Dietary changes of Mediterranean Shags Phalacrocorax aristotelis desmarestii between the breeding and post-breeding seasons in the upper Adriatic Sea

Abstract

Capsule Shags move between breeding and non-breeding areas and this is associated with a significant change in diet.

Aims To determine whether the diet of Shags nesting on islets off the Croatian coast is the same as their diet after the post-breeding move to the Gulf of Trieste.

Methods Diet was determined by the analysis of 611 regurgitated food pellets.

Results A total of 23 988 prey items were identified in the sample of pellets. Post-breeding Shags in the Gulf of Trieste focused on demersal and relatively immobile Gobiidae (81.5% by number, 87.1% by biomass). The most frequent prey species was Gobius niger (70.8% by number). In the breeding season at Oruda island, Croatia, the diet was more varied. Breeding Shags fed on bentho-pelagic, mobile prey such as Atherina boyeri (28.4% in frequency), Serranus hepatus (16.1%) and Crenilabrus tinca (12.0%), while Gobiidae had a dietary frequency of only 18.1%. With respect to biomass the most important prey were Crenilabrus tinca (19.0%) and Serranus hepatus (18.4%).

Conclusion We suggest that the movement of Shags within the Adriatic Sea is driven by dietary requirements. Most previous studies of Shag diet have shown that Shags tend to have a more specialized diet during the breeding season, concentrating upon demersal prey species. However, we have found that birds breeding at the Croatian study colony show dietary diversity. We suggest that lack of dietary specialization is a facultative response to local prey abundance, and is probably the result of over-fishing of demersal species in the areas around the breeding locations in which the birds find suitable sites and are little disturbed by human activity. Shags may move immediately after breeding to the Gulf of Trieste because demersal species are likely to be more abundant there. As a consequence, the diet becomes more specialized and is then more similar to the diet of other populations of Shags.

Seabirds are key players at the top of marine food chains and respond predictably to alterations in prey abundance. They have therefore been used as reliable indicators of changes in fish populations (Hatch & Sanger 1992, Montevecchi 1993, Furness & Camphuysen 1997, Barrett 2002). Changes in a variety of seabird foraging parameters have been successfully used to detect alterations in prey age–class structure (Davoren & Montevecchi 2003), responses of prey populations to climate change (Miller & Sydeman 2004), and variations in the energetic value of prey (Wanless et al. 2005). Furthermore, seabirds' dietary adjustments may indicate shifts in pelagic food webs (Montevecchi & Myers 1996).

European Shags Phalacrocorax aristotelis are foot-propelled diving seabirds (Ashmole 1971), whose diet has been used to assess the recruitment and abundance of fish populations (Barrett et al. 1990, Barrett 1991). In Iceland the diet of this species was used as a measurement of the recruitment in commercially important fish species such as Saithe Pollachius virens and Plaice Pleuronectes platessa (Lilliendahl & Solmundsson 2006). Shag diet varies depending on annual changes in prey availability (Carss 1993) and, to a lesser extent, on location. The species feeds on a wide range of benthic, demersal and schooling pelagic fish, and for this reason it is classified as opportunistic in its feeding habits (Barrett 1991, Grémillet et al. 1998, Velando & Freire 1999). Foraging areas have depths ranging between 7 and 80 m (Guyot 1988, Barrett & Furness 1990, Wanless et al. 1991a, 1993a), with a mean of 30 m (Wanless et al. 1997a), and are usually between 7 and 17 km around the breeding colonies (Wanless et al. 1991b). The average daily food consumption of a Shag is between 16 and 24% of body mass (Barrett et al. 1990). During incubation, Shags are estimated to have an average daily requirement of 389 g of fish, while a bird with three chicks requires about 920 g (Wanless et al. 1993b). The diet of the species has been studied as early as the first half of the last century (Steven 1933), mainly in north-western Europe. In this area the Atlantic subspecies P. aristotelis aristotelis is distributed from Iceland and Northern Scandinavia to the Iberian Peninsula (Nelson 2005). The analysis of dispersal patterns suggests that Iberian Shags are isolated from northern populations (Velando 1997, Velando & Freire 1999). Sandeels Ammodytes spp. dominate the diet of P. aristotelis aristotelis, both in summer (Barrett et al. 1986, Grémillet et al. 1998, Furness & Tasker 2000) and in other seasons (Harris & Wanless 1991). Exceptions to Sandeel dominance were found in Norway, where Gadoids (Gadidae) were the most important prey (Barrett et al. 1990). In Iceland Shags rely heavily on Lesser Sandeels Ammodytes marinus during the breeding season, whereas Bull-rout Myoxocephalus scorpius and Gadoids become increasingly important in autumn and winter (Lilliendahl & Solmundsson 2006). Sandeels are also the dominant prey on the Galician coast of north-west Spain (Velando & Freire 1999). In this area some seasonal differences in diet have been found; in February and March Shags foraged on Gobiidae and Sand Smelts Atherina presbyter and this dietary change may be due to the seasonal changes in the abundance of Sandeel schools. In northern Spain breeding Shags fed on Labridae and Atherinidae (Alvarez 1998).

