Introduction

Flexible time budgets allow individual animals to maintain the nutritional status of their offspring at an acceptable level. Many seabirds, for example, allocate more time to foraging when food density is low or when food availability decreases because of environmental conditions. Seabirds performing long feeding trips do not normally regulate their foraging investments very much to compensate for the current breeding requirements (but see, for example, Bolton 1995; Weimerskirch et al. 1997; Granadeiro et al. 2000) because the nutritional status of their chick is a poor indicator of the energetic needs during the next return to the colony (Hamer and Hill 1994; Weimerskirch et al. 1995). Many short-ranging seabirds, however, do exhibit considerable behavioural flexibility in foraging investment. In particular, Alcidae are known to modify foraging time at the expense of time spent at the colony in order to maintain provisioning rates or reproductive success (Monaghan et al. 1994; Harding et al. 2007; Ronconi and Burger 2008; Burke and Montevecchi 2009). They are able to fine-tune the provisioning to meet the nutritional needs of their chicks (e.g. Bolton 1995; Granadeiro et al. 2000) or to adapt to variable environmental conditions (Waugh and Weimerskirch 2003; Jodice et al. 2006; Burke and Montevecchi 2009).

While short-ranging seabirds that carry multiple prey items to the colony are known to adjust provisioning rates to compensate for a decrease in chick growth or body condition (Tveraa et al. 1998; Wernham and Bryant 1998; Harding et al. 2002), single-prey loaders can only adjust prey length, select for more energy-rich species or increase foraging effort (Harding et al. 2007; Limmer and Becker 2009; Dänhardt et al. 2011), but, compared to multiple-prey loaders, they face restricted abilities to increase the numerical prey input per unit of time. Single-prey loaders may thus show more variation in the time spent foraging, and may react sooner to a poor nutritional status of their offspring than multiple-prey loaders. On the other hand, life-history theory predicts that seabirds are fixed investors that should maximise their own survival independently of breeding requirements (Navarro and González-Solís 2007). Indeed, several studies with experimentally handicapped individuals have shown that seabirds do adjust their breeding efforts to compensate for the extra costs of the handicap, but not at the risk of incurred survival probabilities (Sæther et al. 1993; Weimerskirch et al. 1995, 1999; Navarro and González-Solís 2007; Bijleveld and Mullers 2009). None of these studies were performed on single prey-loaders, while their foraging methods are energetically more expensive as they must return to the colony each time they catch a suitable prey for their offspring, and as such they may reach a certain limit at which their future survival comes at risk more quickly.

The single-prey loading Sandwich Tern Thalasseus sandvicensis is restricted to foraging areas up to 15–25 km from the colony (Veen 1977; Fasola and Bogliani 1990; Stienen 2005). The parents alternate brooding and foraging duties (Veen 1977), and this synchronized behaviour results in a constant presence of one of the adults at the colony site. However, when food intake of the chicks is not sufficient, both parents may leave the colony for foraging and the brood is left unattended more often. Stienen and Brenninkmeijer (2002) suggest that this flexible foraging behaviour allows the parents to buffer against variation in feeding conditions and to counterbalance periods of food shortage for their offspring. In the Sandwich Tern colony of Griend, The Netherlands, kleptoparasitism or food robbery by Black-headed Gulls Chroicocephalus ridibundus seems to be the main driver of parental nest non-attendance. In some years, high proportions of the food that was brought to the tern chicks was robbed by the gulls, causing decreased energy intake by the chicks (Stienen et al. 2001; Stienen and Brenninkmeijer 2002). In such years, parents apparently tried to compensate for the loss by increasing the time spent foraging and leaving their chicks unattended at the nest more often (Stienen and Brenninkmeijer 2002). In this paper, we try to further unravel the decision rules that underlie the flexibility of time allocation to foraging in Sandwich Terns using individually-based data (instead of averaged yearly values as was done by Stienen and Brenninkmeijer 2002) on chick body condition, chick feeding and parental nest attendance, and by comparing the Griend results with unpublished data from Zeebrugge, Belgium.

