Emigration and retention of Palinurus elephas (Fabricius, 1787) in a central western Mediterranean marine protected area

During the last century a large number of marine reserves were established around the world (Jones et al., 1993) in an attempt to halt further deterioration of sensitive habitats or to offer an alternative to traditional management options (Sanchez Lizaso et al., 2000). In the last decade a few efforts have been made to show that the reserves (also called marine protected areas (MPAs) or no-take areas, where all forms of fishing are prohibited) promote an increase in abundance as well as in the mean size of the proSCIENTIA MARINA 71(2) June 2007, 279-285, Barcelona (Spain) ISSN: 0214-8358


INTRODUCTION
During the last century a large number of marine reserves were established around the world (Jones et al., 1993) in an attempt to halt further deterioration of sensitive habitats or to offer an alternative to tradi-tional management options (Sanchez Lizaso et al., 2000).In the last decade a few efforts have been made to show that the reserves (also called marine protected areas (MPAs) or no-take areas, where all forms of fishing are prohibited) promote an increase in abundance as well as in the mean size of the pro-RESUMEN: EMIGRACIÓN Y RETENCIÓN DE PALINURUS ELEPHAS (FABRICIUS, 1787) EN UN ÁREA MARINA PROTEGIDA DEL MEDITERRÁNEO CENTRAL OCCIDENTAL.-Este estudio presenta los resultados de aplicar el modelo multiestado de marcado y recaptura de Arnason Schwartz a una serie de ocho años de datos recolectados en el interior y alrededor de una pequeña área marina protegida (AMP) vedada a la pesca (4 km 2 ) del Mediterráneo central occidental.Desde 1997 a 2004, un total de 4044 ejemplares de Palinurus elephas (Fabr., 1787) fueron marcados, de los cuales 317 fueron recapturados.El modelo más parsimonioso que explicó mejor la variabilidad de los datos fue aquel con una tasa temporal constante de aparente supervivencia y movimiento entre los dos estratos.La ausencia de influencia temporal sobre la tasa de supervivencia aparente en el interior del área protegida, sugirió que el "spillover" y la mortalidad son constantes para cada periodo del estudio.La menor tasa aparente de supervivencia en zonas alrededor de la reserva respecto al interior de la misma (0.26 ± 0.04 (SE) vs 0.94 ± 0.03 (SE)) se considera que es una función del esfuerzo pesquero.Un movimiento continuo de P. elephas a través de los límites de la pequeña AMP fue evaluado.Esta información sobre la retención de langostas en el AMP contribuye a comprender el efecto de la introducción de AMPs en un sistema regulado de pesquería comercial.
tected populations.The effect of a reduction of fishing on density and biomass of fish populations has been thoroughly investigated both in the Mediterranean and in other marine regions (Sánchez Lizaso et al., 2000).Most studies conclude that fishery reserves can increase the fish stocks in neighbouring areas, especially through migration of adults (Dugan and Davis, 1993;Roberts and Polunin, 1993;Rakitin and Kramer, 1996;Guenette et al., 1998).
At present, it is not possible to establish from the published data whether the exporting of specimens from the "source" population to fished areas is a response to resource limitation inside the protected area or only of random movements of adult individuals across the MPA boundaries (Beverthon and Holt, 1957;Kramer and Chapman, 1999).Some studies have only demonstrated that fish or invertebrates tagged and released inside MPAs may be caught outside them (Davis andDodrill, 1980, 1989;Gitschlag, 1986;Yamasaki and Kuwahara, 1990;Hunt et al., 1991;MacDiarmid and Breen, 1993;Attwood and Bennett, 1994;Bohnsack, 1998;Goñi et al., 2006).Some authors reported emigration from reserves to adjacent fished areas (Attwood and Bennett, 1994;Zeller et al., 1998;Johnson et al., 1999;Cole et al., 2000;Martell et al., 2000), while others described bidirectional movements with con-trasting results (Davis and Dodrill, 1989;Rowe, 2001;Zeller et al., 2003;Kelly and MacDiarmid, 2003;Tremain et al., 2004) depending on the species and their lifestage, and on the size of the MPAs (Goñi et al., 2006).Recently, Goñi et al. (2006) used a combination of tag recapture methods, fishing surveys and commercial fishery data to demonstrate the existence of a negative gradient of lobster density up to 4.5 km from the Columbretes Islands Marine Reserve.
The movement of protected species from marine areas, although predicted by theory and modelling (DeMartini, 1993;Man et al., 1995), has rarely been tested and assessed (Russ and Alcala, 1989;McClanahan and Kaunda-Arara, 1996;Hobday et al., 2005;Goñi et al., 2006).Effective tests of the movements are hampered by the difficulty of achieving a good study design and also the lack of a tagged-fish catch time series and fish recovery data (Roberts and Polunin, 1993;McClanahan and Kaunda-Arara, 1996).
The present paper describes an eight-year study, using mark recapture techniques, designed to assess the survival, resighting and movement rate of the spiny lobster, Palinurus elephas, in an MPA and its adjacent fished zone located in the central western Mediterranean.

