoui et al., 2010
Pass between Bagaud and Port-Cros Islands (Port-Cros
National Park, MPA), Provence, France 60-130 Rouanet et al., 2012
Embiez Island, Six-Fours-les-Plages, Provence, France 19 Trigos et al., 2013
Cabrera National Park MPA, Majorca Island, Spain 38 Vázquez-Luis et al., 2014
Javea, Alicante, Spain <10 García-March, pers. comm. (mentioned in
Rouanet et al., 2015)
Moraira, Alicante, Spain 10-120 García-March, pers. comm. (mentioned in
Rouanet et al., 2015)
West Sardinia, Italia 41 Coppa et al., 2015
Mar Menor, Spain 22 Belando et al., 2015
Harbour bay of Favignana island, Italy 110 D’agostaro et al., 2015

653
Medit. Mar. Sci., 19/3, 2018, 642-655
by the population of Souda Bay, which was estimated at
139000 individuals (95% CI: 100600–170400). As Gera
Gulf is part of the Natura 2000 network of protected ar-
eas, contrary to all other known areas in the Aegean Sea
with important fan mussel populations, its importance for
the conservation of the species is high.
Concluding remarks
The key message from the comparison of the de-
sign-based and the model-based approach is that in stud-
ies of animal abundance caution is needed when deciding
to stratify the study area, especially if the prior informa-
tion used for stratification is of low quality. In any case,
sufficient sampling effort should also be focused on the
assumed “low-density” strata, as total abundance there
might end up being of the same order of magnitude or
higher than in the “high-density” strata.
Wave action and poaching have probably contributed
to the size segregation of fan mussels observed in Gera
Gulf, with large individuals being less common in the
very shallow zone. The limited Posidonia oceanica beds
of Gera Gulf largely act as refuge areas for fan mussels,
protecting them from poaching, predation and intense
hydrodynamism. Further research is needed to assess the
level of impact of poaching at population level, as the
analogy with other well-studied areas (i.e. Lake Vouliag-
meni) suggests that population-level impacts of increased
fishing mortality are quite probable. Better law enforce-
ment to confront poaching on the species, and additional
management actions for the protection of the species (see
e.g. Katsanevakis, 2006, 2007a) and its preferred habitats
are needed to conserve this important population, which
may act as a source for neighbouring areas through larvae
spill-over.
It is of great importance to continue monitoring the
fan mussel population of Gera Gulf in the future. Only
with regular monitoring and additional studies will it be
possible to detect population trends and understand the
dynamics of the species. In particular, in view of the
ongoing massive mortality of the species in the western
Mediterranean, urgent adaptation of monitoring plans
to detect mass mortality events in all Mediterranean fan
mussel populations and identify resistant individuals has
been suggested (Vázquez-Luis et al., 2017).
References
Addis, P., Secci, M., Brundu, G., Manunza, A., Corrias, S.
et al., 2009. Density, size structure, shell orientation and
epibiontic colonization of the fan mussel Pinna nobilis L.;
1758 (Mollusca: Bivalvia) in three contrasting habitats in
an estuarine area of Sardinia (W Mediterranean). Scientia
Marina, 73, 143-152.
Akaike, H., 1973. Information theory as an extension of the
maximum likelihood principle. p. 267-281. In: Pro 2nd In-
ternational Symposium on Information Theory, Tsahkadsor,
Armenia, USSR, 2-8 September 1971. Akademiai Kiado,
Budapest.
Basso, L., Vázquez-Luis, M., García-March, J., Deudero, S.,
Alvarez, E. et al., 2015. The pen shell, Pinna nobilis: A re-
view of population status and recommended research pri-
orities in the Mediterranean Sea. p. 109-160. In: Advances
in Marine Biology. Barbara E., C. (Ed). Elsevier Science
Publishing Co Inc, San Diego, United States.
Belando, M.D., García-Muñoz, R., Ramos, A., Franco, I., Ber-
nardeau-Esteller, J. et al., 2015. Distribution and abundance
of Cymodocea nodosa meadows and Pinna nobilis popu-
lations in the Mar Menor coastal lagoon (Murcia, South
East of Spain). PeerJ Pree Prints. 3:e1063v1 https://doi.
org/10.7287/peerj.preprints.1063v1
Bianchi, G., 1992. Demersal assemblages of the continental
shelf and upper slope of Angola. Marine Ecology Progress
Series, 81, 101-120.
Blaschke, T., 2010. Object based image analysis for remote
sensing. ISPRS Journal of Photogrammetry and Remote
Sensing, 65 (1), 2-16.
