Trends in various biological parameters of the European sardine , Sardina pilchardus ( Walbaum , 1792 ) , in the Eastern Mediterranean Sea *

The European sardine (Sardina pilchardus) and anchovy (Engraulis encrasicolus) are the two most important small-sized pelagic species in Greek waters. A study on genetic distances, based on electrophoresis, and of morphometric and meristic characters using multivariate analysis does not indicate separate stocks of sardine in the Aegean Sea (Spanakis et al., 1989). Sardine comprises about 12% of the mean total reported landings (Stergiou et al., 1997a) and 23% of the North Aegean ones (Stergiou and Pollard, 1995). It is mainly caught using purse seiners at depths less than 60 m, making up about 20% of the total purse seine catch, as well as using beach seiners, small netters and, to a lesser extent, trawlers (Stergiou et al., 1997a). Sardine landings exhibit a clear seasonal cycle, with catches being much higher during the October-May period, when sardine schools migrate to the coasts for spawning (Stergiou et al., 1997a; Stergiou and Lascaratos, 1997; Giannoulaki et al., 2001). During 1964-1990, the total sardine landings fluctuated around 10,000-12,000 t and sharply increased to 20,000 t in 1992, following the decline of the anchovy landings from 25,000 t in SCI. MAR., 67 (Suppl. 1): 269-280 SCIENTIA MARINA 2003


INTRODUCTION
The European sardine (Sardina pilchardus) and anchovy (Engraulis encrasicolus) are the two most important small-sized pelagic species in Greek waters.A study on genetic distances, based on electrophoresis, and of morphometric and meristic characters using multivariate analysis does not indicate separate stocks of sardine in the Aegean Sea (Spanakis et al., 1989).Sardine comprises about 12% of the mean total reported landings (Stergiou et al., 1997a) and 23% of the North Aegean ones (Stergiou and Pollard, 1995).It is mainly caught using purse seiners at depths less than 60 m, making up about 20% of the total purse seine catch, as well as using beach seiners, small netters and, to a lesser extent, trawlers (Stergiou et al., 1997a).Sardine landings exhibit a clear seasonal cycle, with catches being much higher during the October-May period, when sardine schools migrate to the coasts for spawning (Stergiou et al., 1997a;Stergiou and Lascaratos, 1997;Giannoulaki et al., 2001).During 1964During -1990, the total sardine landings fluctuated around 10,000-12,000 t and sharply increased to 20,000 t in 1992, following the decline of the anchovy landings from 25,000 t in 1987 to about 10,000 t in 1992 (Fig. 1).The analysis of sardine landings using a variety of statistical tools (Stergiou, 1989(Stergiou, , 1990(Stergiou, , 1991(Stergiou, , 1992;;Stergiou and Christou, 1996;Stergiou et al., 1997b;Stergiou and Lascaratos, 1997) indicate: (a) persistence of landings; (b) landing periodicity of 2-3 and 4-5 years; (c) that climate might affect long-term trends and shortterm variation in the landings; and (d) variability and replacement of anchovy by sardine landings are not due to chance, and that wind activity over the North Aegean Sea might act as a forcing function.
There are many published studies concerning various aspects of sardine biology, ecology and fisheries in Greek Seas (physiology: Machias andTsimenides, 1995, 1996;age, growth and reproduction: Laskaridis, 1948;Tserpes and Tsimenides, 1991;Machias et al., 2001;Koutrakis et al., 2001;Ganias et al., 2000Ganias et al., , 2001b;;migration: Giannoulaki et al., 1999;eggs and larvae: Yannopoulos et al., 1973;Daoulas and Economou, 1986;Machias et al., 2001;Ganias et al., 2001a;genetics: Spanakis et al., 1989;Machias et al., 2001;daily egg production: Somarakis et al., 2001;echo-abundance anddistribution: Machias et al., 1996, 2001;Maravelias et al., 1997;Giannoulaki et al., 2001; fisheries: for a review see Stergiou et al., 1997a).Nevertheless, there is an absence of regular, long time series of data on growth, maturity, biomass and recruitment of sardine, and other pelagic species, throughout the Greek Seas.This is because experimental sampling and surveys are generally highly discontinuous at both temporal and spatial scales, a fact representing an impediment to the study of the effects of environmental variability on small pelagic species.Within this context, an effort was made in 1996 for the establishment of a long time series, the longest one for the Eastern Mediterranean Sea, of various biological parameters of sardine based on biweekly samples collected from the commercial landings of the purse-seine fleet in the Northwest Aegean Sea.
