Effect of refuge supplement on Audouin ’ s gull chick survival *

The Chafarinas Islands archipelago holds important colonies of Audouin’s (Larus audouinii) and yellow-legged gulls (Larus cachinnans), with 3,320 and 3,370 breeding pairs respectively (Ruiz et al., 2000). Earlier studies in different colonies have reported a high degree of interaction between the two species (de Juana et al., 1984; Bradley, 1988; Oro and Martínez-Vilalta, 1994). Recent studies at Chafarinas (González-Solís et al., 1997b) showed that when the two species breed syntopically, yellow-legged gulls interfere with Audouin’s gull. The most characteristic interactions are predation on eggs and chicks, aerial kleptoparasitism, courtship or chick SCI. MAR., 67 (Suppl. 2): 103-108 SCIENTIA MARINA 2003

The effects of yellow-legged gull predation on Audouin's gull productivity depend largely on the relative numbers of the two species (Oro et al., 1996a;González-Solís et al., 1997b).This is because yellowlegged gull individuals predating on Audouin's gulls are specialists in the use of this resource and their numbers are proportional to colony size.Therefore, when Audouin's and yellow-legged gulls breed in imbalaced proportions the effects can range from devastating to negligible, depending on which species dominate (Oro et al., 1996b).
At the Chafarinas Islands, the number of pairs of Audouin's gull in the colony has increased steadily since the first census in the 1960s.Despite some annual fluctuations in breeding numbers, the present population seems to be stabilised around 3,300 pairs.The yellow-legged gull population increased rapidly in the early 1980s (5,000 pairs in 1985), and a culling programme (1987)(1988)(1989)(1990)(1991)(1992) on adults and eggs was carried out to facilitate the recovery of the Audouin's gull local population.During this period, the number of Audouin's gulls rose to 4,000 pairs in 1992 (Álvarez, 1992) and that of yellow-legged gulls dropped to about 1,500 pairs in 1993.However, since then the population of yellow-legged gulls has rapidly recovered to present numbers (about 3,400 pairs), and the ratio between the two species is roughly 1:1.
One of the main factors affecting offspring survival in gulls is the presence of places to hide in the surroundings of the nest (Haymes and Blokpoel, 1978).This is because hides allow better predator avoidance and also minimise thermal stress, thus enhancing chick survival (Parsons, 1982;Parsons and Chao, 1983).At the Chafarinas Islands yellowlegged gulls start to breed one month earlier than Audouin's gulls and in general the territories occupied by the latter are less vegetated (Ruiz et al., 1996); the situation is thus fully suitable for testing the effects of supplementary refuges on Audouin's gull offspring survival and their usefulness as a management tool for conservation purposes in this breeding colony, the second largest in the world (Ruiz et al., 1996).

STUDY AREA AND METHODS
The field work was carried out in 1999 and 2000 at Chafarinas (Melilla: 35º11'N, 3º46'35''E) during the Audouin's gull chick-rearing period (June-July).The study was done at Rey Island on which more than 98% of the Audouin's gulls in the archipelago were breeding during both years.
The north area of the island, which has a poor vegetation cover but holds the largest breeding nucleus (Ruiz et al., 1996(Ruiz et al., , 2000)), was chosen for the experiment.There, we selected an area in which to carry out the experiment (enclosure area), and we used the rest of the subcolony as a reference (reference area) to determine whether the enclosure area was homogeneous with the rest of the subcolony.The enclosure area was 15 x 15 m in size, and was subdivided into nine 5 x 5 m internal enclosures delimited using 0.5 m high wire netting (Fig 1A).The netting was attached to the ground to prevent the movement of chicks between the different internal enclosures.
The experiment aimed to compare the breeding success of Audouin's gull between two areas within the enclosure: (1) control enclosures and (2) refugesupplemented enclosures, where artificial refuges were installed as a potential protection for chicks.
Artificial refuges were made of wooden boards of 1.20 x 1.20 m, supported at each end 20 cm above the ground, thus providing room for chicks to hide.In order to randomise our sample, the experimental area was defined in the field by the end of March, before the start of the laying period.Five and four of the nine 5 x 5 m internal enclosures were randomly selected in 1999 and 2000 respectively.In each of these selected enclosures, two artificial refuges were installed by dividing selected enclosures into 25 1 x 1 m squares excluding the central and the 16 peripheral 1 x 1 m squares and randomly selecting one of the remaining eight to place a supplementary refuge.The second refuge was installed directly opposite the first (Fig. 1B).The refuges were installed before the laying period to avoid the destruction of nests and an excess of disturbance during the incubation, but the netting was not erected until the incubation period.No further human disturbance took place from net placement to ringing of chicks.
In each of the nine enclosures, vegetation cover was estimated using cumulative sub-samples of 1 m 2 .For each 5 x 5 m square we then calculated the percentage of ground covered by plants.In the refuge-supplemented and control enclosures, as well as in a random sample of nests in the reference area, we recorded mean clutch size and mean egg volume in a clutch, though the latter was only measured for three-egg clutches (i.e. the modal clutch size).To estimate mean egg volume in a clutch, we used a combination of two linear measurements: Egg volume = 0.485 x length x width 2 (Ruiz et al., 1996).Breeding success was calculated as an average (i.e.dividing the number of fledglings surviving in every enclosure by the number of nests it contained) and expressed as number of chicks per pair.
Although laying date could be another factor affecting reproductive success (Wendeln, 1997;review in Moreno, 1998), we did not carry out a close surveillance of this factor in order to reduce disturbance to breeding pairs.Nevertheless, previous studies indicated that the main differences in laying date correspond to different subcolonies (Ruiz et al., 1996), so we assumed that breeding pairs within the same subcolony were relatively synchronous.

