Oecologia DOI 10.1007/s00442-015-3334-2
HIGHLIGHTED STUDENT RESEARCH
Diet specialization in a generalist population: the case of breeding great tits Parus major in the Mediterranean area E. Pagani‑Núñez1,2 · M. Valls1 · J. C. Senar1
Received: 16 May 2014 / Accepted: 26 April 2015 © Springer-Verlag Berlin Heidelberg 2015
Abstract The analysis of diet specialization provides key information on how different individuals deal with similar food and habitat constraints within populations. Characterizing parental diet specialization at the moment of breeding, and the consistency of these preferences under different levels of effort, may help us to understand why parents exploit alternative resources. We investigated these questions in a species commonly considered a generalist: a breeding population of Mediterranean great tits Parus major. Our aim was to determine whether they are specialists or generalists at the pair level, and the consistency of this behaviour under different levels of effort. Using proportional similarity and mean pairwise overlap indices, we found that parents showed great variability in prey selection between territories. That is, they displayed a small niche overlap. Interestingly, the most specialized breeding pairs showed a tendency to have larger broods. Additionally, we experimentally manipulated brood size and found that parents showed high short-term consistency in their foraging behaviour. They precisely adjusted the number of Communicated by Ola Olsson. Electronic supplementary material The online version of this article (doi:10.1007/s00442-015-3334-2) contains supplementary material, which is available to authorized users. * E. Pagani‑Núñez [email protected] 1
Evolutionary Ecology Associate Research Unit (CSIC), Natural History Museum of Barcelona, Psg. Picasso s/n., 08003 Barcelona, Spain
2
Behavioral and Community Ecology, Conservation Biology Group, College of Forestry, Guangxi University, No. 100 Daxue Road, Nanning 530005, Guangxi, People’s Republic of China
provisioning trips to the number of nestlings, while they were unable to modify prey proportions or prey size after brood size was changed. We can therefore characterize their foraging strategies as highly consistent. Our results suggest that although the great tit may be considered a generalist at the species or population level, there was a tendency for trophic specialization among breeding pairs. This high inter- and intrapopulation plasticity could account for their great success and wide distribution. Keywords Foraging behaviour · Niche overlap · Niche partitioning · Prey selection · Provisioning rates
Introduction Studying parental foraging behaviour expands our knowledge of the breeding ecology of wild animals (Stephens et al. 2007). At the species level, parents may display a great variety of foraging strategies as a result of divergent patterns of energy investment and prey selection (Drent and Daan 1980; Cuthill and Houston 1997), and in relation to prey availability. This has been shown in diverse animal taxa such as arthropods (e.g. bees, Visscher and Seeley 1982; Schneider and McNally 1993) and mammals (e.g. primates, Pienkowski et al. 1998; Ramos-Fernández et al. 2004). This is also the case for passerine birds, for which provisioning patterns at breeding have been focus of intense research. Several studies have reported highly variable daily provisioning rates (see Pagani-Núñez and Senar 2013 and references therein). These strikingly variable patterns have usually been attributed to prey availability or dynamics. However, in spite of many examples of experimental approaches investigating the consistency of parental foraging behaviour under different temporal constraints or
13
Oecologia
levels of effort, especially in passerine birds (e.g. Tinbergen and Verhulst 2000; Nilsson 2002; Nicolaus et al. 2009, 2015), intrapopulation variability and the plasticity of individual prey selection have rarely been examined. Many studies have analyzed parental responses to changes in brood size (and/or begging intensity) and their consequences for nestling body condition. Most such studies have relied on provisioning rates to evaluate parental foraging behaviour (Smith et al. 1988; Saino et al. 1997; Verhulst and Tinbergen 1997; Kölliker et al. 2000; Tinbergen and Verhulst 2000; Nilsson 2002; Ardia 2007) and have not considered prey selection by parents. Studies considering prey quality have reported almost constant provisioning rates per nestling and prey size irrespective of brood size manipulation (Neuenschwander et al. 2003), or a tradeoff between provisioning rates per brood and prey size. In the latter case, parents often delivered larger prey items at lower provisioning rates (Stoehr et al. 2001; García-Navas and Sanz 2010), a pattern which is sometimes restricted to fathers (Siikamäki et al. 1998, but see also Wright and Cuthill 1990). Nonetheless, different prey types can show different nutritive values (Arnold et al. 2010; Wiesenborn 2012), so parents may also