Fishing has important effects on the demographic structure of the exploited stocks (e.g.Jennings et al., 2001;Stergiou, 2002).The increased fishing effort expended on sardine stocks since the early 1990s (Fig. 1) occurred with a 34% decrease in the North Aegean sardine biomass in 1996 when compared to that of 1995 (Machias, unpubl. data).In this study we tested whether such an increase in the effort expended on sardine populations is reflected in the trends in the biweekly length-frequencies, the mean total length, the exponent b of the lengthweight relationship and the gonadosomatic index of sardine during July 1996-June 2000.In addition, growth and mortality rates were also estimated.

MATERIAL AND METHODS
Biweekly random samples were collected from the commercial landings of the purse-seine fleet at the fishing port of Nea Michaniona (northern rim of the Thermaikos Gulf) from July 1996 to June 2000.Samples were taken within a single day in two or three clusters from random collections of the total landed catch derived from different vessels operating in the Northwest Aegean Sea.No data were collected for the period between 15 th December and 1 rst March of each year, because purse-seine fishing in the Greek Seas is banned during that period.The sampling frequency ranged from 7 to 17 days, with a mean sampling frequency of 15.14 days (± SE=0.44).
Total length (TL) was measured to the nearest mm for 51.246 individuals.In addition, for a subsample of 50 individuals per sampling period, gross body weight (W) and gonad weight (W g ) were also measured to the nearest 0.1 and 0.01 g, respectively.
Sex was macroscopically identified.
The above mentioned data were consequently used for the estimation of the following biological parameters: (a) biweekly length-frequencies; (b) arithmetic mean TL per sampling period; (c) TL-W relationship, W=a(TL) b ; (d) gonadosomatic index (GSI; Nikolsky, 1963): GSI=100(W g /W); and (e) growth parameters, K, L ∞ , C and WP, of the seasonalized von Bertalanffy (1938) growth equation (Pauly, 1998).The latter were estimated from the monthly length-frequencies using various routines of FiSAT (ELEFAN I: "Scan K-values", "response surface" and "automatic search" routines: Gayanilo et al., 1994;Gayanilo and Pauly, 1997).The seasonalized growth equation is (Pauly, 1998): where K is the curvature parameter, L ∞ is the asymptotic length, C is the amplitude of growth oscillation and WP is the winter point related with the summer point t s (WP = t s +0.5) indicating the time of the year when growth is slowest.For the computation of the von Bertalanffy parameters, we used as an initial range of L ∞ the following values: (a) the 95% confidence interval, CI, of the maximum length L max ' estimates; and (b) the estimate of L ∞ derived from the Powell-Wetherall method (Powell, 1979;Wetherall, 1996;Pauly, 1986), applied to all samples pooled together.Both methods are implemented in FiSAT (Gayanilo et al., 1994;Gayanilo and Pauly, 1997) and briefly presented below.L max ' is estimated from the set of n extreme values of the largest specimen in each sample (L*) using the regression: where P is the probability associated with the occurrence of an extreme value, 1/a is a measure of dispersion, and L max ' is the intercept of the regression line with the probability associated with the n th observation.P is computed for any extreme value from P = m/(n+1) where m is the position of the value, ranked in ascending order and n is the number of L* values (Gayanilo et al., 1994;Gayanilo and Pauly, 1997).According to the Powell-Wetherall method, the smallest length fully recruited to the gear (L') and the mean length of a fish with length larger than L' (L mean ) are related as follows: (L mean -L') = a+b (L'), where: Finally, total mortality (Z) estimates were derived from the length-converted catch curve (Pauly, 1984): where N i is the number of fish in length class i, ∆t i is the time needed for the fish to grow through length class i, t i is the age (or the relative age, computed with t 0 = 0) corresponding to the mid-length of class i, and the arithmetic value of the slope b is an estimate of Z. Consequently, natural mortality (M) was estimated from the growth parameters and the mean annual water temperature in the study area (i.e.15°C) using Pauly's (1980) empirical equation, fishing mortality (F) from the relationship F = Z-M and the exploitation ratio (E) from the relationship E = F/Z (Pauly, 1983;Pauly and Morgan, 1987).