It is hypothesized that parental foraging effort increases with a decrease in the body condition of the chicks, and that parental colony attendance is likewise correlated to the body condition of their chicks. We further predict that parents use their own body mass to evaluate future fitness costs (cf. Navarro and González-Solís 2007) so that the degree of flexibility in parental foraging effort depends on adult body reserves. Although regulation of offspring provisioning is generally thought to be related to species-specific constraints imposed by central place foraging, some studies suggest different responses within the same species (Weimerskirch et al. 1999; Phillips and Croxall 2003). To test whether such intra-specific variations in parental responses are linked to different feeding conditions as suggested by Weimerskirch et al. (1999, 2001), we compared the foraging flexibility of Sandwich Terns at two major colony sites with very different feeding characteristics, both in terms of the rate of energy input into the colony and kleptoparasitism by neighbouring gulls.

Methods

Study sites

Research was conducted on Griend from 1992 to 1998 and at Zeebrugge between 2001 and 2007. Griend, an uninhabited island of about 57 ha in the Dutch Wadden Sea (53°15′N, 5°15′E), hosted 5600–8300 pairs of Sandwich Terns during the study period (Table 1). The Sandwich Tern colony at Zeebrugge held 46–4067 pairs and was situated within the port of Zeebrugge (51°21′N, 3°12′E). During the respective study periods, Griend held between 16,000 and 28,500 breeding pairs of Black-headed Gulls, whereas Zeebrugge hosted only 388–2390 pairs. On Griend, the numerical proportion of clupeids transported to the study chicks amounted to 30.8–74.7 %, while generally higher proportions (66.9–91.7 %) of clupeids were brought to the chicks at Zeebrugge (Table 1). Prey length was comparable between the two colonies and normally averaged between 9 and 11 cm. Only in 2006 and 2007, relatively small prey (average of less than 8 cm) were brought to the chicks at Zeebrugge. The number of prey brought to the colony varied between 4.8 and 17.9 prey per day, and was especially high at Zeebrugge in 2006 and 2007 when smaller prey items were brought to the chicks. Sandwich Terns breeding on Griend use the waters around the islands of Terschelling and Vlieland for foraging, and most individuals stay within 15 km distance from the colony (Stienen 2005). At Zeebrugge, smaller prey is caught within a few kilometres of the colony in the port itself or at the port entrance (own observations), while larger prey is caught at sea mainly within 25 km of the colony (Vanaverbeke et al. 2011).

Table 1 Variation in the number of Sandwich Terns Thalasseus sandvicensis and Black-headed Gulls Chroicocephalus ridibundus, and some general feeding characteristics of the Sandwich Tern chicks in the colonies of Griend and Zeebrugge during the study period
Full size table

Body condition of chicks

Each year, one more or less circular enclosure (diameter 3–8 m, height 50 cm) containing all study nests was constructed in/near the centre of a colony or a subcolony which was established during peak laying. The enclosure was constructed just before hatching of the first eggs. During the pre-construction phase of the enclosures, the nests within the selected study sites were marked with numbered sticks, and eggs were marked in sequence of laying with a permanent marker and controlled every 2 or 3 days. At the first control post-hatching, the chicks were aged and individually ringed with a metal ring. For observations on chick feeding and parental behaviour, chicks were dyed with picric acid or ink for individual recognition. To prevent overcrowding and negative effects on chick growth, we regularly put some chicks that were not colour dyed out of the enclosure so that the density inside the enclosure resembled the density outside the enclosure.

Each year, the body mass (±1 g) of the enclosed chicks was measured with an electronic balance preferably every 3 days until the chicks left the enclosure (fledged), died or were preyed upon. Because measurements of body mass were often not performed on days when observations on food were made, body mass was imputed for each missing measurement assuming linear growth between two successive measurements. For each observation day and for each chick (Table 2), body condition index (BCI) was computed using the relative deviation from the average body mass of all chicks (at both study sites and during all study years) that were of similar age and that ultimately fledged (i.e. max. age ≥25 days).