Even though the Atlantic subspecies is well studied, very little is known of the diet, general biology and population of the Mediterranean subspecies P. aristotelis desmarestii, which is endemic to the Mediterranean and Black Seas. The breeding range includes most States along the Mediterranean coast (Croatia, Albania, Bulgaria, Ukraine, Turkey, Cyprus, Egypt, Libya, Tunisia and Algeria). The total population has been estimated at less than 10 000 pairs, half of them breeding within the eastern coast of Spain, Baleares, Corsica, Sardinia, the Tuscany archipelago, Lampedusa, Crete and the islets of the Ionian Sea. Substantial year-to-year fluctuations in breeding numbers have been reported in several colonies, and there appears to be an overall decrease (Aguilar & Fernandez 2002). This occurs mainly because Mediterranean Shags are severely affected by a number of limiting factors, including human disturbance (Guyot 1993), chemical pollution (Lambertini & Leonzio 1986), oil spills (Velando et al. 2005) and accidental catch (Aguilar 1991, Velando & Freire 2002). The subspecies is consequently listed in Annex I of the Birds Directive 147/2009 and is the focus of an Action Plan (Aguilar & Fernandez 2002). To our knowledge, the available data for the diet of the species in the whole Mediterranean rely on the study of Guyot (1988) on the Corsican population, which suggested that Labridae and Ammodytidae were the most important prey. However, a small sample of chick regurgitations from some Sardinian breeding colonies showed the presence of a wider range of species in the diet, including Crenilabrus sp., Mullus surmuletus, Coris julis, Mugil sp., Serranus sp., Sardina pilchardus and Boops salpa (Brichetti et al. 1992).

In the Adriatic Sea, Mediterranean Shags breed in Croatia (2000 pairs) and Albania (5–10 pairs) (Aguilar & Fernandez 2002). From the 1980s onward, the species has become a regular summer visitor in the Gulf of Trieste, north east Italy, with increasing numbers (Utmar 1999). The maximum size of the population approached 2500 individuals in July 2007. This population comes mainly from the Croatian breeding colonies (Sponza et al. 2010). Because of the high sensitivity of Shags to human disturbance during breeding (Guyot 1993), the Gulf of Trieste does not offer suitable and low disturbed places for nesting, but these characteristics are to be found within the Croatian islets. Given Mediterranean Shags' conservation status and the very few studies on this subspecies, we have determined the diet during the breeding season in Croatia and the post-breeding period in the Gulf of Trieste. We use this dietary information to suggest reasons why Shags breed in Croatia but spend the post-breeding period in another area of the Adriatic Sea.

METHODS

Study area

We studied the diet of Mediterranean Shags during the breeding season (January–April) at Oruda island (Lošinj archipelago) (44°33′N, 14°30′E), one of the most important Croatian colonies in the upper Adriatic Sea. In this rocky, low-vegetated island, covering an area of 36 ha, about 100 breeding pairs nest inside bushes of Paliurus spina-Christi and Crataegus monogyna near the sea. The diet of the species in the post-breeding period (May–October) was studied in the Gulf of Trieste (45°39′N, 13°46′E). The two study areas have very different sea bottom depth profiles. The Gulf of Trieste is 20 m deep within 1 km of the coastline and deepest at 25 m. The Lošinj archipelago is characterized by a steep, sloping sea bed. The eastern part of the archipelago is 40–45 m deep within a few metres of the coastline and reaches a maximum depth of 80 m. Conversely, the western side is characterized by shallower depths (20–25 m).

Diet analysis

Diet was characterized by pellet analysis, a non-invasive method that can provide large samples over time. Dietary data were described in a preliminary way by Sponza et al. (2010), in which the key prey were identified by pellet analysis and linked with diving strategies, and this in turn served to characterize the different foraging habitats and benefits of dives. In this paper the entire data set is used for a comprehensive analysis of the diet of Mediterranean Shags during the breeding and post-breeding seasons.

Pellet collection involves little or no disturbance to birds and analyses require minimal laboratory facilities (Carss et al. 1997). This method has been widely used to characterize the diet composition of Cormorants, Shags, Gulls, Terns and Skuas (reviewed in Barrett et al. 2007) and to estimate body sizes and biomasses of the prey eaten (Duffy & Jackson 1986, Dirksen et al. 1995, Velando & Freire 1999, Eschbaum et al. 2003, Gagliardi et al. 2007, Barquete et al. 2008). However, the analysis of pellets has to consider the partial erosion of otoliths caused by the digestive process, so fish lengths derived from otoliths could be under-estimates (Johnstone et al. 1990, Carss et al. 1997). Moreover, the smallest otoliths (i.e. belonging to very small fish) could be completely digested (Johnstone et al. 1990) and otolith wear can vary from bird to bird or from time to time, so the validity of the technique will vary with the species. For example, it was shown that Gadidae, unlike Ammodytidae, have very robust otoliths (Barrett et al. 1990). However, the potential degree of otolith erosion for most of the North Adriatic fish species has not yet been characterized. Unlike some invasive method (Duffy & Jackson 1986, Carss et al. 1997), pellet analysis has been recognized as an appropriate method for the temporal and spatial comparison of the diet of Shags (Duffy & Laurenson 1983, Barrett et al. 1990, Barrett 1991), and it is to this purpose that we use data based upon these techniques in this article.