Borchers, D.L., 1996. Line transect estimation with uncertain
detection on the trackline. PhD Thesis. University of Cape
Town, South Africa, 233pp.
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L.,
Borchers, D.L. et al., 2001. Introduction to distance sam-
pling: estimating abundance of biological populations. Ox-
ford University Press, New York, 448 pp.
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L.,
Borchers, D.L. et al., 2004. Advanced distance sampling:
estimating abundance of biological populations. Oxford
University Press, New York, 434pp.
Burnham, K.P., Buckland, S.T., Laake, J.L., Borchers, D.L.,
Marques, T.A. et al., 2004. Further topics in distance sam-
pling. p. 307-392. In: Advanced distance sampling: estimat-
ing abundance of biological populations. Buckland, S.T.,
Anderson, D.R., Burnham, K.P., Laake, J.L., Borchers,
D.L., Thomas, L. (Eds). Oxford University Press, New
York.
Butler, A.J., Vincente, N., Gaulejac, B., 1993. Ecology of the
pteroid bivalves Pinna bicolor Gmelin and Pinna nobilis.
Marine Life, 3, 37-45.
Catanese, G., Grau, A., Valencia, J.M., García-March, J.M.,
Álvarez, E. et al., 2018. Haplosporidium pinnae sp.nov., a
haplosporidan parasite associated with massive mortalities
of the fan mussel, Pinna nobilis, in the Western Mediter-
ranean Sea. Journal of Invertebrate Pathology, 157, 9-24.
Centoducati, G., Tarsitano, E., Bottalico, A., Marvulli, M., Lai,
O.R., Crescenzo, G., 2007. Monitoring of the endangered
Pinna nobilis Linné, 1758 in the Mar Grande of Taranto
(Ionian Sea, Italy). Environmental Monitoring and Assess-
ment, 131, 339-347.
Cheung, S.G., Shin, P.K.S., 2005. Size effects of suspended
particles on gill damage in green-lipped mussel Pernavir-
idis. Marine Pollution Bulletin, 51 (8-12), 801-810.
Chiles, J.P.,Delfiner, P., 1999. Geostatistics: Modeling Spatial
Uncertainty. John Wiley and Sons. New York, 734pp.
Combelles, S., Moreteau, J.C., Vicente, N., 1986. Contribution
à la connaissance de l’écologie de Pinna nobilis L. (Mollu-
sque : Eulamellibranche). Scientific Reports of Port-Cros
National Park,12, 29-43.
Coppa, S., Cucco, A., De Falco, G., Massaro, G., Camedda,

Medit. Mar. Sci., 19/3, 2018, 642-655
654
A. et al., 2015. Pinna nobilis within a Posidonia oceani-
ca meadow: evidences of how hydrodynamics define this
association in the Gulf of Oristano (West Sardinia, Italy).
PeerJ PrePrints 3:e1074v1. https://doi.org/10.7287/peerj.
preprints.1074v1
D’agostaro, R., Donati, S., Chemello, R., 2015. Density and
distribution patterns of the endangered species Pinna nobil-
is within the harbour bay of Favignana (Egadi Islands MPA)
PeerJ Pree Prints 3: e1552v2 https://doi.org/10.7287/peerj.
preprints.1552v2
Demestre, M., Sánchez, P., Abelló, P., 2000. Demersal fish
assemblages and habitat characteristics on the continental
shelf and upper slope of the north-western Mediterranean.
Journal of the Marine Biological Association of the United
Kingdom, 80, 981-988.
García-March, J.R., 2003. Contribution to the knowledge of
the status of Pinna nobilis (L.) 1758 in Spanish coasts.
Mémoires del’Institut Océanographique de Paul Ricard, 9,
29-41.
García-March, J.R., Kersting, D.K., 2006. Preliminary data
on the distribution and density of Pinna nobilis and Pinna
rudis in the Columbretes Islands Marine Reserve (Western
Mediterranean, Spain). Organisms Diversity & Evolution,
6 (suppl 16), 33-34.
García-March, J.R., Pérez-Rojas, L., García-Carrascosa, A.M.,
2007. Influence of hydrodynamic forces on population
structure of Pinna nobilis L., 1758 (Mollusca: Bivalvia):
The critical combination of drag force, water depth, shell
size and orientation. Journal of Experimental Marine Biol-
ogy and Ecology, 342, 202-212.
Guallart, J., 2000. Seguimiento de Pinna nobilis. p 480–489.