Field methods
The study was carried out in an area of centralwestern Sardinia (central-western Mediterranean) at a depth of 50 to 100 m (Fig. 1).
This area, first identified in 1997, was chosen as an MPA on account of its geomorphological and bionomic characteristics.It is characterised by formations comparable to coastal pre-coralligenous and coralligenous detritus (Peres and Picard, 1964).Beginning from 1998, fishing was prohibited in the ca.4-km 2 area (Regional Law No. 776 of 6-5-1998) (Secci et al., 1999).
From 1997 to 2004, in the period from May to September, an average of 10 fishing samples were carried out each year (except for 2004, when there were only four), following a sampling plan with transects set up in such way as to enable the whole study area to be investigated.These samples were conducted using trammel nets of 1000 m length (nominal mesh from 50 to 73 mm).All the specimens caught inside the area (total 529) and a sample of 3515 individuals (mean carapace length (CL) of 65.17 mm ± 7.31 (SD) for females and 66.22 ± 9.04 (SD) for males, respectively) caught by local fishermen in commercial areas in every year of the experiment (except for 2002 and 2003) were tagged with plastic T-bar-type tags.These were inserted dorso-laterally, using a tagging gun, between the first and second abdominal segments (Campillo et al., 1979).The following parameters were recorded for each specimen: carapace length (CL) in mm, measured along the median line from the top of the rostrum to the posterior edge of the cephalothorax, total length (TL) in cm, measured along the median line from the infraorbital spine to the posterior edge of the telson, total weight (TW) in grams, and sex.After about three days in tanks in order to assess tagging stress (Secci et al., 1999), the tagged specimens were released in the centre of the MPA.On recapture, whenever possible, the same biometric parameters as those recorded at release were noted for each specimen.A distinction was made between recaptures caught inside the area (IN) and those in the neighbouring fished zone (OUT) (the zone around the MPA within a radius <5 km from the centre of the area), on the basis of reports from professional fishermen.In this neighbouring area, commercial fishing was carried out mainly by the same boats as those involved in the sampling inside the area.All these boats belong to the fishermen of "Coop.Su Pallosu".
As few data were available, in the following analysis they were processed as a whole, without distinguishing their original site of capture (i.e.IN or OUT).