In the Gulf of Trieste, during the 2005 post-breeding period (May–October), we collected 23–30 pellets per month at each of the three roosts, except for May at the beginning of the season (14–15 pellets) (Table 1). The three roosts are: Isonzo river mouth (45°43′N, 13°33′E), Filtri (45°44′N, 13°39′E) and Lazzaretto (45°36′N, 13°42′E). At the first site Shags roost on beached, dead trees and sand banks, while in the latter two sites, birds rest on floating platforms used for mussel farming. We used a canoe to reach sand banks, beached trunks, and floating platforms, minimizing disturbance to birds. We analysed a total of 486 items. In Croatia, during the 2006 breeding season (January–April) we analysed a total of 125 pellets (Table 2 ), collected monthly near the nests at the Oruda colony. During the monthly visits we observed adult birds in breeding plumage only. All samples were fresh and showed no signs of deterioration. Upon collection, all specimens were individually sealed in a plastic bag and tagged with an identification number and then stored and frozen until processing. The probability of analysing pellets regurgitated by the same individual was negligible, given the high numbers of Shags present in the two study areas, and given that each collection was carried out in different sites and periods.

Table 1. Dietary composition of Mediterranean Shags Phalacrocorax aristotelis desmarestii in the Gulf of Trieste during the 2005 post-breeding season, expressed as numerical frequency (%N), biomass (%B) and frequency of occurrence (%O) percentage values.

	May			June			July			August			September			October
No. of pellets	44			89			90			90			90			83
No. of prey	1207			2613			3503			4591			4110			4692
	%N	%B	%O	%N	%B	%O	%N	%B	%O	%N	%B	%O	%N	%B	%O	%N	%B	%O
Gobiidae
Gobius niger	56.42	39.97	100.00	71.76	44.10	100.00	80.02	51.66	100.00	62.88	54.93	100.00	77.20	48.77	100.00	69.29	60.70	96.40
Zosterisessor ophiocephalus	5.80	29.07	65.90	6.77	33.29	89.70	6.54	28.12	85.50	6.69	28.34	90.00	11.92	32.96	92.20	4.26	19.88	83.10
Gobius bucchichii	9.78	3.79	54.50	7.65	3.31	52.30	1.88	1.18	30.00	0.35	0.26	13.30	0.75	0.48	20.00	0.40	0.35	13.20
Gobius cruentatus	2.90	8.80	45.40	1.80	5.22	31.80	0.94	3.24	23.30	0.63	3.92	22.20	0.39	2.03	14.40	0.15	1.30	6.00
Gobius cobitis	0.25	1.35	6.80	0.50	2.92	11.30	0.31	2.14	11.10	0.09	1.68	3.30	0.32	3.77	13.30	0.13	2.40	6.00
Pomatoschistus minutus	–	–	–	0.04	0.02	1.10	0.03	0.01	1.10	–	–	–	0.02	0.00	11.10	0.40	0.16	10.80
Pomatoschistus marmoratus	–	–	–	0.38	0.09	6.80	0.03	0.01	1.10	0.48	0.18	7.80	0.29	0.05	10.00	0.06	0.02	3.60
Triglidae
Lepidotrigla cavillone	–	–	–	–	–	–	–	–	–	0.02	0.01	1.10	–	–	–	–	–	–
Cepolidae
Cepola rubescens	1.08	3.16	25.00	1.15	2.78	25.00	1.34	3.33	22.20	0.74	2.73	18.90	0.95	4.23	22.20	0.94	5.06	24.10
Pleuronectidae
Platichthys flesus	0.17	0.57	2.30	–	–	–	0.03	0.19	1.10	0.35	1.06	3.30	–	–	–	–	–	–
Soleidae
Solea vulgaris	–	–	–	–	–	–	0.31	0.35	5.50	0.15	0.09	2.20	0.44	0.38	6.60	1.13	0.99	7.20
Sparidae
Oblada melanura	0.08	0.18	2.30	–	–	–	0.03	0.13	1.10	–	–	–	–	–	–	–	–	–
Diplodus sargus	–	–	–	–	–	–	0.03	0.12	1.10	–	–	–	–	–	–	–	–	–
Diplodus annularis	–	–	–	–	–	–	0.03	0.27	1.10	–	–	–	–	–	–	0.04	0.24	2.40
Pagellus erythrinus	1.24	1.02	6.80	–	–	–	0.06	0.07	2.20	–	–	–	–	–	–	0.64	0.91	10.80
Pagrus pagrus	0.08	0.48	2.30	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Dentex dentex	–	–	–	–	–	–	–	–	–	0.02	0.15	1.10	–	–	–	–	–	–
Lithognathus mormyrus	–	–	–	–	–	–	–	–	–	0.09	0.34	1.10	–	–	–	–	–	–
Boops salpa	0.17	0.31	2.30	–	–	–	0.03	0.09	1.10	–	–	–	–	–	–	–	–	–
Labridae
Crenilabrus tinca	1.08	0.60	15.90	1.91	1.06	19.30	0.77	0.73	24.40	0.78	0.52	25.50	0.41	0.19	10.00	0.60	0.61	16.80
Serranidae
Serranus hepatus	8.12	6.64	45.40	2.64	2.65	37.50	2.37	3.06	37.80	0.96	1.34	18.90	1.56	2.38	31.10	0.40	0.73	12.00
Gadidae
Trysopterus minutus	1.82	0.31	11.30	1.19	0.16	15.90	0.66	1.21	17.80	0.37	0.94	16.70	1.07	2.82	20.00	0.23	0.81	9.60
Odontogadus merlangus	0.33	0.03	4.50	0.46	0.62	4.50	0.06	0.05	1.10	–	–	–	0.05	0.90	1.10	0.02	0.03	1.20
Maenidae	0.58	0.65	9.10	0.46	0.62	10.20	0.09	0.22	2.20	0.13	0.42	3.30	0.10	0.18	1.10	0.13	0.38	2.40
Engraulidae
Engraulis encrasicholus	1.24	0.62	13.60	2.22	2.23	19.30	1.26	2.29	18.90	0.41	0.93	8.90	0.44	0.48	12.20	0.21	0.24	2.40
Mugilidae	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	0.09	0.93	1.20
Atherinidae
Atherina boyeri	8.37	1.72	18.20	0.80	0.22	7.90	2.74	0.82	5.50	24.40	1.71	37.80	3.99	0.17	10.00	20.74	3.90	53.00
Carangidae
Trachurus mediterraneus	0.41	0.70	9.10	0.19	0.36	4.50	0.34	0.26	6.60	0.46	0.44	3.30	0.10	0.20	3.30	0.13	0.37	2.40
Moronidae
Dicentrarchus labrax	0.08	0.04	2.30	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Sphyraenidae
Sphyraena sphyraena	–	–	–	0.04	0.06	1.10	0.09	0.40	2.20	–	–	–	–	–	–	–	–	–
Belonidae
Belone belone	–	–	–	0.04	0.30	1.10	0.03	0.07	1.10	–	–	–	–	–	–	–	–	–

