Temporal variation on the diet of the South American Tern ( Sterna hirundinacea , Charadriiformes : Laridae ) on its wintering grounds Variación temporal en la dieta del Gaviotín Sudamericano ( Sterna hirundinacea

The diet of the South American Tern (Sterna hirundinacea) and its seasonal variation during the 2005 and 2006 non-reproductive seasons in the Uruguayan Atlantic coast was analyzed. Diet was assessed by the analysis of pellets collected in the Rocha lagoon sandbar, a major tern roosting area in Uruguay, aiming to analyze the hypothesis that terns feed mainly on the Argentine Anchovy (Engraulis anchoita) during the wintering period. A total of 844 pellets were collected, 442 in 2005 and 402 in 2006. Diet was composed of fi sh (88 %), insects (9 %) and crustaceans (3 %). The main fi sh species consumed was the Argentine Anchovy (77.7 %), followed by Marini’s Anchovy (Anchoa marinii) (7.9 %) and the Striped Weakfi sh (Cynoscion guatucupa) (3.6 %). Insects and crustaceans were the main items in two of the groups of pellets collected during the study period. Despite this temporal variation in the diet, the Argentine Anchovy was the main prey item consumed during both seasons. These results support the hypothesis that South American Terns strongly depend on anchovies as trophic resource.


INTRODUCTION
Seabirds are top predators that can provide substantial information about changes in lower trophic levels, which is fundamental for the comprehension of marine ecosystems (Montevecchi & Myers 1996, Fur ness & C a m p h u y s e n 1 9 9 7 ) .T h e c o m p o s i t i o n of seabirds' diet is usually variable among years, seasons or even weeks (Shealer 2002).Long-term quantitative studies on the diet of seabirds, which detail prey size and their energetic contents, are impor tant in order to provide information on diet variations in relation to the availability and abundance of prey (Barrett et al. 2007).The diet of many species of migratory seabirds, including terns, is well known during the breeding season, but information is scarce during the non-breeding season (Barrett et al. 2007).However, the diet of terns can be studied relatively easy in their roosting areas through pellet analysis (Barrett et al. 2007).
T h e S o u t h A m e r i c a n Te r n ( S t e r n a hirundinacea Lesson) is a migrator y species endemic to South America.In Ur uguay it is a winter visitor (March-September) that uses the coastal areas along the Atlantic coast for roosting and feeding (Escalante 1970, Azpiroz 2003, Alfaro & Clara 2007).Migratory movements of this species, as well as the origin of the fl ocks observed in Uruguay and in the region during winter, are poorly known.However, due to their phenology some authors have proposed that they come from the breeding colonies in Patagonia, Southern Argentina (Escalante 1970, Favero et al. 2000, Bugoni & Vooren 2005, Alfaro & Clara 2007, Gatto et al. 2009).Rocha lagoon is one of the most important roosting areas for this species in Uruguay, where flocks of approximately 2000 individuals are common during the austral winter (Alfaro & Clara 2007).Favero et al. (2000) studied the diet of the South American Tern in three different wintering areas in Buenos Aires Province, Argentina, and determined that the diet was composed of fi sh (mainly the Argentine Anchovy), crustaceans and insects.Temporal variation was not analyzed by these authors but they found significant variation in the propor tion of fish consumed between the localities analyzed.Diet on breeding grounds in Patagonia has not been published yet, but unpublished information confi rms it is mainly composed of the Argentine Anchovy (Gatto & Yorio 2009).
The Argentine Anchovy (Engraulis anchoita Hubbs & Marini) is distributed in neritic areas of the south-western Atlantic Ocean from Rio de Janeiro (Brazil) to San Jorge Gulf (Argentina) (Hansen 2000).This species reproduces in Uruguayan coastal areas during winter (June-August), then, during spring and summer adults migrate southward to Patagonia, but larval and juvenile age classes stay in the area (Nion & Ríos 1991).Although it is recorded during the entire year in Uruguayan waters, its abundance is signifi cantly higher during winter and mainly composed of adults (Ehrhardt et al. 1977, 1978, Sánchez & de Ciechomski1995, Hansen & Madirolas 1996, Hansen et al. 1996, Cousseau & Perrotta 2000).
Because the Argentine Anchovy was the main prey consumed by the South American Ter n in Ar gentina and their seasonal movements between the breeding and nonbreeding areas coincide with tern movements, the possibility of a predator-prey relationship between these two species was suggested (Favero et al. 2000, Silva Rodríguez et al. 2005).Similar predator-prey relationships have been observed in the breeding areas of pelicans and terns (e.g., Anderson et al. 1982, Schaf fner 1986).In these cases, a positive relationship between anchovy abundance and the reproductive success and abundance of seabirds was observed (Anderson et al. 1982, Schaffner 1986).Favero et al. 2000 suggested that during migration between breeding areas in Patagonia and wintering areas in Buenos Aires and Uruguay, South American Terns are taking advantage of the anchovies, following this abundant resource.According to this suggestion, we tested the hypothesis that South American terns feed mainly on Argentine Anchovies during the non-breeding season in Uruguay.To evaluate this, we determined diet composition, main prey, temporal variation, and feeding strategy of the South American Tern during two years in the Atlantic coast of Rocha, Uruguay.The technique used to assess the diet was the analysis of terns' pellet content, which has some biases associated that were also evaluated.