In: Control y Seguimiento de los Ecosistemas del R.N.C.
de las Islas Chafarinas. Informe GENA S.L. para O.A.P.N.
(Ministerio de Medio ambiente), Madrid.
Hastie, T.J., Tibshirani, R.J., 1990. Generalized additive models
(monographs on statistics and applied probability). Chap-
man & Hall, London, 352pp.
Hedley, S.L., Buckland, S.T., 2004. Spatial models for line
transect sampling. Journal of Agricultural, Biological, and
Environmental Statistics, 9, 181-199.
Hedley, S.L., Buckland, S.T., Borchers, D.L., 2004. Spatial
distance sampling models. p. 48–70. In: Advanced distance
sampling: estimating abundance of biological populations.
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake,
J.L., Borchers, D.L., Thomas, L. (Eds). Oxford University
Press, New York.
Hendriks, I.E., Sintes, T., Bouma, T., Duarte, C.M., 2007. Ex-
perimental assessment and modeling evaluation of the ef-
fects of seagrass (P. oceanica) on flow and particle trap-
ping. Marine Ecology Progress Series, 356, 163-173.
Hendriks, I.E., Cabanellas-Reboredo, M., Bouma, T.J., Deude-
ro, S., Duarte, C.M., 2011. Seagrass meadows modify drag
forces on the shell of the fan mussel Pinna nobilis. Estua-
rine, Coastal and Shelf Science, 34, 60-67.
Katsanevakis, S., 2006. Population ecology of the endangered
fan mussel Pinna nobilis in a marine lake. Endangered Spe-
cies Research, 1, 51-59.
Katsanevakis, S., 2007a. Growth and mortality rates of the fan
mussel Pinna nobilis in Lake Vouliagmeni (Korinthiakos
Gulf, Greece): a generalized additive modelling approach.
Marine Biology, 152 (6), 1319-1331.
Katsanevakis, S., 2007b. Density surface modelling with line
transect sampling as a tool for abundance estimation of ma-
rine benthic species: the Pinna nobilis example in a marine
lake. Marine Biology, 152 (1), 77-85.
Katsanevakis, S., 2009. Population dynamics of the endangered
fan mussel Pinna nobilis in a marine lake: a metapopulation
matrix modelling approach. Marine Biology, 156 (8), 1715-
1732.
Katsanevakis, S., 2016. Transplantation as a conservation ac-
tion to protect the Mediterranean fan mussel Pinna nobilis.
Marine Ecology Progress Series, 546, 113-122.
Katsanevakis, S., Thessalou-Legaki, M., 2009. Spatial distri-
bution and abundance of the endangered fan mussel Pinna
nobilis in Souda Bay (Crete Island, Greece). Aquatic Biol-
ogy, 8, 45-54.
Katsanevakis, S., Maravelias, C.D., Damalas, D., Karageorgis,
A.P., Anagnostou, C. et al., 2009. Spatiotemporal distribu-
tion and habitat use of commercial demersal species in the
eastern Mediterranean Sea. Fisheries Oceanography, 18
(6), 439-457.
Katsanevakis, S., Poursanidis, D., Issaris, Y., Panou, A., Petza,
D. et al., 2011. ‘Protected’ marine shelled molluscs: thriv-
ing in Greek seafood restaurants. Mediterranean Marine
Science, 12, 429-438.
Katsanevakis, S., Weber, A., Pipitone, C., Leopold, M., Cronin,
M. et al., 2012. Monitoring marine populations and com-
munities: review of methods and tools dealing with imper-
fect detectability. Aquatic Biology, 16, 31-52.
Krebs, C.J., 1999. Ecological Methodology. Benjamin/Cum-
mings, Menlo Park, 620pp.
Marques, F.F.C., Buckland, S.T., 2004. Covariate models for
the detection function. p. 31-47. In: Advanced distance
sampling: estimating abundance of biological populations.
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake,
J.L., Borchers, D.L., Thomas, L. (Eds). Oxford University
Press, New York.
Mazaris, A.D., Almpanidou, V., Giakoumi, S., Katsanevakis,
S., 2018. Gaps and challenges of the European network
of protected sites: spatial characteristics and structural
complexities of marine areas. ICES Journal of Marine Re-
search, 75, 190-198.
McDonald, L.L., 2004. Sampling rare populations. p. 11-42. In:
Sampling rare or elusive species. Thompson, W.L. (Ed). Is-
land Press, Washington.
Morrison, M.L., Block, W.M., Strickland, M.D., Kendall, W.L.,
2001. Wildlife study design. Springer -Verlag, New York,
386pp.