Table 2. Dietary composition of Mediterranean Shags Phalacrocorax aristotelis desmarestii at the breeding colony at Oruda, Croatia, during the 2006 breeding season, expressed as numerical frequency (%N), biomass (%B) and frequency of occurrence (%O) percentage values.

	January			February			March			April
No. of pellets	30			30			30			35
No. of prey	900			758			971			643
	%N	%B	%O	%N	%B	%O	%N	%B	%O	%N	%B	%O
Gobiidae
Gobius niger	10.67	2.83	66.60	5.01	1.24	44.80	30.07	4.30	76.60	13.84	2.40	58.80
Zosterisessor ophiocephalus	0.11	1.16	3.30	–	–	–	–	–	–	–	–	–
Gobius bucchichii	–	–	–	–	–	–	0.10	0.07	3.30	–	–	–
Gobius cruentatus	–	–	–	0.13	0.66	3.40	–	–	–	–	–	–
Pomatoschistus minutus	1.89	0.15	10.00	0.26	0.01	3.40	2.27	0.15	6.60	–	–	–
Pomatoschistus marmoratus	0.67	0.09	6.60	0.79	0.04	13.80	2.16	0.08	16.60	–	–	–
Cepolidae
Cepola rubescens	0.33	3.09	3.30	2.11	5.35	6.90	9.78	25.64	26.60	4.04	3.94	2.90
Soleidae
Solea vulgaris	0.22	0.38	3.30	–	–	–	–	–	–	–	–	–
Scorpaenidae
Scorpaena scrofa	–	–	–	–	–	–	0.10	0.59	3.30	–	–	–
Bothidae
Arnoglossus laterna	–	–	–	–	–	–	0.21	0.29	6.60	–	–	–
Sparidae
Sparus auratus	–	–	–	–	–	–	–	–	–	0.31	2.22	5.80
Oblada melanura	–	–	–	2.11	10.96	13.80	0.21	0.91	6.60	1.40	6.38	5.80
Diplodus sargus	0.11	3.82	3.30	0.92	1.26	3.40	–	–	–	–	–	–
Diplodus annularis	1.89	7.89	20.00	5.54	16.23	44.80	0.93	3.05	13.30	0.93	3.93	14.70
Charax puntazzo	0.11	1.22	3.30	–	–	–	–	–	–	–	–	–
Pagellus erythrinus	8.00	13.68	56.60	5.28	6.72	37.90	7.62	11.53	50.00	16.02	17.09	64.70
Lithognathus mormyrus	0.11	2.42	3.30	0.13	0.97	3.40	–	–	–	1.24	9.63	11.70
Labridae
Crenilabrus tinca	9.11	16.67	60.00	3.83	11.58	62.10	12.67	18.29	83.30	24.73	27.23	94.10
Serranidae
Serranus hepatus	13.33	17.15	46.60	23.75	31.22	65.50	18.64	21.09	63.30	7.15	6.56	38.20
Maenidae	6.89	19.02	36.60	5.80	5.94	37.90	5.05	9.98	40.00	5.29	11.55	41.10
Engraulidae
Engraulis encrasicholus	–	–	–	–	–	–	0.41	0.76	3.30	0.31	0.46	5.80
Mugilidae	–	–	–	1.06	1.57	3.40	–	–	–	–	–	–
Atherinidae
Atherina boyeri	45.11	8.89	60.00	40.90	4.10	48.30	8.34	1.12	46.60	20.68	2.39	38.20
Carangidae
Trachurus mediterraneus	–	–	–	–	–	–	–	–	–	0.16	0.40	2.90
Moronidae
Dicentrarchus labrax	1.44	1.56	36.60	2.37	2.15	34.50	1.44	2.16	33.30	3.58	2.21	23.50
Scombridae
Scomber scomber	–	–	–	–	–	–	–	–	–	0.31	3.60	5.80

