Standardised Catch Rates for Swordfish (Xiphias Gladius) from the Italian and Greek Fisheries

SCRS/2007/107 COLLECT. VOL. SCI. PAP. ICCAT, 62(4): 1074-1080 (2008)

STANDARDIZATION OF SWORDFISH (XIPHIAS GLADIUS) CATCH RATES FROM THE GREEK AND ITALIAN MEDITERRANEAN LONGLINE FISHERIES

George Tserpes1, Panagiota Peristeraki1, Antonio Di Natale2, Antonia Mangano2

SUMMARY

Indices of abundance of swordfish (Xiphias gladius) from the Greek long-line fisheries operating in the eastern Mediterranean and the Sicilian ones exploiting the Tyrrhenian Sea and the Straits of Sicily are presented for the period 1987-2005. Annual standardized indices were estimated by means of Generalized Linear Modeling techniques and the predictor variables included the Year, Month and Area of fishing. Results did not demonstrate the presence of any particular trend over time.

RÉSUMÉ

Ce document présente les indices d’abondance de l’espadon (Xiphias gladius) des pêcheries palangrières grecques opérant dans la Méditerranée orientale et des pêcheries siciliennes exploitant l’espadon dans la mer Tyrrhénienne et le Détroit de Sicile au titre de la période 1987-2005. Les indices annuels standardisés ont été estimés au moyen de techniques de modèle linéaire généralisé et les variables de prédiction incluaient l’année, le mois et la zone de pêche. Les résultats n’ont signalé la présence d’aucune tendance particulière dans le temps.

RESUMEN

Se presentan, para 1987-2005, índices de abundancia de pez espada (Xiphias gladius) de las pesquerías griegas de palangre que operan en el Mediterráneo oriental y las pesquerías sicilianas que explotan el pez espada en el Mar Tirreno y el Estrecho de Sicilia. Se estimaron los índices anuales estandarizados por medio de técnicas de modelación lineal generalizada y las variables independientes incluían Año, Mes y Área de pesca. Los resultados no demostraron la presencia de ninguna tendencia particular en el tiempo.

KEYWORDS

Swordfish, Mediterranean, catch/effort

1. Introduction

Swordfish (Xiphias gladius) is a commercially important migratory fish heavily fished in the Atlantic and Mediterranean. Greece and Italy are among the most important swordfish producers in the Mediterranean and, in the latest years, account for about 50-60% of the total Mediterranean production. (Anon. 2006)

Greek swordfish fisheries exploit the eastern part of the Mediterranean basin covering a large area, extending from the east Ionian to the Levantine seas. The gear used is drifting surface long-lines. Italian fisheries mainly exploit the Adriatic, Ionian and Tyrrhenian seas using surface long-lines and gillnets.

The aim of the present work is to estimate annual standardised abundance indices based on commercial catch per unit effort (CPUE) data series obtained from the main Greek and Italian long-line fleets. In principle, it is attempted to update a previously estimated series (Tserpes et al., 2004) through the use of additional data. The work is focusing in the long-line gear, as gillnets have suffer several modifications after the driftnet ban resulting in large inconsistencies in the gear used. Data have been analysed by means of commonly used Generalised Linear Modelling (GLM) techniques.

1 Hellenic Centre for Marine Research, P.O. Box 2214, 71003 Iraklion, Greece. E-mail:[email protected] 2 Aquastudio, Messina, Italy. 50 2. MATERIALS AND MMTHODS

CPUE data have been collected in the frames of past European and national projects and included spatial and temporal information on catch and effort data, in as much as possible detail, i.e. on an individual boat trip basis. CPUE observations were expressed in terms of kg/1000 hooks. Sampling, which was based on information from landings on pilot ports, covered the period 1987-2005 and included the main Greek and Italian fleets exploiting different areas of the central and eastern Mediterranean (Figure 1).