Modelling procedure
Recapture histories of the tagged spiny lobsters were analysed using Arnason-Schwartz (AS) markrecapture models (Arnason, 1973;Schwartz et al., 1993), a multistate generalisation of Cormack-Jolly-Seber (CJS) models (Cormack, 1964;Jolly, 1965;Seber, 1965).The CJS model focuses on "typical" open-sea population mark-recapture models in which the probability of an individual being seen is defined by two parameters: the probability of the animal surviving and remaining in the sample area (Φ) (this is an "apparent survival" since it is not possible to distinguish between losses due to death and those due to permanent emigration), and the probability of the animal being encountered (p), conditional on it being alive and in the sample area.The multistate model adds a further level of biological realism by incorporating movement data (Ψ) (Brownie et al., 1993).Models were fitted using the MARK program (White and Burnham, 1999).
The fully parametised AS model can be represented by Φ(t,s)p(t,s)Ψ(t,s).Thus, the likelihood of survival, resighting and movement is a function of time(t) and strata (IN = inside MPA; OUT = neighbouring fished zone).With 7 encounter occasions, 1 group, and 2 strata, the unconstrained saturated model had 42 parameters.All models were initially structured using the identity design matrix and sink function.
Model selection was guided primarily by the knowledge of the biology of the marked lobsters' population.The aim of the model selection is to identify the biologically meaningful model that explains the significant variability in the data but excludes unnecessary parameters (the principle of parsimony), i.e. additional parameters that cannot be justified on the basis of actual data (cf.Burnham et al., 1987).Parsimony was assessed using the adjusted form of the Akaike Information Criteria (AICc), incorporating an adjustment for lesser lack of fit of the saturated model (Burnham et al., 1995;Anderson et al., 1998), ΔAICc, i.e. the difference in AICc between two models and the normalised AICc weights which provide a relative weight of evidence for a particular model that best describes the data (Burnham and Anderson, 1998).Finally, the most parsimonious model was tested for the most important biological question by comparing it with neighbouring ones using the likelihood ratio test (LRT) (Lebreton et al., 1992).

RESULTS
During the eight years of experimentation (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004), a total of 4044 specimens were tagged, of which 317 were recaptured, with proportionally more males.Altogether, 153 tagged lobsters were caught inside the MPA and 164 in the neighbouring areas from 1998 to 2004 (Tables 1 and 2).The number of recaptures from IN and OUT zones was substantial during the survey years except in the last two years, when it began to decrease.
The recapture data do not appear to indicate fidelity to the site of original capture (Table 2).Given the limited number of data, the following model was thus applied to all the data, without making any distinction according to the site of original capture.

Model selection
The model which best explains the variation in the data while using the fewest parameters is Model 2 with 18 parameters (Table 3), which is approxi-  The best model shows the apparent survival to be constant over time in each stratum (in the no-take area and its surrounding areas it is 0.94 ± 0.03 (SE) and 0.26 ± 0.04 (SE), respectively) although the next best model could lead one to presume that there are temporal changes of survival in the neighbouring fished zone.The likelihood ratio test (LRT) indeed shows the absence of a particularly significant difference in the fit between the two models (χ 2 = 7.908 df = 5 P = 0.1614).
The most parsimonious model shows that the resighting probability varies during the sampling years in the two strata.The values registered in the OUT zone are greater than those in the IN zone, where the values ranged from 0.06 ± 0.03 SE to 0.308 ± 0.05 SE, and from 0.13 ± 0.06 SE to 0.96 ± 0.15 SE, respectively.In both strata the values decreased in the last two years (Fig. 2).These data confirm the recapture trend of tagged specimens observed during the eight annual samples.
The probability of movement from the MPA to the neighbouring fished zone and vice versa appears constant over the years, showing mean values of 0.28 ± 0.03 and 0.31 ± 0.12, respectively.