compensate for a decrease or increase in their provisioning rate by adjusting prey proportions rather than prey sizes (Magrath et al. 2004). Therefore, it is necessary to assess intrapopulation variability and the plasticity of individual prey selection in order to fully understand parental feeding behaviour and foraging strategies. For instance, previous research highlighted a link between different foraging strategies and productivity in herbivorous mammals (White 1983; Spalinger and Hobbs 1992), which display less sophisticated foraging skills than insectivorous birds. The high variability of parental foraging strategies within and among populations of many species suggests that the optimal foraging strategy may depend on the environmental conditions present (Stephens et al. 2007; Naef-Daenzer 2012). Parents would also modify their foraging strategies within populations, adjusting prey proportions and prey sizes to changing environmental conditions or levels of effort. Animals exploit diverse food resources by implementing different foraging strategies. Sometimes there are several profitable food types, so different species specialize in different food types according to their foraging and breeding strategies (Heithaus et al. 1975). On the other hand, generalist and specialist species may show divergent distribution patterns and population trends depending on how they exploit their habitat (Seamon and Adler 1996), and different populations of the same species may display either a generalist or a specialist strategy depending on the habitat (Quevedo et al. 2009). There is a lack of research, however, on the variability of these foraging strategies and their consequences for fitness within populations. Breeding
13
passerines use a wide variety of prey as food sources (e.g. Barba and Gil-Delgado 1990; Grundel 1990; Blondel et al. 1991; Banˇbura et al. 1994; Gilroy et al. 2009; Maziarz and Wesołowski 2010). However, in most cases it is also unclear whether parents may be considered generalists or whether different pairs specialize in different prey types within this wide range of variation (Estes et al. 2003; Bolnick et al. 2003; Woo et al. 2008; Newsome et al. 2009). In the Mediterranean area, bird populations rely more often on alternative prey than populations of the same species inhabiting northern latitudes. This variability in foraging strategies fits with the diverse and variable prey distribution that is characteristic of the Mediterranean area (Banˇbura et al. 1994; Blondel et al. 2010). In the case of the great tit Parus major, we recently reported that a population inhabiting a Mediterranean mixed forest presented high consistency in parental effort and prey selection over time (PaganiNúñez and Senar 2013) and consumed a wide variety of prey (Pagani-Núñez et al. 2011; Pagani-Núñez and Senar 2014). The great tit shows a wide geographical distribution across Europe and Asia (see e.g. Kvist et al. 2007), and it is possible to find viable populations in a wide variety of habitats (forest, agricultural and urban landscapes). Since different habitats may require different breeding and/or foraging strategies, characterizing the intrapopulation variability of parental foraging behaviour could represent a reliable approach to explaining and understanding animal distribution patterns (Naef-Daenzer 2012). In this study, our main objective was to estimate individual diet specialization with respect to the mean diet of the population in order to assess the variability in foraging strategies among breeding pairs. This allowed us to characterize each pair as more or less generalist or specialist. To experimentally evaluate the consistency of these strategies, we also evaluated the consistency of parental foraging behaviour by means of a brood size manipulation experiment. We hypothesized that, regardless of prey availability (and without underestimating its relevance in shaping population patterns), parents with a pure generalist foraging strategy should respond to brood size manipulation by adapting prey proportions or prey size to an increase or decrease in their provisioning rate (Grieco 2001; Stoehr et al. 2001; García-Navas and Sanz 2010). In reduced broods, we could expect an increase in provisioning rates per nestling and/or in the delivery of high-quality prey such as caterpillars and spiders, while we could expect the opposite response in increased broods. Conversely, parents with a specialist foraging strategy should not respond to our experimental manipulation (in relation to prey quality) because specialization would imply a loss—to some extent—of ability to shift individual dietary preferences (Trowbridge 1991; Nosil 2002), or may result in non-overlapping foraging niches between individuals (Bolnick et al.
Oecologia
2003; Newsome et al. 2009; Vander Zanden et al. 2010; Quiroga et al. 2011). Finally, we ascertained how intra- and interpopulation variability in parental foraging strategies could help to explain the distribution patterns of great tits.