In the case of Greece, a total of 3804 observations were analyzed. These covered the activities of the two main swordfish fleets operating in the country, the fleets of Kalymnos and Hania. Generally, catches of both fleets represent 50-70% of the total Greek production (Tserpes et al., 2002). These fleets mainly exploit the central, southeastern Aegean Sea but occasionally extend their activities to the northern Aegean and Levantine basin. Fishing is carried out using drifting surface long-lines through February to September while is prohibited by law from October to January. In the last five years, the traditional long lines have been modified, resembling the ones used for the tuna fishery in the Atlantic. The modified gear, which is known as American-type long-line, is set deeper than the traditional one and uses fluorescent material to attract the fish. Catchability differences among the different long-line types were taken into account when creating the global data set. Specifically, based on the results of a previous study (Tserpes & Peristeraki, 2004), observations from the American-type long-line were multiplied by 0.64 to reflect the estimated 36% catchability differences among gears.

In the case of Italy, a total of 2276 observations were analyzed from the Sicilian long-line fleet. The Sicilian fleet, which is among the larger swordfish fleets in the Mediterranean, mainly exploits the Tyrrhenian Sea but occasionally expands its activities into a much wider area. The fishery mainly operates from August to December while the driftnet fishing season usually lasts from April to August.

CPUE data were analysed, separately by country, by means of Generalised linear modelling (GLM) techniques (McCullagh and Nelder 1983). Based on the deviance residuals plot, models assuming a Gamma error structure with a log link function were found to be the most appropriate for all examined data sets. The models included year, month and area as main effects and the month-area interaction. Interaction terms including the “Year” effect were not examined as it was considered that they could bias annual standardised estimates. Model fitting was accomplished using the statistical package S-plus, following the approaches described by Venables and Ripley (1997).

3. Results and discussion 0.1 Eastern Mediterranean (Greek longliners)

A total of 3804 data records were analysed that were collected in the period 1987-2006 with the exception of 1989, 1996 and 1997. Five fishing areas were considered: A = Cretan sea, B = Central Aegean, C = South- eastern Aegean, D = Levantine and E = North Aegean (see also Figure 1). There is no any outstanding feature in the deviance residual plot suggesting that the model is inappropriate for the observations (Figure 2). The analysis of deviance table indicated that the model explained about 15% of the total variation. All effects were significant on the 95% statistical level (Table 1).

The effect of the significant predictors on CPUE is shown on the y-axis for different values of the predictor (x- axis) (Figure 3). CPUE levels do not show any particular trend among years while the beginning and end of the fishing season seem to be more productive. The standardised CPUE indices were calculated by rescaling the “year” effect estimates to the average estimated CPUE rate; estimated indices are given in Table 3.

3.2 Central Mediterranean (Sicilian longliners)

A total of 2276 data records were analysed that were collected throughout the year from 1991 to 2005, with the exception of 1993 and 1996. Three fishing areas were considered: I = Straits of Sicily, J = South Tyrrhenian and K = Central Tyrrhenian seas (see also Figure 1). The model provided a good fit to the data as it was demonstrated by the deviance residual plot (Figure 4). The analysis of deviance table indicated that it explained about 23% of the total variation. All factors were significant on the 95% statistical level (Table 2).

The effect of the significant predictors on CPUE is shown on the y-axis for different values of the predictor (x- axis) (Figure 5). Similarly to the Greek longliners, the CPUE levels do not show any particular trend among 51 years. CPUE indices were calculated by rescaling the “year” effect estimates to the average estimated CPUE rate; estimated indices are given in Table 3.

References

ANON. 2006. Report for Biennial Period 2004-05, Part II (2005), Vol. 3: pp 33-45.

MCCULLAGH, P. and J.A. Nelder. 1983. Generalized Linear Models. Chapman and Hall, London.

VENABLES, W.N. and B.D. Ripley. 1997. Modern Applied Statistics with S-PLUS, Second Edition. Springer.

TSERPES, G., P. Peristeraki and A. Di Natale. 2001. Size distribution of swordfish landings in the central and eastern Mediterranean. Collect. Vol. Sci. Pap. ICCAT, 52(2): 733-739.

TSERPES, G., P. Peristeraki, C. Koutsikopoulos, G. De Metrio, A. Di Natale, J.M. De La Serna, D. Macias, and J.M. Ortiz de Urbina. 2002. The swordfish fishery in the Mediterranean. Final report of the EU Project 99/032. 75p.