Table 2 The yearly number of chicks of which the nutritional state was monitored and observational protocols were made, and the yearly number of adults that were caught on the nest in the colonies of Griend and Zeebrugge in May and June
Full size table

Body condition of adults

Since body mass of parents of individually followed chicks was largely lacking, we used averaged yearly values of adult body mass (BMAavg) as a proxy for yearly body condition. Each year, 9–216 adults (mean = 66.6 ± 53.7 SD) were randomly caught at the nest during May and June of which the body mass (±1 g) was measured with an electronic balance (Table 2). In 2002, when the Zeebrugge colony was reduced to just 46 pairs, no attempts were made to catch adults. All adults were caught a few days before expected hatching during the pipping process of the egg(s). In this way, we accounted for possible confounding effects of decreasing body mass during incubation (Wendeln and Becker 1996).

Chick feeding and parental attendance

Food transport to the chicks was followed from a hide placed at a few metres from the enclosure. Observational protocols lasted 60–1084 min per day (mean = 642.1 min ± 336.4 SD); protocols of less than 1 h were omitted from the analyses. We recorded prey species (two types, namely Clupeidae Clupea harengus/Sprattus sprattus and Ammodytidae Ammodytes spp./Hyperoplus lanceolatus), prey length and the fate of the prey (consumed or not consumed by the chick). Prey species, other than Clupeidae and Ammodytidae, were omitted from the analyses since they amounted to less than 1.1 % of all food items supplied to the chicks (Stienen et al. 2000). Prey size was estimated in quarters of the parent’s bill size (bill length = 5.43 ± 0.25 cm; Stienen et al. 2000), and estimating prey size was calibrated among the observers. The first author (E.S.) participated in observations at both study sites during all years while A.B. and W.C., respectively, participated in observations on Griend and at Zeebrugge. Other more occasional observers (seven on Griend and three at Zeebrugge) are mentioned in the Acknowledgements.

The fate of the food items brought to the colony was categorized as consumed or not consumed by the chick. A very small proportion of the prey items brought to Sandwich Tern chicks (<2 % of all prey brought to the chicks on Griend) was dropped on the ground or was eaten by the parent itself, but the main reasons for not consuming a prey were robbery of food items by Black-headed Gulls or sometimes other kleptoparasites, losing prey out of the observer’s sight when the prey-carrying parent was chased by a kleptoparasite or eating of prey by the parent when it was attacked by a kleptoparasite (Stienen et al. 2000). The proportion of prey items that was consumed by the chick can thus be used as an estimate of the kleptoparasitic pressure. To calculate the daily amount of energy transported to the chicks (DET), the daily amount of energy actually consumed by the chick (DEI) and the proportion of energy that was consumed by the chick (CP), prey were transformed into energy values according to the allometric relationships given by Stienen and Brenninkmeijer (2002).

During the protocols, the presence of one or more parents near the nest was monitored continuously. If one or more parents were present in the (sub)colony under study, there was regular interaction with the chick or among partners so that parents could be easily linked to an individual chick. When no parent was present near the nest, most often the chick was hiding in the vegetation or crouched motionless on the ground. In this paper, we distinguish between the presence (one or two parents present near the nest) and absence (no parent present near the nest) of parents, in the latter case assuming that both parents went out foraging. Daily energy supply per parent (DETpar) was computed as follows:

$${\rm DET}^{{\rm par}}=({\rm DET})/(2 \times {\rm PU}+P)\times 100$$

where PU = proportion of the time the chick was left alone at the colony site and P = the proportion of the time the chick was attended by one parent.

Statistical analysis

Data on chick body condition as well as observational data on parental nest attendance and food were restricted to nests with only one chick. Chicks younger than 15 days were excluded from all statistical analysis as they are usually attended by at least one parent (cf. Stienen and Brenninkmeijer 2002). Since most second-hatched chicks die within 2 weeks post-hatching (Stienen and Brenninkmeijer 2006), a restriction to singletons only led to a relatively small reduction of the entire dataset (in this study, 2.3 % of the data were not used).