DISCUSSION
Tagging and recapture methods, such as the Cornack-Jolly-Seber (CJS), have always been considered basic elements for assessing the probability of a population surviving (Williams et al., 2002).However, recent developments of the CJS have shown that the capture and survival rates of an opensea population can be affected by a certain amount of individual variability (Williams et al., 2002).Thus, statisticians and ecologists have attempted, in recent years, to widen the CJS model by introducing "multistate models", in which, in addition to environmental and temporal factors, capture, survival rates and movements of specimens are considered in relation to the individual variability inherent in a population (Bonner and Schwarz, 2006).
The present study describes the results of applying the Arnason Schwartz multistate mark-recapture model (Arnason, 1973;Schwartz et al., 1993) to a temporal series of data covering eight years of experimentation.This method allowed us to examine the apparent survival rate and movement trends of P. elephas in an MPA of the central-western Mediterranean and in the fished areas surrounding it.
The most parsimonious model which best explains the data variability is that of temporally constant rates of apparent survival and movement in each of the two strata, together with an abruptly decreasing trend in the probability of encountering tagged specimens, which was observed in the last two years of experimentation.
The absence of any temporal influence in the apparent survival rate inside the no-take area suggests that spillover and mortality are constant for each period of study.The high value could be also linked to the fact that the introduced specimens, aged between 3 and 4 years (Marin, 1987;Follesa et SCI. MAR., 71(2) ., 2003), do not represent the main prey of certain species, such as Scorpaena scrofa, Scorpaena porcus, Diplodus sargus and the cephalopod Octopus vulgaris, which are present in the area.These predators prey mainly on recently settled juvenile P. elephas (Diaz et al., 2005).The apparent survival rate also remained practically constant over the years in the fished zone surrounding the MPA, although with a value lower than in the MPA.This datum confirms the continuous emigration of P. elephas across the boundary of the small MPA.The lobsters move from inside the MPA to the outside regions, where they are liable to be caught by a small number of fishermen mainly belonging to the same cooperative.Despite this, from the first year of experimentation their fishing effort probably had an effect on the survival of the animals in the OUT zones.Also, the variations in the resighting rate values were correlated with the various degrees of fishing effort and with the reporting rate in the IN and OUT areas.
For the immediate and substantial spillover of a species with a medium-low mobility, like P. elephas, into the surrounding fished zone to occur, the protected area needs to be on about the same spatial scale as the annual movement of individuals (Childress, 1977;Sánchez Lizaso et al., 2000;Kelly et al., 2002).In a reserve substantially greater than the scale of lobster movement, the majority of lobsters will be retained within the boundaries, thus limiting spillover.
The small size of our study area (4 km 2 ) seems to have ensured, even in the first few years of experimentation, the movement of lobsters from the IN zone to the OUT ones, at a rate that appeared to be constant throughout the eight-year period.Considering, in fact, that all the lobsters recaptured had covered a mean distance of ca.1.8 km/yr and that the distance a specimen would have to cover in order to move outside the area (calculated as a straight line from the centre of the area) would be between 1.4 and 1.6 km, it appears clear that the spillover of adult specimens outside the area proved to be consistent after one year of the area being set up.
The results of the present study are encouraging for the potential enhancement of spiny lobster populations.This is the first time that CSJ modelling has been applied to an adult population of lobsters.The relatively small standard errors associated with apparent survival, resighting and also movement rates show that even with a small sample size tests can be powerful.Robust and quantitative informa-tion on the movements of lobsters across the MPA boundary has been acquired.However, the lower survival rate estimated in the surrounding fished areas than in the MPA ones could indicate that the imposition of MPAs should be associated with a controlled fishery pressure in the adjacent fished areas, depending on the state of depletion in the fishery before their introduction.
Hence, even if our results cannot be extended as a generalisation to other protected areas, we hope that this information will make a useful and important contribution to the wider debate on the effect of introducing MPAs into a managed commercial fishery of rock lobsters.

Fig. 1 .
Fig. 1. -Map of the Sardinian protected area and surrounding coastline.
FIG. 2. -Probability of resighting (p) of spiny lobsters in the restocking area (a) and the neighboring fishing zone (b). al

TABLE 1 .
-Number of specimens tagged each year (total, males and females).

TABLE 2
. -Breakdown of the recaptures recorded inside (IN) and outside (OUT) the MPA in the years of experimentation (no data available for 1997).