TSERPES, G., P. Peristeraki and A. Di Natale. 2004. Standardised catch rates for swordfish (Xiphias gladius) from the Italian and Greek fisheries operating in the central-eastern Mediterranean. Collect. Vol. Sci. Pap. ICCAT, 56(3): 850-859.

TSERPES, G., P. Peristeraki. 2004. Catchability differences among the longlines used in the Greek swordfish fishery. Collect. Vol. Sci. Pap. ICCAT, 56(3): 860-863.

52 Table 1. Analysis of deviance table for the Gamma-based GLM model fitted to long-line CPUE data from the Greek fleets.

Source of df Deviance Res.df Res.Dev F Value Pr(F) Variation NULL 3803 2229.131 Year 16 220.1936 3787 2008.938 25.3197 0.00 Month 7 18.5730 3780 1990.365 4.8815 0.00 Area 4 54.8425 3776 1935.522 25.2251 0.00 Month:Area 28 35.7977 3748 1899.724 2.3521 0.00

Table 2. Analysis of deviance table for the Gamma-based GLM model fitted to long-line CPUE data from the Sicilian fleet.

Source of df Deviance Res.df Res.Dev F Value Pr(F) Variation NULL 2275 1349.936 Year 12 152.9211 2263 1197.015 33.4993 0.00 Month 11 52.6684 2252 1144.347 12.5866 0.00 Area 2 86.6775 2250 1057.670 113.9269 0.00 Month:Area 14 16.8570 2236 1040.813 3.1652 0.00

Table 3. Standardized abundance indices by year and fleet. Indices are expressed in terms of kg/1000hooks.

Year Greek LL Sicilian LL 1987 129.34 1988 157.95 1989 1990 136.65 1991 175.33 112.96 1992 48.17 108.14 1993 130.10 1994 186.91 101.64 1995 104.72 127.76 1996 1997 80.42 1998 197.87 134.69 1999 162.36 158.74 2000 131.48 64.11 2001 137.39 123.00 2002 104.87 195.00 2003 137.64 87.11 2004 132.05 120.05 2005 135.90 132.31 2006 129.06

53 Figure 1. Map of the central-eastern Mediterranean indicating the main areas exploited by the studied fleets. A = Cretan sea, B = Central Aegean, C = Southeastern Aegean, D = Levantine, E = North Aegean, I = Straits of Sicily, J = Southern Tyrrhenian and K = Central Tyrrhenian. 3 2 s l a u 1 d i s e R

e c 0 n a i v e D 1 - 2 - 3 -

50 100 150 200 250 Fitted : Year + Month + Area + Month * Area

Figure 2. Residual deviance of the generalized linear model fitted to Greek long-line data, plotted against predicted CPUE. The fitted line represents a locally weighted smoother.

54 plot represents the contribution of the corresponding variable to the fitted predictor. The fitted values are values fitted The predictor. fitted the to variable corresponding indicate errors. lines twostandard the broken to zeroand adjusted average the of contribution the represents plot 3. Figure partial for A rea partial for Y ear -1.0 -0.5 0.0 0.5 -1.0 -0.5 0.0 0.5 1.0 Generalized linear model derived significant effects on CPUE index for the Greek longline data. Each data. longline Greek the for index CPUE on effects significant derived model linear Generalized Area Year 55

partial for M onth -1.0 -0.5 0.0 0.5 1.0 Month 2 1 s l a u d i s e R 0

e c n a i v e D 1 - 2 -

0 100 200 300 Fitted : Year + Month + Area + Month * Area

Figure 4. Residual deviance of the generalized linear model fitted to Sicilian longline data, plotted against predicted CPUE. The fitted line represents a locally weighted smoother.

Year Month 0 . 1 2 5 . h r t 0 a n 0 e o Y

r o 0 f . o

l a i 2 a t i - r t r a a p 5 . p 0 - 4 - 0 . 1 -

Area 1 a 0 e r A

1 l - a i t r a p 2 -

Figure 5. Generalized linear model derived significant effects on CPUE index for the Sicilian longline data. Each plot represents the contribution of the corresponding variable to the fitted predictor. The fitted values are adjusted to average zero and the broken lines indicate two standard errors.