The data on singletons ≥15 days old (Table 2) consisted of 1418 measurements of 232 individual chicks of chick body condition (BCI) and parental nest non-attendance (PU). For analyses of daily energy transport (DET), daily energy consumption (DEI) and the proportion consumed (CP = DEI/DET × 100 %), the year 1998 was omitted because in that year we had no reliable estimates of prey length, leaving 1246 measurements of 211 individual chicks for analysis of DET, DEI and CP.

The effects of chick age (AGE), BCI, colony site (KOL entered as a factor variable) and a random effect of INDIVIDU (to control for repeated samples of individuals) on the proportion of time the chick was left unattended at the nest (logit transformed PU), daily energy transport to the chick (DET), daily energy consumption of the chicks (DEI) and the proportion of the energy actually consumed by the chick (CP) were analysed using a generalised linear mixed model (GLMM) in S-PLUS v.6.2.

To examine the combined effects of the nutritional state of the chicks and the adult body condition on the proportion of parental non-attendance, for each year the average body mass of adults caught on the nest in May and June (BMAavg), the averaged condition of chicks (BCIavg) and the averaged parental nest non-attendance (logitPUavg) were entered into a linear regression model. To control for repeated measurements of BCI and logitPU on individual chicks, these parameters were first averaged per individual chick after which these values were averaged per year.

Results

Nutritional state of the chicks

Individual measurement of body condition index (1246 measurements) of the studied Sandwich Tern chicks ranged from −0.70 to 0.35. Average body condition of chicks was slightly higher at Zeebrugge (Fig. 1), but the difference was not significant (BCI = −0.02 ± 0.11 SD and 0.00 ± 0.12 SD, respectively, on Griend and at Zeebrugge; n = 231 chicks, Student’s t test, t = 1.48, P = 0.17). Also, averaged yearly body condition of the chicks (BCIavg) did not differ between the two colonies (−0.03 ± 0.05 and −0.01 ± 0.09 SD, respectively, on Griend and at Zeebrugge; n = 14, Student’s t test, t = 0.5, P = 0.60).

Fig. 1
figure 1

Body condition index BCI (average ± SD, respectively for Griend and Zeebrugge) of Sandwich Tern Thalasseus sandvicensis chicks at Zeebrugge and on Griend in relation to the age (in days) of the chicks (above). Below the number of measurements

Full size image

Parental colony attendance

Parental non-attendance (logitPU) at the colony significantly increased with chick age (AGE), and was negatively correlated with body condition index (BCI) of the chick (Table 3; Fig. 2). Furthermore, we found significant effects of the colony site (KOL) and the interaction term AGE × KOL. On Griend, the probability of being left alone at the colony strongly increased with chick age and with decreasing BCI, whereas at Zeebrugge the effect of age was much smaller (Fig. 2).

Table 3 GLMM coefficient estimates for the effect of chick age (AGE), chick body condition index (BCI) and colony site (KOL: Griend vs. Zeebrugge) on parental nest non-attendance (logitPU; n = 1418 observation days)
Full size table
Fig. 2
figure 2

The effects of chick age (in days) and chick body condition index (BCI) on parental nest non-attendance in the colonies of Griend and Zeebrugge. Values represent averages + SE. Nd no data

Full size image

Averaged yearly parental non-attendance (logitPUavg) was significantly higher on Griend than at Zeebrugge (corresponding PU, respectively, 29.8 % ± 24.5 and 8.4 % ± 10.7 SD; n = 14, Student’s t test, t = 2.9, P = 0.01).

Effects of parental condition

Body mass of individual adult Sandwich Terns (BMA) varied between 195 and 312 g. Averaged yearly BMA (BMAavg) was consistently higher in the Griend colony (Fig. 3) where adults weighed on average 14.5 g more than at Zeebrugge (n = 868, Student’s t test, t = 11.4, P < 0.001). There was no relationship between the averaged yearly BMAavg and the averaged yearly body condition of chicks (BCIavg; Pearson regression, R 2 = 0.005, P = 0.81).