METHODS
Diet of the South American Tern was assessed by the analysis of regurgitated pellets collected during two consecutive winter seasons (2005 and 2006) in their roosting areas on Rocha lagoon sandbar (34º31'-34º38' S and 54º12'-54º22' W) (Fig. 1).Rocha lagoon is a coastal, shallow and saline lagoon with a sandbar in the southern part.This sandbar is temporarily connected with the Atlantic Ocean through a mouth located at the southwestern tip of the sandbar (Fig. 1).The tern's roosting areas are located on the sandbanks that appear inside the lagoon close to the mouth (where the lagoon water level is shallow) and in the sand dunes and shorelines along the sandbar.
Pellet analysis is a common, relatively easy and non-invasive way of studying the diet of many species of seabirds.Biases associated with this technique include pellet identifi cation, digestibility of the otoliths, underrepresentation of soft prey species, identifi cation of the prey, and sample fi delity (Duffy & Jackson 1986, Barrett et al. 2007).All of these biases were considered in this study as described in the following methodology.
A total of 844 pellets was collected from May to September Pellets were individually stored in plastic bags and frozen at -20 ºC until analysis.Total length and width of 107 pellets and the dry weight of 224 pellets were recorded as part of pellet description and separation from other species.A stereomicroscope (14x) was used to separate and identify the remains found in each pellet (e.g., insect's jaws and heads, and fi sh's otoliths).Sagitta otoliths, crustacean and insect remains were identifi ed using bibliography (Torno 1970, Freyre & Mollo 1987, Martínez & Monasterios de Gonzo 1991, Volpedo & Echeverria 1997) and compared with reference material from three scientific collections: INIDEP (Instituto Nacional de Investigación y Desarrollo Pesquero, Mar del Plata, Argentina), UNMLP (Universidad Nacional del Mar del Plata, Argentina) and UDELAR (Universidad de la República, Facultad de Ciencias, Uruguay).
The number of fi sh per pellet was estimated using the maximum number of hypural bones, otolith pairs or eye lense pairs.The total number of insects was estimated counting the maximum number of heads, proboscis, and pairs of eyes or jaws.Finally, the number of crustaceans was estimated using pairs of crab's chelipeds, number of entire heads or crustacean's juveniles or estimating the number of adults considering a minimum of one adult per pellet if only exoskeleton remains were present.Other soft prey species that could be detected in the diet are Cephalopods that can be detected and identifi ed using their beaks.With all the items used to identify soft prey species, we expected no signifi cant underrepresentation of them.
Otolith length was measured to estimate the total length and mass of each fi sh species using regression equations previously described (Baldás et al.1997, Bugoni 2001, Bugoni & Vooren 2004, Saccardo et al. 2005, Mauco 2006) (Table 1).When an otolith was broken or eroded, the total length was estimated comparing them with well preserved otoliths of the same species.For fish species without regressions available and with less than 1 % of the abundance (N %) in the diet, their mass was estimated to be the same as the mean weight of the other fi shes.Crustacean mass was estimated using the mean mass of fresh adults and juveniles collected in the study area.Insect mass was estimated in a similar way, weighing fresh insects of different taxonomic categories and assigning to each prey insect the mean mass which corresponds to its species.
For each item, the frequency of occurrence (F) (calculated as the number of pellets in which the item was found), the abundance by number (N), and the mass (M) (estimated by the above equations) was calculated (Duffy & Jackson 1986, Bugoni & Vooren 2004).The importance of each item in the diet was determined using the index of relative importance IRI = F % (M % + N %), where F %, M % and N % are the percentage of frequency of occurrence, mass and abundance by number in the pooled sample (Bugoni & Vooren 2004).Inter-annual variation of each category N %, F %, M % and IRI % of the most important species was analyzed using the G-test of independence (G = 2 ∑ [observed values x ln (observed values/expected values)]) based on the null hypothesis of no signifi cant differences between years (Sokal & Rohlf 1987).
To analyze the feeding strategy (specialistgeneralist) of the South American Tern, we used the modifi cation of Costello's graphical analysis as proposed by Amundsen et al. (1996).In this method the preyspecifi c abundance and the frequency of occurrence are plotted in a two-dimensional graph.Prey-specifi c abundance (P i ) is the percentage of a prey taxon among all prey items in only those pellets in which the prey occurs, P i = (∑S i /∑S ti ) 100 where S i is the pellet content (in terms of number) of prey i, and S ti the total pellet content in only those pellets with prey i.The distribution of the points on the graph indicates the feeding strategy, prey importance, and niche width of the predator.The method was designed to analyze stomach contents but in this case it is used to analyze pellet content.Sample rarefaction curves (sample vs. prey species richness) were built to test sample fi delity (if the number of pellets collected per sample was enough to reach prey richness) (Gotelli & Colwell 2001).Temporal variability of the diet was analyzed using Shannon diversity index H (H = -∑ (ni n -1 ) ln(ni n -1 ), where ni = number of individuals of the species i in a sample, and n = total number of individuals registered in the sample) and Shannon equitability (H/ln(S), S = total richness of the sample).We compared the confi dence intervals of each index through Boostrap analysis using PAST software (Hammer et al. 2001).Temporal variation in the importance by number (N %) of the main fi sh items was analyzed with a G-test of independence (Sokal & Rohlf 1987).