Materials and methods Fieldwork and diet recording During the spring of 2012, we monitored 182 nest boxes at Can Catà field station (45°27′N, 2°8′E; Catalonia, northeast Spain), a predominantly evergreen mixed forest. Nest boxes were checked twice weekly to determine laying and hatching dates and brood size. Habitat structure showed a marked gradient of variation in relation to tree composition. The proportion of oaks (Quercus spp.) ranged from 5 to 95 % and correlated negatively with the altitude above sea level (82–219 m above sea level; see Pagani-Núñez et al. 2011, 2014). The main tree species were Aleppo pine Pinus halepensis, holm oak Quercus ilex and the oak Q. cerrioides. Nestling diet was successfully recorded in 31 nests by means of Micro-D cameras attached to the covers of the nest boxes and focused on the entrance (see PaganiNúñez and Senar 2014 for more details on camera properties and filming procedures). We filmed for 48 h in each nest when nestling age was 10–14 days (Pagani-Núñez and Senar 2013). Cameras were installed and camouflaged the day before recording to minimize the impact of installing the device. Recording was performed for five hours each day, from 7 a.m. to 12 a.m. We used full clock hours to standardize for interindividual differences in the time at which recording began. The sex of each parent was easily identified from the recordings. In addition, most individuals had been previously marked with numbered PVC rings. We determined prey type and size for each feeding action of each parent within this time window. Prey size was determined according to a semi-quantitative scale with three values: 1 = small; 2 = medium; 3 = large (Barba et al. 1996). In this work, we define brood size as the number of nestlings in the nest at the time of recording. Diet overlap and trophic specialization We assessed between-pair trophic specialization in relation to population niche breadth by means of the proportional similarity index, PSi = 1 − 0.5∑j │pij − qij│ = ∑j min(pij, qij), and the mean pairwise overlap (see Bolnick et al. 2002 for more details). The former is a measure of the overlap between the diets of an individual and the population; the latter is the mean PSi between all pairwise combinations of individuals, or—in our case—pairs (Bolnick et al. 2002). Both indices were obtained through IndSpec, version 4.0
(Bolnick et al. 2002). These indices provide information on the use of trophic resources by individuals properly framed in the context of a concrete population, and range from 0 to 1, where 1 indicates that the individual’s diet is equal to the mean diet of the population and lower values indicate that the individual specializes with respect to the population mean. To perform this analysis, we used nestling diet proportions on control days (5 h of recordings from 7 a.m.; see Pagani-Núñez and Senar 2013), which represented the normal conditions for parents. We used mean values for each pair instead of individual data because our aim was to assess the diet overlap between territories (that is, between pairs or nests). In addition, we standardized prey proportions to the mean size of each kind of prey in each nest (following Barba et al. 1996, as previously described): we multiplied mean prey size by the number of prey items and calculated new percentages on this basis. We divided prey into fourteen classes, which can be listed as follows according to their functional groups and their importance in the nestling diet: caterpillars (Lepidoptera larvae), chrysalises (Lepidoptera pupae), moths and butterflies (Lepidoptera adults), other larvae, spiders (Araneae), spiders’ eggs, grasshoppers (Orthoptera), stick insects (Phasmida), mosquitoes (Diptera), bees and flying ants (Hymenoptera), beetles (Coleoptera), unidentified arthropods, fruits and formless remains. We used the most detailed classification to investigate diet specialization and how parents combine different prey classes, but we simplified it to facilitate the assessment of parental responses to our experimental manipulation. We recorded a total of 2362 feedings. The mean number ± SE of total feedings per pair across 5 h of recording was 76.19 ± 7.00, while the mean number of feeding trips per nestling was 15.81 ± 1.41. These analyses included a total of 62 parents from 31 nests. Brood size manipulation experiment In addition to the degree of diet specialization among pairs, we assessed the consistency of their foraging behaviour under different levels of effort. We performed a brood size manipulation experiment in 18 nests when nestlings were 10–14 days old. Our sample consisted of 36 parents and 94 nestlings (a mean of 5.22 nestlings per nest), 18 of which were moved between nests to experimentally modify brood size. We used pairs of nests with the closest hatching dates (up to 2 days) and with brood sizes that were the same or similar, thus decreasing the number of available nests from the initial 31 which we successfully filmed. That is, some nests remained unpaired due to these limitations and were not used in the experiment. Although there was great interpair variability, we collected our sample across the whole breeding season (we expect that this minimized the impact of reducing the initial sample). We randomly selected and
13
Oecologia