Fig. 3
figure 3

Average body mass of adult Sandwich Terns (BMA ± SE) on Griend (1992–1998) and at Zeebrugge (2001, 2003–2007)

Full size image

On Griend logitPUavg correlated with BCIavg but not with BMAavg, whereas at Zeebrugge neither BCIavg or BMAavg significantly contributed to explain the variation in logitPUavg (Table 4; Fig. 4). When data of both colonies were entered into a single regression analyses, logitPUavg significantly correlated with BMAavg, but no longer with BCIavg (Table 4). Further adding the factor variable colony (KOL) resulted in a significant effect of KOL (P = 0.01), while BMAavg and BCIavg were no longer significant (P = 0.27 and 0.10, respectively).

Table 4 Linear regression analysis testing the combined effect of the averaged yearly adults’ body mass (BMAavg) and the averaged yearly body condition index of the chicks (BCIavg) on averaged yearly parental nest non-attendance (logitPUavg)
Full size table
Fig. 4
figure 4

Relationship between averaged yearly parental nest non-attendance (logitPUavg) and averaged yearly body mass of adults (BMAavg), body condition index of the chicks (BCIavg), food consumption probability, daily energy transport (DETavg) and daily energy intake (DEIavg) at Zeebrugge (ZB; circles) and on Griend (GR; dots). Significant correlations are shown by lines. Note that if the year 1996 is omitted all relationships become stronger and even the relationship between BMAavg and logitPUavg becomes highly significant

Full size image

Energy transport to the colony and energy consumption by the chicks

Daily energy transport to the chick (DET) was significantly higher on Griend than at Zeebrugge (respectively 269.5 ± 98.9 and 239.4 ± 101.7 kJ/day, t = 2.47, n = 211 chicks, P = 0.01), but because prey loss was much higher on Griend (compare CP in Fig. 5) daily energy intake (DEI) was significantly lower than at Zeebrugge (173.2 ± 62.5 and 211.3 ± 89.9 kJ/day, t = 3.52, n = 211 chicks, P < 0.001). Because of the lower colony attendance on Griend, the daily energy transport that was realised per parent (DETpar) did not differ between the two colonies under study (respectively 217.7 ± 69.6 and 230.6 ± 98.0 kJ/day for Griend and Zeebrugge, t = 1.1, df = 209, P = 0.28).

Fig. 5
figure 5

Daily energy transport to the chick, daily energy intake by the chick and the proportion of energy consumed by the chick in relation to chick age and parental nest non-attendance (PU) in the colonies of Griend (left) and Zeebrugge (right). Values represent averages + SE

Full size image

On Griend but not at Zeebrugge, DET increased with chick age (Table 5; Fig. 5). In none of the colonies did increased parental non-attendance result in a significant increase of the energy transport to the chick. On Griend, consumption probability (CP) of the prey significantly correlated with logitPU and with the interaction term logitPU × AGE, with the latter reflecting an increasingly negative effect of parental non-attendance on CP with increasing age of the chick (Fig. 5).

Table 5 GLMM coefficient estimates for the effect of chick age (AGE) and parental nest non-attendance (logitPU) on the daily energy transport (DET), energy intake (DEI) and consumption probability (CP) of chicks on Griend (left panels) and at Zeebrugge (right panels)
Full size table

The positive effect of chick age on DET and CP (although not always significant) resulted in a significant effect of AGE on the daily energy intake (DEI) in both colonies (Table 5; Fig. 5). At Zeebrugge, the interaction term AGE × logitPU also had a significant effect on DEI, showing an increasingly positive effect of parental nest non-attendance on the energy intake with aging of the chick (Fig. 5).

The averaged yearly proportion of energy that was actually consumed by the chick (CPavg) strongly correlated with parental non-attendance (logitPUavg), whereas DETavg was not related to logitPUavg (Fig. 4). Only for Griend was DEIavg significantly, but negatively, correlated with logitPUavg.