RESULTS
In total, the TL of sardine ranged between 8 and 21 cm (mean length = 14.48 cm ± SE = 0.64).The annual length-frequency distribution exhibited a mode at 15 cm in 1996-1997 and at 14 cm in 1997-2000 (Fig. 2).During July 1999-June 2000, the percentage of the smallest individuals (TL<13 cm) was higher whereas that of the larger ones lower than those in the previous years (Fig. 2).The biweekly length-frequency distributions (Fig. 3) displayed modes at 15-16 cm for the first year, at 14-15 cm for the second and third year and at 13-14 cm for the last study year during May to September.The smallest individuals (12-14 TL cm for 1996-1999 and 11-13 cm TL for 1999-2000) were recorded during October-April whereas during the last study year very small individuals (TL<11 cm) were also recorded during July-October.
The exponent b of the length-weight relationship was 3.041 for females, 3.116 for males and 3.153 for sexes combined (Fig. 5).The exponent b for sexes combined exhibited a seasonal cycle, with the highest values generally occurring, with few exceptions, from March to September-October, and also increased significantly (r=0.358,n=77, P<0.002) with time (Fig. 6).It was also negatively correlated with the mean TL over the study period (r=-0.29,n=77, P<0.05).
The GSI (Fig. 7) also exhibited a clear bimodal seasonal cycle, with peaks during November-April and lowest values during May-September, with the importance of the secondary peak increasing with time.No relationship was found between GSI and the exponent b of the length-weight relationship (r = -0.13,n=77, P>0.05).

DISCUSSION
The annual length-frequencies of sardine were unimodal (Fig. 2), a fact clearly indicating that landings are based exclusively on one year-class, as is also the case in the NE Aegean (i.e., Thracian Sea: Koutrakis et al., 2001;Kallianiotis et al., 2001), the Central Aegean Sea (Ganias et al., 2001a) and the northwestern Mediterranean (Pertierra and Perrotta, 1993).The maximum recorded TL for the last study year was 18.9 cm (Fig. 2), which was larger than that recorded in the Thracian Sea during October 2000-July 2001 (TL=17.3;Koutrakis et al., 2001), a fact indicating that a further decline in TL may have taken place in 2001.The maximum recorded TL in the study area was also larger than that reported from the Ionian and Central Aegean Sea during 1999-2001(18 and 18.4 cm, respectively: Machias et al., 2001)).
The exponent b of the length-weight relationship for sexes combined was similar to that estimated for sardine in the Ionian and Central Aegean Seas during 2000-2001 (b=3.15;Machias et al., 2001) and slightly smaller than that estimated for October 2000-July 2001 in the Thracian Sea (b=3.29;Koutrakis et al., 2001).
In the study area, sardine was characterized by high K and L ∞ values, when compared to estimates from other areas, being comparable to those estimated from otoliths' readings in the Western Mediterranean (Table 1).The North Aegean Sea, including the Thermaikos Gulf, is generally considered as a relatively eutrophic area, being characterized by higher plant pigment concentrations, influenced by nutrient input from the Black Sea, freshwater runoff and anthropogenic eutrophication, than those prevailing in the remaining parts of the Greek Seas (Stergiou and Georgopoulos, 1993;Stergiou et al., 1997a).Accordingly, the relative abundance and density of small pelagic fishes, as indicated by echo surveys, are higher in the North Aegean Sea than in the remaining parts of the Greek Seas (Stergiou et al., 1993;Machias et al., 1996;Machias et al., 2001).
Sardine, like other temperate clupeids, is a multiple spawner producing several discrete batches of oocytes within the reproductive period (Ganias et al., 2001a;Somarakis et al., 2001), relying on energy stored as fat rather than liver glycogen for reproduction, with individuals not spawning at periods of high food abundance, but storing energy as fat for later reproductive activity (Somarakis et al., 2000).The analysis of the sardine GSI in the study area (Fig. 7) indicated that spawning takes place during the colder part of the year, in November-April.This is also true of the Central Aegean and Ionian Seas (Machias et al., 2001), the Thracian Sea (Koutrakis et al., 2001) as well as in other areas of the Mediterranean Sea and off the Moroccan and Spanish coasts of the Atlantic Ocean (e.g.Delgado and Fernandez, 1985;Alvarez and Alemany, 1996;Quintanilla and Perez, 2000;Froese and Pauly, 2000; FishBase online: www.fishbase.org).In contrast, spawning in the North Sea and Black Sea takes place in the summer (June to August: FishBase online).It must be noted that based on the biweekly GSI, the reproductive period of sardine in the North Aegean extends over 4-6 months (Fig. 7) as is also the case for the Thracian Sea (Koutrakis et al., 2001) and the Central Aegean Sea (Machias et al., 2001).In contrast, in the Ionian Sea the reproductive activity of sardine extends over about 2-4 months which are shifted towards spring (Machias et al., 2001;Somarakis et al., 2001).Finally, it is noteworthy that the increase in the importance of the secondary GSI peak with time (Fig. 7) must be related to the decrease in the mean TL with time (Fig. 4), given that smaller sardines generally spawn later than larger ones (Machias et al., 2001).It must be pointed out that although GSI describes the reproductive condition of sardine before and after the peak of reproduction it does not adequate-ly describe the reproductive condition within the reproductive period (Ganias et al., 2001a).