moved two nestlings from one nest to the other for 24 h, so we obtained a simultaneous experimental increase and decrease in brood size in the two complementary nests (“reduced” = −2 or “increased” = +2). For instance, a pair of nests with five nestlings each were adjusted to have three nestlings in one and seven in the other. The experiment was restricted to two nestlings to keep brood size within the natural range of variation for this species (Orell et al. 1996). If the nests did not have the same brood size, we reduced the larger nest and increased the smaller nest of the experimental pair. We recorded the “normal” and the “experimental” day for each nest. Thus, the control for each pair was their own “normal” foraging behaviour. Moreover, we performed the experiment before or after recording the normal day in successive pairs of nests. After the recording, we placed both of the moved nestlings back in their original nests. Provisioning rates per hour and provisioning rates per nestling, prey proportions and mean prey size were determined for each individual. In this case, prey were classified as caterpillars, spiders or others in order to focus on the most relevant prey types, following our previous procedures (Pagani-Núñez and Senar 2014; PaganiNúñez et al. 2014). Statistical analysis Once we had characterized the trophic structure of our study population, we aimed to identify which extrinsic and/or intrinsic factors determined this structure. We used an Akaike’s information criterion (AIC) approach to assess which intrinsic (brood size and foraging strategy) and extrinsic (date and habitat structure) factors were more relevant in shaping PSi values. The AIC provides information on how well a model based on a specific variable (or set of variables) fits to a particular dataset; it is a measure of the uncertainty (or entropy) in the data which the model does not account for (Burnham and Anderson 2002). The smaller the value of the AIC, the better the fit of the model to the data. Models that differ by less than two in their AIC scores (ΔAIC ≤2) fit the data similarly well. Moreover, we computed the AIC weights (which can be interpreted as conditional probabilities) for each model (Wagenmakers and Farrell 2004). Our analysis was based on a normal distribution with a log link function. The dependent variable was PSi value (after being arcsine square-root transformed), and independent variables included the number of prey types in the nestling diet, brood size (number of nestlings during recording on control days), date (number of days from 1st April), proportion of the trees surrounding the nest boxes that were Quercus spp. (to control for habitat quality, see Pagani-Núñez and Senar 2014), provisioning rate per nestling, and mean prey size. We included variables
13
that characterize parental foraging strategies (provisioning rates and prey size) and brood size as intrinsic factors to ascertain the energetic costs and the consequences for fitness of different foraging strategies. The number of prey types selected by each pair helped us to quantitatively interpret the significance of our PSi values; in other words, whether being more specialist directly implied the selection of fewer prey types. We included the date of breeding and the proportion of the trees surrounding the nest boxes that were Quercus spp. as extrinsic factors, since these features strongly affect prey composition (see e.g. Pagani-Núñez and Senar 2014). Moreover, if these environmental features are correlated with PSi values, it could indicate a strong effect of prey availability on parental foraging strategies. Breeding date and habitat structure may determine brood size (see e.g. Perrins and McCleery 1989; Lambrechts et al. 2004), so we assessed the relationships among these variables beforehand. We found that pairs inhabiting territories with lower proportions of Quercus spp. showed a tendency to raise larger brood sizes (n = 31, r = −0.32, t = −1.80, p = 0.08) and to breed earlier in the season (n = 31, r = 0.32, t = −1.91, p = 0.07). Brood size did not correlate with date (n = 31, r = −0.07, t = −0.40, p = 0.69). There was considerable time overlap among nests (see Appendix A1). In order to investigate which prey types were most commonly captured together, we used a principal components and classification analysis (PCCA). Using a batch of variables, PCCA performs a principal component analysis and then shows how these variables are interrelated. We used the numbers of each prey type in each nest as variables for analysis, but we excluded the four least common prey types (Hymenoptera, Diptera, chrysalises and unidentified insects; each of these represent around 1 % of the nestling diet) to increase the statistical power of the PCCA. This approach allowed us to probe how and why parents combine several food sources. Based on our previous research, we would not expect caterpillars and spiders to be associated with each other, or with other relevant prey types such as grasshoppers or moths. Caterpillars and spiders have been described as key food sources for many bird species, and this finding would justify considering three functional food sources (caterpillars, spiders and others) when investigating diet composition. Moreover, the approach helped us to ascertain whether parents combine their prey according to availability patterns, or whether they select prey based on their characteristics (e.g. if they simultaneously focus on these similar prey types when foraging). Finally, we analyzed the consistency of parental foraging behaviour within our population. To analyze parental responses to our experiment, we used a repeated measures analysis of variance approach. We used a control and an experimental day for each pair; that is, the control for