Table 3. Descriptive analysis of prey size (fish length, cm) in the diet of Mediterranean Shags Phalacrocorax aristotelis desmarestii in the Gulf of Trieste and at the breeding colony at Oruda, Croatia.

Family	Gulf of Trieste						Croatia
	N	Mean	sd	Min.	Max.	Mode	N	Mean	sd	Min.	Max.	Mode
Atherinidae	2475	5.02	1.82	2.25	10.91	4	930	5.55	1.62	2.96	11.08	4
Gobiidae	16 430	7.87	2.95	1.47	24.60	6	591	4.36	1.62	1.58	18.74	4
Labridae	171	7.59	1.96	3.18	14.57	7	393	9.49	2.42	6.36	23.94	8
Maenidae	42	11.16	3.31	5.76	27.52	9, 10	197	9.49	3.73	1.83	21.25	8
Moronidae	1	–	–	–	–	–	68	10.28	1.94	6.46	15.65	10
Serranidae	377	8.78	1.13	5.36	11.57	8	527	8.31	0.66	6.27	10.55	8
Sparidae	62	10.94	4.12	5.88	22.07	9	411	12.10	4.51	3.95	25.10	9
All families	20 051	7.61	3.03	1.47	29.03	6	3131	7.54	3.58	1.58	25.10	4, 8

We processed all pellets in the laboratory following a dissolution method described by Privileggi (2003). Each pellet was individually stored in a beaker (50 ml), soaked in a detergent solution, and then left to settle for about 20 hours. After the mucous had been dissolved, Isopod parasites and Nematode worms were removed and, after decanting, otoliths and any skeletal structures useful for prey identification were separated out. We used a filter in order not to lose the smallest otoliths during the outflow. To remove any mucous residue and to bleach bones, each sample was treated with a solution of sodium hydroxide (NaOH, 10%) and sodium hypochlorite (NaClO, 50%). This method ensured the conservation of the smallest bony fragments during the mucous dissolution. We then rinsed all samples with water and air dried them at room temperature before storage and examination. Where possible, we paired otoliths. We also accurately separated and identified (according to our own reference collection and to Härkönen 1986, Keller 1993, Veldkamp 1995) otoliths, scales, opercula and vertebrae to the lowest taxonomic level, by using a stereomicroscope with an ocular micrometer (6–32× zoom lens). The reference collection of diagnostic bones (Privileggi 2003) includes 79 marine (1107 individuals) and 45 freshwater fish species (1465 individuals) of different size classes. We evaluated the total length and mass of each prey by means of both original (Privileggi 2003) and published (Keller 1993, Volponi 1994, Veldkamp 1995) equations that relate otoliths length to the body size and the mass of fish. We produced an equation for each fish species, except Cepola rubescens, Belone belone, Gobius bucchichii and Sphyraena sphyraena, which were excluded from this analysis. While analysing eroded otoliths, particular attention was paid to the comparison of the morphological characters of outline, sulcus, rostrum and ostium with the reference collection, in order to estimate and adjust the size of the otoliths.

The level of taxonomic resolution for identification was different for some prey types. In this paper we treated the family Mugilidae as one category, as the species that make up this family are difficult to identify because of the extreme similarity of the otoliths. Moreover, the family Maenidae is represented essentially by Maena maena and Maena chryselis, whereas many families are represented by one species only (i.e. Atherinidae by Atherina boyeri, Labridae by Crenilabrus tinca, Serranidae by Serranus hepatus) (Tables 1 & 2).