Discussion

Importance of chick body condition

The main finding of this study is that, in Sandwich Terns, the proportion of time that chicks were left unattended at the nest negatively correlated with chick body condition. Assuming that time spent outside the colony is used for foraging, this suggests that Sandwich Terns do adjust parental foraging effort to the nutritional status of their chicks. This was expected, as seabirds, and especially short-ranging seabirds (Hamer and Hill 1994; Weimerskirch et al. 1995), must be flexible and adjust foraging effort in response to variable conditions that affect the growth and survival of their offspring in order to be successful (Cairns 1987). Such plasticity in parental investment has already been found in many other short-ranging seabirds (Monaghan et al. 1994; Bolton 1995; Hamer et al. 1998; Tveraa et al. 1998; Wernham and Bryant 1998; Granadeiro et al. 2000; Harding et al. 2002; Ronconi and Burger 2008; Burke and Montevecchi 2009). Only a few studies on short-ranging seabirds have not found a relationship between parental foraging effort and the body condition of their chicks (Ricklefs and Schew 1994; Takahashi et al. 1999).

However, the relationship between body condition and parental non-attendance strongly differed between the two study sites, with Sandwich Tern parents from Griend apparently reacting much more pronounced to a backlog in the development of their chicks than parents at Zeebrugge. Consequently, averaged yearly parental non-attendance on Griend was significantly higher than at Zeebrugge, while the average body condition of the chicks did not differ between the two colonies. On Griend, but not at Zeebrugge, averaged yearly parental non-attendance strongly correlated with the average body condition of the chicks. Our results thus suggest that, within the same species, parents may respond differently to changes in the body condition of their offspring. Studies on other seabirds found that the differences in the degree of parental investment may depend on the build-up of energy reserves of the parents themselves as well as on the body condition of their chicks (e.g. Erikstad et al. 1998; Navarro and González-Solís 2007). When having sufficient body mass reserves and with their own future survival not at risk, seabirds can increase foraging effort in an attempt to counterbalance a backlog in the body condition of their offspring. Weimerskirch et al. (1999, 2001) therefore suggested that parental foraging flexibility in seabirds critically depends on local conditions influencing food abundance, which could explain why some studies found different responses within the same species (Phillips and Croxall 2003; this study). In the closely related Common Tern Sterna hirundo, Wendeln (1997a, b) found that intra-specific differences in adult body mass as well as in colony attendance were strongly related to different foraging strategies. Prior to egg-laying, birds using marine feeding areas were heavier and started laying earlier than limnetic foraging adults (Wendeln 1997b). During incubation, body mass of females that specialised on marine prey strongly decreased (Wendeln 1997a), so that during chick feeding adults at marine sites were lighter, and colony attendance was lower than at freshwater sites (Becker et al. 1997). Also, in this study, we found a pronounced and consistent difference in adult body condition between the two breeding sites with incubating parents on Griend weighing on average 14.5 g more than at Zeebrugge, suggesting strong site-specific differences in food availability. It is, however, not clear whether and how this relates to the strong differences in colony attendance of the parents. If we used data of both colonies, and included chick body condition and adult body mass in a linear regression analysis, parental nest non-attendance did correlate with adult body mass. However, the relationship did not hold when the analyses were performed separately for the two colonies under study. On Griend, there was one year (1996) in which adult body mass was exceptionally low while parental non-attendance was the highest recorded in this study, suggesting that factors other than adult body mass played an even more important role in determining foraging effort.