Oecologia
each pair was their own natural behaviour. The provisioning rate per brood in the control and the experimental day were our dependent variables. Our within-subject factor was the comparison between these variables before and after the experiment, which we referred to as “time”, while the experimental group (experiment: “reduced” or “increased” if the dependent variable differed between experimental groups) and the sex (male or female if the sexes differed a priori) were the between-subject factors. We assessed their interactions to ascertain if the parents responded to our manipulation (time × experiment) and if there was a sex-related response (sex × experiment and time × sex × experiment) or some bias in our experimental procedure. We did the same for each of the following dependent variables: provisioning rate per nestling; proportion of caterpillars, spiders and others; mean prey size. A chi-square normality test showed that provisioning rate per brood [control day: χ2= 2.81, df = 4 (adj.), p = 0.59; experimental day: χ2 = 4.10, df = 3 (adj.), p = 0.25], provisioning rate per nestling [control day: χ2 = 2.41, df = 2 (adj.), p = 0.30; experimental day: χ2 = 2.61, df = 2 (adj.), p = 0.27], and mean prey size [control day: χ2 = 3.25, df = 3 (adj.), p = 0.35; experimental day: χ2 = 4.30, df = 3 (adj.), p = 0.23] were normally distributed. Moreover, prey proportions were arcsine square-root transformed. We used Statistica 6.0 to perform our analyses (StatSoft 2001), except for the PSi values, which were computed with R (R Development Core Team 2014).
Results Caterpillars were the main prey type, representing half of all food (55 %, Appendix A2). Spiders were the second most common prey type (14 %, Appendix A2), and the proportions of the food represented by the remaining prey types ranged from 1 to 7 % (Appendix A2). However, nestling diet was extremely variable among territories. For instance, caterpillars ranged from 16 to 93 % of the food, and not all prey types were present in all nests (Appendix A2). That is, different pairs selected different combinations of prey types. The similarity index score (SI, population mean of PSi) was 0.72, with PSi values for pairs ranging from 0.53 to 0.84 (Fig. 1). The mean pairwise overlap was 0.63. In certain cases, pairs of nests which overlapped in time and space displayed dissimilar PSi values (Appendix A1). Our AIC approach showed that the best model explaining the variability in PSi values included only brood size (Table 1), although the top seven models fitted the data similarly well (ΔAIC ≤2, Table 1). In the top model, the confidence interval of the coefficient included zero (lower CL 95 %: −0.05, upper CL 95 %: 0.001). Indeed, PSi values were negatively related to brood size (Fig. 2), suggesting that breeding
Fig. 1 Histogram showing the distribution of proportional similarity index values within our study population. Data correspond to 31 pairs of Mediterranean great tits Parus major and are based on natural brood sizes
pairs with more specialized diets than the overall population more frequently had larger broods. The second model included brood size and the proportion of the trees surrounding nest boxes that were Quercus spp. (Table 1), and the third included the proportion of the trees surrounding nest boxes that were Quercus spp. (Table 1). Models 4–7 included brood size and, respectively, provisioning rate per nestling (Table 1), mean prey size (Table 1), date (Table 1) and number of prey classes (Table 1). The first two axes of the principal component and classification analysis explained the most variance: 50.73 % of the variation in nestling diet (Fig. 3). The first component showed positive associations among the majority of prey types (Table 2). Stick insects were collected independently of those types, while moths and butterflies were selected in opposition to that majority of prey types (Table 2). The second component showed positive associations among caterpillars, fruits and other larvae (Table 2). On the other hand, spiders and their eggs seemed to be collected independently (Table 2). Finally, grasshoppers, stick insects, moths and butterflies, beetles and formless remains were collected in opposition to the first group (Table 2). The third component provided further confirmation that stick insects were often collected independently (Table 2), while the fourth and fifth components showed secondary associations between moths and butterflies and other larvae, and between caterpillars and grasshoppers in opposition to stick insects (Table 2). Thus, there were three main assemblages among the prey types, although two of them (stick insects and moths and butterflies) could be segregated into a fourth group and showed secondary associations (Fig. 3). Our repeated measures approach showed that parents reduced the number of feeding trips after we reduced brood
13
Oecologia Table 1 Results of our Akaike’s information criterion (AIC)-based method of assessing which intrinsic (brood size and foraging strategy) and extrinsic (date and habitat structure) factors were more relModel no.