General aspects of the diet
T h e p e l l e t s c o n t a i n e d f i s h r e m a i n s (scales, ver tebrae, otoliths, eye lens, jaws and undigested fish tails), insect remains, crustacean remains and feathers.Fish remains were found in 83.6 % of all pellets, otoliths in 50.2 %, insects in 16.1 %, and crustaceans in 12.4 %.The mean number of prey species per pellet was 4.27 ± 5.33 (SD) (N = 832).This amount was biased by a high abundance of insects found in the diet in 2005.Considering only fi sh and crustaceans, the mean number of prey species per pellet is 3.02 ± 2.13 (SD) (N = 771).Mean pellet length and width was 18.68 ± 3.26 mm (SD) (N = 107) and 11.84 ± 1.9 mm (SD) (N = 107), respectively.Mean dry weight was 0.57 ± 0.21 g (SD) (N = 224).
Ten fi sh species and 13 insect items were found in the diet during the 2005 season.The Regressions (otolith-total length, otolith-body mass, total length-body mass) used for the main fi shes found in the diet of the South American Tern.TL = total length, OL = otolith length, M = body mass.ªBaldás et al. (1997( ), °Bugoni (2001)), * Mauco (2006), b Saccardo et al. (2005).
Regresiones (otolito-largo total, otolito-masa corporal, largo total-masa corporal) utilizadas para los principales peces encontrados en la dieta del Gaviotín Sudamericano.TL = largo total, OL = largo del otolito, M = masa corporal.ªBaldás et al. (1997( ), °Bugoni (2001))  Composition of the diet and frequency of occurrence (F, F %,), abundance (N, N %), mass (M, M %) and Index of relative importance (IRI, IRI %) for each of the items in the diet of the South American Tern in Rocha, Uruguay.Items in bold have more than 1 % of the IRI in the diet.Asterisk * shows the species which had their body masses estimated using the mean mass of all other fi shes.
Composición de la dieta y frecuencia de ocurrencia (F, F %), abundancia (N, N %), masa (M, M %) e índice de importancia relativa (IRI, IRI %) para cada uno de los ítems encontrados en la dieta del Gaviotín Sudamericano en Rocha, Uruguay.Los ítems en negrita fueron los que presentaron más del 1 % del IRI en la dieta.El asterisco * indica las especies en las cuales la masa corporal tuvo que ser estimada utilizando el promedio de la masa de todo el resto de los peces encontrados en la dieta.most important prey species was the Argentine Anchovy (70.2 %) followed by the Giant Water Bug (16 %) and Marini's Anchovy (9 %) (Fig. 2).In 2006 the diet was composed of nine fi sh species, six insect species, and one crustacean species.The most important prey species was again the Argentine Anchovy (63 %) followed by the Mole crab (24.4 %), the Striped Weakfi sh (7 %) and Marini's Anchovy (4.6 %) (Fig. 2).There were no signifi cant differences between the IRI % of the Argentine Anchovy (EAN) and Marini's Anchovy (AMA) between years (EAN, G = 0.396; AMA, G = 1.464; df = 1, P > 0.05 for both), however, the Striped Weakfish IRI % differed between years (G = 5.963, df = 1, P < 0.05).Fish mean total length was 103.6 ± 30.6 mm (SD) (N = 453), and 89.6 ± 35.8 mm (SD) (N = 491) in 2005 and 2006, respectively.The largest insect found in the diet was the Giant Water Bug, with a total body length of 20.8 mm, and the smallest corresponded to a species of the Family Curculionidae (4.6 mm total body length).The Mole crab was the only species of crustacean found in the diet.Its estimated mean size was 20 mm total body length and 3.0 g mass in adults, and 5.5 mm and 0.05 g in juveniles (we considered juveniles to those crabs whose exoskeleton was still soft).
Fish mean body mass consumed during the 2005 season was 9.8 ± 6.8 g (SD) (N = 453), while it was 7.7 ± 6.9 g (SD) (N = 491) in 2006.Insects body mass ranged from 0.02 to 0.62 g; however, the most abundant insect in the diet was the Giant Water Bug with a mass of 0.2 g.
The feeding strategy diagram for the South American Tern (N = 844 pellets) shows an accumulation of points in the lower left corner with a spread to the upper left and to the center of the graph (Fig. 3).According to the model, this demonstrates a situation in which there is a high between-phenotype contribution (BPC) to the niche width, which means that individual predators have specialized on different prey types, the Giant Water Bug, the Mole crab and the Argentine Anchovy.However, there is a clear dominance of the Argentine Anchovy but also many rare prey species.In general, the graph shows a mixed feeding strategy, with var ying levels of specialization and generalization on different prey types.