Additional costs of food robbery

In most years, Sandwich Terns on Griend brought proportionally more sandeels and longer prey items to their chicks than at Zeebrugge (Table 1). Although longer prey are more susceptible to robbing, the difference in prey length cannot explain the large difference in the proportion of energy that was actually consumed by the chicks. Given the average prey lengths as presented in Table 1, and using the relationship between prey length and consumption probability as found by Stienen et al. (2001), average consumption probability would amount to 0.70 ± 0.03 on Griend (close to the values shown in Fig. 5) and 0.72 ± 0.07 at Zeebrugge (much lower than in Fig. 5). We instead believe that the higher consumption probability at Zeebrugge was due to the fact that this colony held much lower numbers of Black-headed Gulls (Table 1). During incubation and the onset of chick-rearing, the gulls protect the terns against predators (Veen 1977), but, with the aging of the tern chicks, a number of gulls develop into specialised parasites of fish meant for the tern chicks (Stienen et al. 2001). The lower gull density and the fact that the tern colony at Zeebrugge was never located within the gull colony as was always the case on Griend, but instead at the edge, make it less likely that neighbouring gulls would develop into specialised pirates. On Griend, on average more than three Black-headed Gulls tried to rob a fish from an arriving Sandwich Tern parent (Stienen and Brenninkmeijer 1999), while at Zeebrugge normally none or only one pirate was active (own observations). Despite the fact that daily energy transport to the chicks on Griend was higher, the low energy consumption probability even resulted in a lower energy intake rate than at Zeebrugge.

The strong relationship between parental nest non-attendance and energy consumption probability suggests that Sandwich Tern parents react to high rates of food loss to kleptoparasites by increasing the time spent foraging and by leaving the chick unattended at the colony more often. When a prey is successfully delivered to the chick, brooding and foraging duties are normally changed between the partners (own observations). When the prey was stolen, however, the parent that was initially carrying the prey normally returned to the foraging area, and when its subsequent return took too long, the attending parent often left the colony to forage (own observations). The low nest attendance on Griend might thus merely be a response to high rates of food loss which in turn causes a decrease in chick body condition.

Effects of adult body condition

Since we found a significant relationship between adult body mass and energy consumption probability, it cannot be excluded that the relationship between parental nest non-attendance and energy consumption probability reflects a difference in the degree of parental foraging investment due to variation in the build-up of energy reserves of the parents. Here, one must bear in mind that the correlation between parental nest attendance and adult body mass was much weaker than the correlation with consumption probability (Fig. 4), mainly because the data from Griend in 1996 (relatively low adult body mass and high parental nest non-attendance) did not fit very well. Nevertheless, one can think of several plausible explanations for a direct effect of adult body mass on parental nest non-attendance and a possible link with food consumption probability. Firstly, this can be expected when parents have already incorporated the expected additional costs for rearing young (i.e. expected increase in foraging effort due to food loss to kleptoparasites) before hatching of the eggs by investing in extra body mass (Sénéchal et al. 2011), and by assuming that parents on Griend in 1996 did not correctly predict the robbing pressure. The rate of kleptoparasitism experienced during courtship feeding or the number of Black-headed Gulls present in the colony might, for example, be used as a proxy for the robbing pressure expected during chick rearing. Also, differences in the age composition or experience of the adults may explain the difference in provisioning rates and parental plasticity between Griend and Zeebrugge as was found in Common Terns Sterna hirundo (Limmer and Becker 2009).

Another explanation would be that parents use their own body mass to evaluate future fitness costs and adjust parental foraging effort accordingly (cf. Navarro and González-Solís 2007). If at the same time high non-attendance at the colony evokes high rates of food parasitism, a high parental body condition would eventually result in low food consumption probabilities of their offspring. The latter is plausible, since prey delivery seems easier when one of the parents attends the chick (own observations). When a prey-carrying parent arrives at the colony, the attending parent normally repositions the chick to an open spot where the prey-carrying parent can land easily. After landing, the chick is normally positioned inbetween the two parents who form a protective shield around the chick with their open wings. When no parent is attending the chick, not only will this protective behaviour be absent but in addition neighbouring adults often do not allow an unattended chick to wait for the arriving parent at an open spot, making successful delivery of the prey less easy.

We therefore conclude that our study strongly suggests that parental nest non-attendance in Sandwich Terns is basically a response to compensate for the food loss to kleptoparasitic gulls in order to meet the energy requirements for the growth of the chicks, but that an additional role of adult body condition cannot be excluded.