Variable 1
Variable 2
1 2
Brood size Brood size
evant in shaping proportional similarity (an index of diet specialization) values among 31 breeding pairs of great tits Parus major df
AIC
ΔAIC
wi(ΔAIC)
% Quercus spp.
1 2
−55.80 −54.73
0.00 1.06
0.23 0.13
Provisioning rate per nestling Mean prey size Date Brood size Mean prey size
2 2 2 2 2
3
% Quercus spp.
4 5 6 7 8
Brood size Brood size Brood size No. of prey types % Quercus spp.
9
% Quercus spp.
Provisioning rate per nestling
10
Brood size
Date
Variable 3
1
2 % Quercus spp.
3
−54.59
−53.83 −53.82 −53.81 −53.80 −53.31
−52.96
−52.96
1.21
0.12
1.97 1.98 1.99 2.00 2.49
0.09 0.08 0.08 0.08 0.07
2.83
0.06
2.84
0.06
We included variables that characterize parental foraging strategies (provisioning rate per nestling and mean prey size) and brood size as intrinsic factors. We included the date of breeding, a measure of habitat structure (% of trees surrounding nest boxes that were Quercus spp.) and the number of prey types as extrinsic factors. We show the top ten models in order of statistical power
Fig. 2 Relationships between proportional similarity index (y-axis) and natural brood size (x-axis) during recordings of 31 pairs of great tits Parus major
size, and vice versa (reduced broods: from 37.95 ± 5.51 to 27.35 ± 2.90 feeding trips vs. increased broods: from 34.12 ± 5.57 to 48.62 ± 6.59 feeding trips; Table 3; Fig. 4). Experimental groups did not differ a priori in provisioning rate (Table 3). Provisioning rate per brood did not differ between the sexes and did not vary in relation to sex after our experimental procedure (all p > 0.24; Table 3). Additionally, provisioning rate per nestling did not differ a priori between experimental groups, nor between the sexes, and remained constant after we modified brood size (reduced broods: from 6.90 ± 5.91 to 7.54 ± 4.42 feeding trips per nestling vs. increased broods: from 7.62 ± 4.50 to 7.64 ± 3.96 feeding trips per nestling, all p > 0.24; Table 3). The proportion of caterpillars differed between experimental groups prior to experimental procedures
13
Fig. 3 Relationships among the ten most important prey types delivered to nestlings by 31 Mediterranean great tit Parus major pairs in natural conditions based on principal component and classification analysis. Prey types that are closer on the diagram were combined across the whole study period more often, while prey types that are farther apart were combined less often. Prey types close to the centre of the diagram were delivered independently
(reduced broods: 37.95 ± 5.51 % vs. increased broods: 52.37 ± 0.06 %; Table 4). The proportion of caterpillars delivered by parents did not differ between the sexes and remained constant after our experimental manipulation of brood size (all p > 0.18; Table 4). Females delivered more spiders than males (females: 18.80 % ± 0.03 vs. males: 12.03 % ± 0.02; Table 4), but the proportion of spiders
Oecologia Table 2 Results of the principal component and classification analysis of the ten most common prey classes within our population Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 % Var explained 27.55 No. caterpillars 0.50 No. spiders 0.79 No. fruits 0.38 No. grasshoppers 0.46 No. stick insects −0.01 No. moths and but- −0.43 terflies
23.17 0.46 0.01 0.74 −0.62 −0.34 −0.17
13.50 0.30 −0.17 0.09 −0.09 0.74 −0.63
10.09 0.27 0.03 −0.38 −0.18 −0.21 −0.46
8.29 −0.46 0.24 −0.07 −0.47 0.42 0.00
No. other larvae No. beetles No. spider eggs
0.39 0.28 0.75
0.09 −0.19 0.29
0.74
−0.02 0.37 −0.21
−0.55 −0.37 0.26
No. formless remains
0.62 −0.65 −0.03
0.00
0.12
−0.47
−0.30
We used the number of prey of each type delivered by 31 pairs of great tits Parus major to their nestlings (during 5 h of recording) as variables for analysis, to show how parents combined the various food sources
Table 3 Repeated measures analysis of variance using provisioning rate per brood or nestling for 19 pairs of great tit Parus major parents as the dependent variable, experimental group (reduced or increased) as the within-subject factor, and sex (male or female) as the betweensubject factor
Provisioning rate per brood Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment Provisioning rate per nestling Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment
SS
F1,32
p level
1159.21 1353.34 67.60 127.21 0.54 2800.04 20.54
1.43 1.68 0.39 0.16 0.16; Table 4). Finally, mean prey size did not differ a priori between experimental groups, nor between the sexes, and also remained constant after we modified brood size (all p > 0.09; Table 4).