Sample fi delity
Most of the species accumulation cur ves were uniform with around fi ve to six expected species (Fig. 4).However, samples (one sample means the pool of pellets collected on one day) from July 2005 and September 2006 showed a trend to increase, indicating that sample size (number of pellets collected per day) was insuffi cient.Those samples had lower number of pellets than the others: 27 in July and 40 in September.According to the shape of the curves, between 60 and 80 pellets would have been required to reach 60 to 100 % of the expected species richness.Nevertheless, depending on the species of prey ingested lower samples sizes could be enough.The sample of 5 August 2005 (161 pellets) showed the highest richness, caused by a great variety of insects consumed.In contrast, May 2006 (105 pellets) showed the lowest expected richness, due to the consumption of almost exclusively one species, the Mole crab.

Temporal variation
The Shannon diversity index was not signifi cantly different between most of the samples from 2005 (Boot P < 0.05), while in 2006 the diversity of May and July were statistically different from the majority of the samples (Boot, P < 0.05) (Table 3).The low evenness and diversity of 5 August 2005 and May 2006 samples was due to the dominance in abundance of the Giant Water Bugs in August (44.5 % IRI) and the Mole crab in May (IRI % 99.9).September and June 2005 were also different from the others because of their high diversity and evenness of fi sh prey.
During the 2005 season the N % of the Argentine Anchovy, Marini's Anchovy and Striped Weakfi sh (CGU) was statistically different between samples (EAN, G = 67.865;AMA, G = 29.563;CGU, G = 7.977; df = 4 for EAN and AMA and df = 3 for CGU, P < 0.05 for all) (Fig. 5).However, in 2006 the proportion of fi sh consumed was uniform along the season (EAN, G = 1.997;AMA, G = 5.345; df = 2, P > 0.05 for both), although an increase in the abundance of the Striped Weakfi sh by the end of the season was clearly observed (G = 14.984, df = 2, P < 0.01) (Fig. 7).The main difference in the N % of fi sh consumed between years was the increase of the abundance of the Rough Scad (Trachurus lathami Nichols) and the South American Silverside (Odontestes argentinensis Valenciennes) in 2006 (Fig. 5).