Each prey type in the diet was estimated as numerical frequency (N), biomass (B), and frequency of occurrence (O, number of samples containing each prey type). Contingency tables tested by chi-square test (χ ²) were used to compare the numerical frequencies (not percentages) of different prey types between the two study areas and the monthly differences within each area. For this purpose, we grouped the different prey species by family. Families representing less than 1% at both study areas were categorized as ‘Other species’. Other statistical differences were assessed with Mann–Whitney U test and Spearman's rank correlation.

As regards prey size, we report the descriptive analysis only. Consequently we do not compare the two study areas and the different months within each study area, in order not to go beyond the limits of the pellet analysis method.

The significance threshold was set at P < 0.05 and the analysis was performed using spss 13.0 and statistica 7.1 software.

RESULTS

Analysis of 611 pellets revealed 23 988 identified prey. Of these, 20 716 belonged to 31 taxa in the Gulf of Trieste, whereas 3272 belonged to 26 taxa in Croatia (Tables 1 & 2). In the Gulf of Trieste, Gobiidae were the focal prey (81.5%N). The most captured species was the Black Goby Gobius niger (70.8%N), which occurred (%O) in 99.2% of the pellets analysed. The second in importance (11.9%N) were Atherinidae (i.e. Atherina boyeri) (Table 1). Conversely, in Croatia, Atherinidae were the most important prey (28.4%N), followed by Gobiidae (18.1%N), Serranidae (16.1%N) (i.e. Serranus hepatus), Labridae (12.0%N) (i.e. Crenilabrus tinca), and Sparidae (12.6%N) (Fig. 1, Table 2).

Figure 1. Numerical frequency (%N) and biomass (%B) percentages of prey in the diet of Mediterranean Shags Phalacrocorax aristotelis desmarestii in the Gulf of Trieste and at Oruda breeding colony, Croatia. Fish families are listed in the key.

Diet composition was significantly different between the two study areas (χ ² = 8894.5, df = 8, P < 0.0001) (Fig. 1). The main contributions to the total Chi-square were given by 5 (Sparidae, Gobiidae, Serranidae, Labridae, Maenidae) over nine categories. However, the main contributions were first due to the high consumption of Sparidae (family contribution/total contribution = 1860.7/8894.5) and then to the low consumption of Gobiidae (family contribution/total contribution 1346.3/8894.5) in Croatia with respect to the Gulf of Trieste. In this latter area, the differences between months were significant (χ ² = 2463.9, df = 40, P < 0.0001). In the context of a huge exploitation of Gobiidae, both by number and biomass (Table 1, Fig. 1), the monthly differences in frequency were principally due to Atherinidae during the whole period (family contribution/total contribution = 1649.8/2463.9). Such differences were particularly evident in August, when we recorded at the same time the lowest frequencies (with respect to the expected values) for Gobiidae (monthly family contribution/monthly total contribution = 60.5/692.9) and the highest for Atherinidae (monthly family contribution/monthly total contribution = 595.5/692.9). A high consumption of Serranidae was recorded in May (monthly family contribution/monthly total contribution = 263.2/385.9).

In Croatia, the differences between months were significant (χ² = 879.8, df = 24, P < 0.0001). Atherinidae (family contribution/total contribution = 280.9/879.8), Gobiidae (family contribution/total contribution = 223.2/879.8), and Labridae (family contribution/total contribution = 135.5/879.8) contributed the most to the differences. The highest contribution was due to March, when we observed an increase in the consumption of Gobiidae (monthly family contribution/monthly total contribution = 146.3/371.1) and a concurrent decrease of Atherinidae (monthly family contribution/monthly total contribution = 137.8/371.1).

In order to assess a possible relationship between frequency and biomass values, we correlated the overall data for the two study areas (Fig. 2). We found a significant correlation in the Gulf of Trieste only (Spearman's rank correlation: N _{Gulf of Trieste}  = 9, r _s  = 0.88, P < 0.01; N _Croatia  = 9, r _s  = 0.35, P = 0.36). However, at both sites, we showed two atypical points of the two more abundant taxa (circles in Fig. 2a), that are Gobiidae and Atherinidae. In Croatia, the effect of both values lead to high frequency and low biomass values. In the Gulf of Trieste this trend was confirmed by Atherinidae, while a high frequency of Gobiidae corresponded to high biomass values. If we removed these exceptions, the two regressions became similar and overlapping (Spearman's rank correlation: N _{Gulf of Trieste}  = 7, r _s  = 0.96, P < 0.001; N _Croatia  = 9, r _s  = 0.86, P < 0.02) (Fig. 2b).

Figure 2. (a) Frequency/biomass correlations in the two study areas. Black points indicate values for each one of the nine fish families. Black circles indicate the value for Gobiidae, while dashed circles indicate Atherinidae. (b) Frequency/biomass correlations with Gobiidae and Atherinidae values removed (Cep, Cepolidae; Lab, Labridae; Mae, Maenidae; Mor, Moronidae; Oth, Other species; Ser, Serranidae; Spa, Sparidae).