Discussion Our findings support the idea that Mediterranean great tits can be considered diet generalists, while we recorded a tendency for diet specialization among breeding pairs. Interestingly, pairs raising larger broods showed more specialized diets, and diet specialization did not correlate with habitat structure or breeding date. However, pairs showed a tendency to raise larger brood sizes in territories with lower proportions of Quercus spp., and they occupied territories with lower proportions of Quercus spp. earlier in the season. These correlations with environmental features may help us to interpret our results. Previous research has stressed the key roles of habitat structure (Schneider and McNally 1993; Pienkowski et al. 1998) and the timing of breeding (Beja 1996; Eide et al. 2004; Rutz and Bijlsma 2006) in shaping prey availability and breeding performance of wild animals. This pattern is particularly marked in insectivorous birds due to their dependence on insect prey, which vary in their peaks of abundance (NaefDaenzer et al. 2000; Tremblay et al. 2003; Blondel 2007; Eeva et al. 2009). The great tit and its close relative the blue tit Cyanistes caeruleus (hereinafter “tits”), though usually considered diet generalists, mainly rely on caterpillars to feed their offspring (Naef-Daenzer et al. 2000; Tremblay et al. 2003; Blondel 2007; Eeva et al. 2009). Accordingly, it has been suggested that there is a lack of fit between their life-history strategies and environmental conditions at both extremes of their distribution ranges (Sanz 1998; Rytkonen and Orell 2001), and especially within evergreen forests (Sanz 1998; Fargallo 2004; Lambrechts et al. 2004). As has been shown, we could expect suboptimal conditions early or late in the season (before or after the peak in caterpillar abundance) and in territories with a higher proportion of Aleppo pines (where food is usually scarce). In this regard, the lack of alternative, profitable prey types to caterpillars strongly constrains tit breeding performance (Rytkonen and Krams 2003), while the presence of such alternative prey types is what guarantees the viability of southern marginal populations (Blondel et al. 1991; Banˇbura et al. 1994; Ziane et al. 2006; Pagani-Nuñez et al. 2011). In line with this idea, our findings hint that diet specialization could operate as the main mechanism through which great tits avoid these habitat and temporal constraints, allowing them
13
Oecologia Table 4 Repeated measures analysis of variance using the proportion of caterpillars or spiders or other prey delivered by 19 pairs of great tits Parus major to their nestlings or the mean prey size as the dependent variable, experimental group (reduced or increased) as the within-subject factor, and sex (male or female) as the between-subject factor
Fig. 4 Provisioning rates per brood in natural conditions and after the experimental procedure of increasing or decreasing the number of nestlings in the nests of 18 pairs of Mediterranean great tits Parus major. Filled circles correspond to reduced broods while empty squares correspond to increased broods
to increase or maintain their numbers under suboptimal conditions. In other words, in spite of the limiting effect of habitat structure and date on brood size, which is particularly evident in the Mediterranean area, more specialized pairs kept broods of comparable sizes to central European populations. On the other hand, increased interpair variability would be expected if parents raising larger broods have less time to search for food and have to exploit narrower foraging niches. Moreover, within species, niche partitioning may be promoted by spatial and temporal segregation (Cherel et al. 2009), and individuals may specialize in the exploitation of certain habitats (Paterson et al. 2012; Rousseau et al. 2012). In any case, great tit parents displayed a gradient from generalist to specialist foraging strategies (a pattern also found in other systems, e.g. Cherel et al. 2009), which was mainly determined by parental investment in reproduction. This dynamic pattern of diet specialization fits the diverse and variable prey distribution that is characteristic of the Mediterranean area (Blondel et al. 2010), highlighting that diversity of prey may also promote diversity of foraging strategies within species and populations (Sinclair et al. 2003; Duffy et al. 2007). The lack of plasticity of parental foraging behaviour in the short term may hint at why the relationship between brood size and foraging niche is so relevant. Several studies performed in different environments and populations of great and blue tits have reported contrasting results in relation to the capacity of parents to adjust their investment in reproduction to environmental conditions (Sanz 1998; Rytkonen and Orell 2001; Tremblay et al. 2003; Mägi et al. 2009; Atiénzar et al. 2010). We found that great tit parents, irrespectively of sex, rapidly modified their behaviour