DISCUSSION
The diet of the South American Tern on the Rocha coast during the two studied seasons was mainly composed of the Argentine Anchovy followed by other species of fi sh, insects and crustaceans, as had already been obser ved on the nearby Atlantic coast of Buenos Aires Province by Favero et al. (2000).However, during each season the proportion of anchovies consumed was variable.Two other fi sh species, Marini's Anchovy and Striped Weakfi sh, were also important in the diet, being occasionally found in the same or in higher proportions than the Argentine Anchovy.
The distribution of the Argentine Anchovy along the coast of Argentina and Ur uguay varies throughout the year as a consequence of reproductive and feeding migrations, resulting in dif ferential distribution patterns of age   et al. 1977, 1978, Hansen & Madirolas 1996).Those variations in Anchovy abundance are probably the cause of the variations obser ved in the diet of the terns between years and seasons.Many seabirds are generalist and opportunistic; consequently their diet correlates positively with the abundance of their prey (Montevecchi & Myers 1996, Furness & Camphuysen 1997).According to the feeding strategy diagram, the South American Tern has a mixed feeding strategy, which according to the approach by Amundsen et al. (1996) can be considered a generalist strategy but with different individuals specialized on different food types (Amundsen et al. 1996).Therefore, the consumption of prey other than the Anchovy could be the consequence of a combination between an increase in the abundance of the new prey, better accessibility to it and/or a decrease in the relative abundance of the Anchovy.
Two other important species consumed by the South American Tern on the Rocha coast were the Giant Water Bug and the Mole crab.The consumption of insects and crustaceans found in some of the samples could be due to an elevated abundance of them in the area, but not necessarily due to a decrease in the Anchovy's abundance.Many studies about the diet of Common Terns (Sterna hirundo Linnaeus) in the region also found large quantities of insects during certain periods (Mauco et al. 2001, Bugoni & Vooren 2004, Mauco & Favero 2004, Mauco 2006).In those cases the presence of insects in the diet has been explained by the effect of meteorological and/or biological factors which generate increases in their abundance (Mauco et al. 2001, Bugoni & Vooren 2004, Mauco & Favero 2004, Mauco & Favero 2005, Mauco 2006).
Mole crabs were previously recorded in the diet of gulls and terns (Mueller 1976, Blokpoel et al. 1989, Blokpoel et al. 1992).In fact, this feeding behavior is not unusual and occurs when there is an increase in the abundance of this crustacean species (Muller 1976, Blokpoel et al. 1989, Blokpoel et al. 1992).The preference for Mole crabs instead of fi sh is because they are easier to catch when they are abundant and they comprise a good quality meal (Mueller 1976).Studies about the Mole crab in Uruguay conclude that their reproductive cycle is complex and variable and depends on the physical characteristics of the environment where they live (Delgado & Defeo 2008).Periods of high abundance of this species have been detected mainly between February and April and less frequently between May to June (Delgado & Defeo 2006).Terns consumed Mole crabs in May 2006, suggesting high abundance of them around these dates.
Rarefaction cur ves showed that samples from June 2005 and September 2006 did not reach the asymptote, indicating that the sample size at these days was insuffi cient.However, the number of pellets needed to reach the expected species richness of prey depends on the type of prey eaten, which varies between samples.On 5 August, if we had collected 80 pellets we could only had reached 60 % of the expected richness.However, this sample size is suffi cient to reach the expected richness for most of the other samples collected.This rarefaction analysis showed two key points to take into account when developing diet studies; it is important to calculate how many pellets are needed to detect most of prey species, and it is necessary to collect samples throughout the season because of prey variability.Temporal variation in the diet of the South American Tern was mainly due to the ingestion of insects and crustaceans.This marked shift in the diet shows the ability of terns to quickly change their foraging strategy to exploit an occasional resource.Despite these few exceptions, the dominant species in the diet of South American Terns was the Argentine Anchovy.This result supports our hypothesis about the Anchovy as the main item in the diet of South American Terns in Uruguay, and also supports the hypothesis that terns follow the Anchovy's migration and come from the tern breeding colonies in Patagonia.Recent studies about the diet of South American Terns in the reproductive areas in Patagonia also found the Argentine Anchovy as their main prey (Gatto & Yorio 2009).This study confi rms that there is a strong predator-prey interaction between the South American Tern and the Argentine Anchovy throughout their distribution.
, 442 in 2005 and 402 in 2006 non-breeding seasons.To determine the temporal variation of the diet during each season, pools of pellets were collected throughout the season.Dates and number of pellets collected at each sampling date are as follows: 1 June (27), 20 July (102), 5 August (161), 16 August (75) and 9 September (77) in 2005, and 25 May (105), 3 July (116), 3 August (141) and 5 September (40) in 2006.All pellets were collected from mono-specifi c fl ocks in order to avoid confusion with pellets of other, similar species.

Fig. 1 :
Fig. 1: Map of the study area.The circle shows the South-West tip of the Rocha lagoon sandbar where South American Terns' pellets were collected in 2005 and 2006.Mapa del área de estudio.El círculo muestra la punta suroeste de la barra de la laguna de Rocha donde los bolos del Gaviotín Sudamericano fueron colectados en 2005 y 2006.

Fig. 2 :
Fig. 2: Index of relative importance (IRI %) of the most important prey species in the South American Tern diet during 2005 and 2006 seasons.

TABLE 3
Shannon's diversity index and evenness from 2005 and 2006 samples.