Considering the ecology of prey (benthic, bentho-pelagic and pelagic), we found significant differences between the two areas (χ ² = 7213.2, df = 2, P < 0.0001). In the Gulf of Trieste, Shags strongly preferred benthic fishes, both in terms of frequency (83.0%) and biomass (91.2%), whereas in Croatia there was a higher consumption of bentho-pelagic prey (frequency 46.4%, biomass 78.0%). In fact the main contributions to the total Chi-square were due to the low frequency of benthic (category contribution/total contribution = 1194.1/7213.2) and a high consumption of bentho-pelagic prey (category contribution/total contribution = 4526.6/7213.2) in Croatia.

We determined a prey size mode of 6 cm in the Gulf of Trieste, whereas in Croatia we recorded two modal sizes, at 4 and 8 cm. The larger prey sizes are recorded in May and June. Moreover, prey lengths decreased significantly within the season (Spearman's rank correlation: n = 6, r _s  = – 0.83, P < 0.05) (Fig. 3).

Figure 3. Monthly mean (± sd) size (length of fish, cm) of prey consumed by Mediterranean Shags Phalacrocorax aristotelis desmarestii in the Gulf of Trieste and at Oruda breeding colony, Croatia.

DISCUSSION

European Shags are opportunistic predators, and the variability in composition of their diet in different locations is related to geographical differences in the available potential prey (Barrett 1991, Velando & Freire 1999). Moreover, the species has a high flexibility of feeding grounds, both in terms of diving depths and bottom sediment characteristics (Guyot 1988, Wanless et al. 1991b, Velando & Freire 1999). Our study supports this evidence. Recognizing the limitations of the pellet analyses used, we highlight a wide prey spectrum, mostly composed of fish species that are found at a diversity of depths and in different habitats, including pelagic prey (such as Atherina boyeri), bentho-pelagic fishes from sandy and rocky bottoms (such as Pagellus erythrinus and Serranus hepatus) and demersal species from sandy-mud bottoms (such as the family Gobiidae). We never observed Shags following fishing trawlers nor feeding on by-catch discards. As also reported by Oro & Ruiz (1997) and Arcos et al. (2002), we consider this behaviour as very occasional in our study population.

Are Shags really opportunistic predators?

Although Shags have flexible feeding habits that allow them to exploit both benthic and pelagic resources (Grémillet et al. 1998), they should be considered as predominantly benthic feeders (Cramp & Simmons 1977, Barrett et al. 1990, Wanless et al. 1991a, 1993a, 1998, Watanuki et al. 2008). Diet analyses in most of Europe actually show that Shags rely on Sandeels (Ammodytes spp.) (Barrett et al. 1986, Harris & Wanless 1991, 1993, Grémillet et al. 1998, Velando & Freire 1999, Furness & Tasker 2000, Lilliendahl & Solmundsson 2006). Even if Sandeels sometimes form schools within the water column (Jonsson 1992), Shags catch these prey on or near the bottom (Harris & Wanless 1991, Wanless et al. 1991a, 1997b). We confirm the importance of the sea bed for Mediterranean Shags, by showing that these birds exploit prey in the Gulf of Trieste which are essentially benthic. Similarly, in Croatia, Shags forage mostly (69.0% by number) on benthic and bentho-pelagic fishes. Moreover, we emphasize that, from May to October, the period with the larger presence of Shags (Sponza et al. 2010), the Gulf of Trieste is one of the most important Adriatic areas for the abundance of pelagic ephemeral species such as Engraulis encrasicolus, Sardina pilchardus and Scomber scomber. In late spring, these species leave the central Adriatic areas and reach the Gulf of Trieste's shallow waters for spawning (Skrivanic & Zavodnik 1973, Orel & Zamboni 2004). While they represent 74% of the north Adriatic fish landings (Prestamburgo et al. 2005), these fishes represent less than 1% and 0.5% by number of Shags' diet in the Gulf of Trieste and Croatia, respectively. We therefore suggest that Shags aim at diving towards the sea bed to exploit mainly benthic, but also bentho-pelagic, food resources, and do not take advantage of abundant pelagic prey.

Diet specialization

Sandeels play a key role in the diet of Shags in most European waters and are the basis of a feeding specialization during the chick rearing period (Harris & Wanless 1993, Velando & Freire 1999). This specialization is probably linked to the high energetic value and the abundance of these fishes, which also represent the focal prey of many seabird species (Rindorf et al. 2000). Contrary to what has been found in most studies, we did not find any prey specialization during the breeding season. Rather, in Croatia, the prey spectrum is particularly wide, as we recorded the prevailing importance of five families (Sparidae, Gobiidae, Serranidae, Labridae, Maenidae).