In general, growth is affected by a variety of factors, such as food quantity and quality, temperature and environmental heterogeneity, and growth itself affects maturation (Wootton, 1990).The marked decline in the mean sardine TL during 1996-2000 (Fig. 3) followed: (a) the sharp increase in the sardine landings during 1992-1997 with a concurrent decline in anchovy landings (Fig. 1) (see also Koutrakis et al., 2001;Kallianiotis et al., 2001); and (b) the 34% decrease in the North Aegean sardine biomass in 1996 when compared to 1995 (A. Machias, unpubl. data).In addition, sardine discarding has been greatly reduced in recent years, with the length structure of discarded specimens being similar to that of the retained ones (Kallianiotis et al., 2001).The above-mentioned facts together with the high exploitation rate estimated for the study period (E=0.76)indicate that intense fishing, by removing the largest individuals, may be implicated in such a decline in the maximum and mean TL of sardine during 1996-2000.Fishing generally affects the age and length structure of fisheries stocks (e.g.Jennings et al., 2001;Stergiou, 2002) and fishinginduced declines in the length structure have also been observed for other sardine species of the world (e.g.Japanese sardine: Kawasaki, 1992;Hiyama et al., 1995).Yet, the effect of other factors, such as for instance adverse environmental conditions and low food density (e.g. for North Aegean Sea: Somarakis et al., 1997Somarakis et al., , 2000;;Stergiou and Lascaratos, 1997;Japanese sardine: Wada and Kashiwai, 1991;Hiyama et al., 1995), acting in a synergetic or not fashion with fishing, cannot be ruled out.In any case, given that the reproductive traits (i.e., fecundity, spawning frequency, egg size) of sardine are lengthdepended (Ganias et al., 2001b;Somarakis et al  2001), a decrease in the mean TL of sardine may lead to a decline in the total population reproductive output and offspring survival (Trippel et al., 1997).This, coupled with climate-induced changes in the biological production of the North Aegean Sea (e.g.Stergiou et al., 1997a) may lead to long term fluctuations in sardine abundance.These issues require further studies.
The decline in the mean TL (Fig. 3) went along with an increase in the exponent b of the lengthweight relationship (Fig. 5), which may indicate a density-dependent effect.In general, densitydependent effects on growth are commonly reported for fish populations (e.g.sardine: Wada and Kashiwai, 1991;Hiyama et al., 1995;herring: Heath et al., 1997;silver hake: Ross and Almeida, 1986;eel: De Leo and Gatto, 1996).The underlying mechanism involves relaxation of competition for available food resources at low population densities (e.g.Jennings et al., 2001).Thus, a decrease in the abundance of large sardines in the North Aegean could result in more quantity of food being available for small sardines and/or small individuals gain faster weight in the absence of intra-specific competition.Yet, density-dependence is not always easy to prove (Jennings et al., 2001) and in fact other mechanisms for density-dependent growth may also be implicated (e.g. the relationship between school size and population density and its effect on growth: Bakun, 1989;Bakun and Cury, 1999; amplified size-dependent predation during population declines: Bakun, 1989).This clearly indicates the need for an "integrated ecosystem view of marine resource population dynamics" (Bakun, 1989), such as the paradigmatic approach undertaken for the Peruvian anchovy (Pauly and Tsukayama, 1987;Pauly et al., 1989), which will account for spatial and temporal variability in a variety of both biological and environmental parameters.
The trends in the biological characteristics of sardine reported here must be taken into account when designing management plans, and their incorporation into various assessment models may considerably reduce uncertainty in estimations and increase predictability.The fact that there were only four years of data does not allow us to statistically address many of the issues discussed above.However, monitoring of the biological parameters of sardine in the study area is ongoing, and an effort will be made to relate such changes with environmental ones.
FIG. 6. -Temporal variation in the exponent b of the length-weight relationship of sardine, Sardina pilchardus, in the Northwest Aegean Sea, 7/1996-6/2000. ., TRENDS IN VARIOUS BIOLOGICAL PARAMETERS OF SARDINE 277