Statistical methods
To determine whether the study area was representative of the subcolony, we compared clutch size and mean egg volume between the experimental and the reference area using the Mann-Whitney and ttests respectively.
The Mann-Whitney U statistic was also used to assess differences in vegetation cover and clutch size between the refuge-supplemented and control enclosures, whilst we used the t-test to compare mean egg volumes.The breeding success in the refuge-supplemented and control enclosures was compared using a 2 x 2 contingency table stratified by year.The homogeneity of the strata was evaluated with a Breslow-Day test and the Mantel-Haenszel χ 2 statistic was used to test the association between breeding success and presence of artificial refuges.
In both study years, breeding success (chicks/pair) in the refuge-supplemented enclosures was higher than in the control enclosures, but such differences were much greater in 1999 than in 2000 (Table 3).We did not find significant differences between the strata defined by year and enclosure (χ 2 = 1.482; d.f = 1; p = 0.224).Thus, we performed a general analysis of the relationship between breeding success and refuge availability.There was a positive and statistically significant association between breeding success and presence of supplementary refuges (Mantel-Haenszel: χ 2 = 21.906;d.f.= 1; p < 0.0001).The stratified odds-ratio statistic of association was 2.384 , indicating that the probability of fledging in enclosures with additional refuges was roughly double that in control enclosures.

DISCUSSION
Chick survival depends on several factors: parental quality, food availability, clutch size, hatching order, predation rates, diseases and weather conditions (Hébert, 1987;Bolton, 1991;Hario, 1994;Kilpi, 1995;Kilpi et al., 1996;Monaghan et al., 1998).In our study, there were no significant differences in clutch size and mean egg volume between the enclosure and reference areas, indicating that parental quality was similar between them.Similarly, there were no differences for the same parameters and vegetation cover between the control and refuge-supplemented squares within the enclosure zone.We therefore considered the whole enclosure and the reference areas to be fairly similar, from the point of view of both breeders and habitat quality (i.e.vegetation cover) during the laying season.Therefore, the overall significantly higher chick-survival rate in the refuge-supplemented enclosures suggests a direct beneficial effect of such refuges.However, the fact that in 2000 this effect was largely diluted (Table 3) also indicates that factors other than predator avoidance or reduction of thermal stress account for chick survival.
Several studies on Audouin's gull and other Mediterranean seabirds have shown a relationship between food availability and fisheries activity near the colonies, which are usually well reflected in some breeding parameters and/or behavioural patterns (Bosch et al., 1994;Castilla and Jiménez, 1995;Oro, 1995;Oro et al., 1995;Oro et al., 1996a;Oro et al., 1996b;Arcos and Oro, 1996;González-Solís et al., 1997a;Oro and Ruiz, 1997;Oro et al., 1997;González-Solís et al., 1999).Although quantitative data on Moroccan fisheries and their catches near the Chafarinas Islands were not available, a lower activity of local fisheries due to bad weather 106 J. PRIETO et al. conditions and an associated food-shortage period due to scarcity of discards were observed in 2000.This was probably a major factor causing a lower breeding success in that season and diluting the effect of refuges.Moreover, other side-effects of bad weather (e.g.chilling) cannot be ruled out as accounting for the greater mortality in 2000.Thus, despite the overall relationship between breeding success and refuge availability, our data show that other factors affected the survival probability of Audouin's gull chicks at the Chafarinas Islands.Food availability is one of the main factors affecting breeding success, although its influence depends on the phase of the breeding period affected.In the Ebro Delta, a site located 800 km further north, which holds the largest colony of Audouin's gull in the world, clutch size was reduced when the laying period was affected by food scarcity, and productivity fell when the chick rearing period was affected (Oro et al., 1996a).
In our study there were no differences in egg or clutch size between 1999 and 2000, so any food shortage period at Chafarinas would only have affected the chick-rearing period.Low food availability may influence feeding rate, adult attendance, intraspecific kleptoparasitism and predation by yellow-legged gulls.We suggest that at Chafarinas food availability was the main factor affecting Audouin's gull breeding success.However, other factors such as refuge availability, which significantly increases offspring survival by providing protection against predation and overheating, could also be important, particularly in the breeding seasons without the influence of factors like reduced food availability.
Other studies have shown benefits of refuge availability for chicks of terns or gulls (Parsons and Chao, 1983;Burness and Morris, 1992;Recorbet and Bonaccorsi, 1996), but our study is the first to quantify such benefits for the survival of Audouin's gull chicks.The results suggest that artificial refuges might be useful tools for the management and protection of this vulnerable gull species.

TABLE 1 .
-Percentage distribution of Audouin's gull clutch size in the enclosure area (control and refuge-supplemented enclosures) and the reference area in both years of study.CS1 to CS ≥ 4 refer to clutch sizes from one to four or more eggs respectively.