The role of Gobiidae

Feeding specialization is conversely recorded during the post-breeding period. In the Gulf of Trieste, Shags indeed focused on demersal Gobiidae, both in terms of frequency (81.5%) and biomass (87.1%). We report slight exceptions to such dominance in August and October, when there was a decrease of Gobiidae and an increase in Atherinidae. The species belonging to both families are sedentary and perform short movements (Riedl 1991). Gobiidae are strictly demersal and poor swimmers (Louisy 2005), whereas Atherinidae are confined in estuaries and coastal lagoons (Louisy 2005). They aggregate in schools which lay at different depths (Riedl 1991). Given the role of Gobiidae, the possible temporary differences in the availability of these demersal prey could lead Shags to forage in the upper layers of the water column as well, thus exploiting other fish. Similarly, Lilliendahl & Solmundsson (2006) noted momentary changes in Shags' diet as a consequence of possible variations in the availability of their focal prey (Sandeels). Shags also shift between Gobiidae and Atherinidae in Croatia during March. The significant increase of Gobiidae could be linked to the Shags' efforts to focus on these demersal prey. Since the April data do not substantiate a further increment, this pattern could be simply linked to a temporary accessibility of these fishes. The likely non-prevalence of a fish species over another could lead Croatian Shags to exploit the currently available prey, which would bring about an increase in dietary diversity. This is particularly evident in March and April, which is at the peak of chick rearing (Sponza et al. 2010). It is important to stress that Lilliendahl & Solmundsson (2006) recorded a high dietary overlap between nestlings and adult Shags in Iceland. Moreover, Velando & Freire (1999), even if suggesting some differences in diet between nestlings and adults in Spain, showed that chicks were fed almost exclusively with Sandeels. In May and June, the main chick rearing period in Spain, Sandeels represented more than 80% by number of adults' diet. We hence maintain that Shags' dietary variability in Croatia reflected nestlings' diet.

The importance of Gobiidae can also be inferred from the analysis of frequency/biomass ratio, which highlights how these prey are a discriminating factor at both study areas. If we remove Gobiidae values, the diet indeed tends to a similar frequency/biomass ratio in the two areas. On the contrary, we record a positive effect of Gobiidae in the Gulf of Trieste only. This is due to the larger size of Gobies in the Gulf of Trieste with respect to Croatia.

Prey size

Prey sizes are slightly shorter by comparison with the prey of Shags breeding in Scotland (9.7 cm) (Wanless et al. 1993a) and Spain (9.8 cm) (Velando & Freire 1999). The most frequent prey length in the Gulf of Trieste is 6 cm, whereas in Croatia we detected two modes (4 and 8 cm length), which are similar to the modal lengths (6 and 8 cm) recorded by Wanless et al. (1993b).

Nonetheless, the most interesting aspect is the trend of the monthly mean size of prey between the two areas. We suggest that there is an increase in prey length at the end of the breeding season (April) which is probably linked to chick rearing. Conversely, larger prey are taken in May and June in the Gulf of Trieste. Thereafter, the prey size decreases. We could assume that this trend is probably related to an abundance of adult Gobies in May and June due to spawning (Patzner 2007).

Ecological perspectives

Shags summering in the Gulf of Trieste come from the Croatian breeding colonies (Sponza et al. 2010). These post-breeding movements are linked to a more profitable foraging activity in the Gulf of Trieste with respect to Croatian waters, due to shallower depths (maximum 25 m versus 80 m in Croatia) and reduced physiological stress during diving (Sponza et al. 2010). These movements have not been always a habitual pattern, since they increased consistently from the 1980s to become a real migratory movement at the present time (Sponza et al. 2010). During the 1980s and 1990s, there was a substantial decrease (67%) of demersal fisheries landings along the eastern Adriatic Sea coastline (Slovenia, Croatia, Bosnia-Herzegovina, Montenegro and Albania) (Mannini et al. 2005). We suggest that there has been over-fishing in Croatian waters and the intensive use of unselective bottom trawls have lead to a shortage of demersal fish, thus reducing the probability of Shags showing feeding specialization during breeding. The consequent necessity to exploit bentho-pelagic resources possibly forced Shags to a broader and more variable diet. Conversely, although feeding specialization is attainable in the Gulf of Trieste, Shags cannot nest there because of human disturbance and a lack of suitable sites. Nonetheless, the specialization on demersal Gobiidae allows Shags to efficiently recover from the costs of the breeding season in Croatia (Sponza et al. 2010).

This study suggests that the diet of Mediterranean Shags can be used to index the availability of fish in marine ecosystems (Thayer & Sydeman 2007, Mills et al. 2007) and to provide useful information about the fish resource conditions that can be incorporated in fisheries management models (Hislop & Harris 1985, Hatch & Sanger 1992, Barrett 2002, Roth et al. 2007).

ACKNOWLEDGEMENTS

We thank C. Trani for help in collecting pellets. Particular thanks are due to J. Kralj and M. Kulieri'c of the Institute of Ornithology of Zagreb (Croatia) for permissions and logistic support. At Veli Lošinj (Croatia) we thank Blue World staff, in particular N. Rako, P. Makelworth and Karlo, for bringing us to and from Oruda island. A PhD scholarship was granted to B. Cimador by the University of Trieste. We finally thank E.A. Ferrero for valuable criticism and we are also indebted to the anonymous reviewers for constructive suggestions and comments on a previous version.