Rabaoui, L., Tlig-Zouari, S., Katsanevakis, S., Ben Hassine,
O.K., 2010. Modelling population density of Pinna nobilis
(Mollusca: Bivalvia) in the eastern and south-eastern Tu-
nisian coasts. Journal of Molluscan Studies, 76, 340-347.
R Core Team, 2015. R: A Language and Environment for Statis-
ticalComputing. Vienna: RFoundation for StatisticalCom-
puting. Available at https://www.R-project.org/.
Rouanet, E., Astruch, P., Bonhomme, D., Bonhomme, P., Ro-
geau, E. et al., 2012. Suivi de l’herbier de Posidonie de la
passe de Bagaud, impact de l’ancrage (Parc national de Port-
Cros, Var, France). Partenariat Parc national de Port-Cros -
GIS Posidonie, GIS Posidonie publ., Marseille, 81 pp.

655
Medit. Mar. Sci., 19/3, 2018, 642-655
Rouanet, E., Trigos, S., Vicente, N., 2015. From youth to death
of old age: the 50-year story of a Pinna nobilis fan mussel
population at Port-Cros Island (Port-Cros National Park,
Provence, Mediterranean Sea). Scientific Reports of Port-
Cros National Park, 29, 209-222.
Seber, G.A.F., 1982. The estimation of animal abundance and
related parameters. Macmillan, New York, 654pp.
Thomas, L., Buckland, S.T., Rexstad, E.A., Laake, J.L., Strind-
berg, S. et al., 2010. Distance software: design and analysis
of distance sampling surveys for estimating population size.
Journal of Applied Ecology, 47, 5-14.
Thorson, G., 1950. Reproductive and larval ecology of marine
bottom invertebrates. Biological Reviews, 25 (1), 1-45.
Trigos, S., Vicente, N., García-March, J.R., Jímenez, S., Torres,
J. et al., 2013. Presence of Pinna nobilis and Pinna rudisin
the Marine Protected Areas of the North Western Mediter-
ranean. 3rd International Marine Protected Areas Congress
(IMPAC3), 21-27 October 2013, Marseille.
Vázquez-Luis, M., March, D., Alvarez, E., Alvarez-Berastegui,
D., Deudero, S., 2014. Spatial distribution modelling of the
endangered bivalve Pinna nobilis in a Marine Protected
Area. Mediterranean Marine Science, 15 (3), 626-634.
Vázquez-Luis, M., Borg, J.A., Morell, C., Banach-Esteve, G.,
Deudero, S., 2015. Influence of boat anchoring on Pinna
nobilis: a field experiment using mimic units. Marine and
Freshwater Research, 66, 786-794.
Vázquez-Luis, M., Álvarez, E., Barrajón, A., García-March,
J.R., Grau, A. et al., 2017. S.O.S. Pinna nobilis: a mass
mortality event in western Mediterranean Sea. Frontiers in
Marine Science, 4:220. doi: 10.3389/fmars.2017.00220.
Vicente, N., 2009. Poursuite de l’inventaire des populations de
Pinnanobilis sur les sites de Port-Cros et de Porquerolles.
Rapport Parc National de Port-Cros, Hyères, 1-35.
Vicente, N., 2010. Inventaire de Pinna rudis et comparaison
avec les densités de Pinna nobilis dans la Réserve Natu-
relle de Scandola. Contrat n°867/08, Rapport final, Office
de l’Environnement de Corse, Corte, 31pp.
Vicente, N., Moreteau, J.C., Escoubet, P., 1980. Etude de l’évo-
lution d’une population de Pinna nobilis L. (Mollusque
Eulamelibranche) au large de l’anse de la Palud (Parc natio-
nal de Port-Cros). Scientific Reports of Port-Cros National
Park, 6, 39-67.
Wood, S.N., 2000. Modelling and smoothing parameter estima-
tion with multiple quadratic penalties. Journal of the Royal
Statistical Society - Series B Statistical Methodology, 62,
413-428.
Wood, S.N., 2006. Generalized additive models: an introduc-
tion with R (texts in statistical science). Chapman & Hall/
CRC, Boca Raton, FL.
Zavodnik, D., Hrs-Brenko, M., Legac, M., 1991. Synopsis on
the fan shell Pinna nobilis L. in the eastern Adriatic Sea. p.
169-178. In: Les Espèces Marines à Protéger en Méditerra-
née. Boudouresque, C.F., Avon, M., Gravez, V. (Eds). GIS
Posidonie, Marseille.