13
% Caterpillars Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment % Spiders Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment % Others Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment Mean prey size Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment
SS
F1,32
p level
0.16 0.36
HIGHLIGHTED STUDENT RESEARCH
Diet specialization in a generalist population: the case of breeding great tits Parus major in the Mediterranean area E. Pagani‑Núñez1,2 · M. Valls1 · J. C. Senar1
Received: 16 May 2014 / Accepted: 26 April 2015 © Springer-Verlag Berlin Heidelberg 2015
Abstract The analysis of diet specialization provides key information on how different individuals deal with similar food and habitat constraints within populations. Characterizing parental diet specialization at the moment of breeding, and the consistency of these preferences under different levels of effort, may help us to understand why parents exploit alternative resources. We investigated these questions in a species commonly considered a generalist: a breeding population of Mediterranean great tits Parus major. Our aim was to determine whether they are specialists or generalists at the pair level, and the consistency of this behaviour under different levels of effort. Using proportional similarity and mean pairwise overlap indices, we found that parents showed great variability in prey selection between territories. That is, they displayed a small niche overlap. Interestingly, the most specialized breeding pairs showed a tendency to have larger broods. Additionally, we experimentally manipulated brood size and found that parents showed high short-term consistency in their foraging behaviour. They precisely adjusted the number of Communicated by Ola Olsson. Electronic supplementary material The online version of this article (doi:10.1007/s00442-015-3334-2) contains supplementary material, which is available to authorized users. * E. Pagani‑Núñez [email protected] 1
Evolutionary Ecology Associate Research Unit (CSIC), Natural History Museum of Barcelona, Psg. Picasso s/n., 08003 Barcelona, Spain
2
Behavioral and Community Ecology, Conservation Biology Group, College of Forestry, Guangxi University, No. 100 Daxue Road, Nanning 530005, Guangxi, People’s Republic of China
provisioning trips to the number of nestlings, while they were unable to modify prey proportions or prey size after brood size was changed. We can therefore characterize their foraging strategies as highly consistent. Our results suggest that although the great tit may be considered a generalist at the species or population level, there was a tendency for trophic specialization among breeding pairs. This high inter- and intrapopulation plasticity could account for their great success and wide distribution. Keywords Foraging behaviour · Niche overlap · Niche partitioning · Prey selection · Provisioning rates
Introduction Studying parental foraging behaviour expands our knowledge of the breeding ecology of wild animals (Stephens et al. 2007). At the species level, parents may display a great variety of foraging strategies as a result of divergent patterns of energy investment and prey selection (Drent and Daan 1980; Cuthill and Houston 1997), and in relation to prey availability. This has been shown in diverse animal taxa such as arthropods (e.g. bees, Visscher and Seeley 1982; Schneider and McNally 1993) and mammals (e.g. primates, Pienkowski et al. 1998; Ramos-Fernández et al. 2004). This is also the case for passerine birds, for which provisioning patterns at breeding have been focus of intense research. Several studies have reported highly variable daily provisioning rates (see Pagani-Núñez and Senar 2013 and references therein). These strikingly variable patterns have usually been attributed to prey availability or dynamics. However, in spite of many examples of experimental approaches investigating the consistency of parental foraging behaviour under different temporal constraints or
13
Oecologia
levels of effort, especially in passerine birds (e.g. Tinbergen and Verhulst 2000; Nilsson 2002; Nicolaus et al. 2009, 2015), intrapopulation variability and the plasticity of individual prey selection have rarely been examined. Many studies have analyzed parental responses to changes in brood size (and/or begging intensity) and their consequences for nestling body condition. Most such studies have relied on provisioning rates to evaluate parental foraging behaviour (Smith et al. 1988; Saino et al. 1997; Verhulst and Tinbergen 1997; Kölliker et al. 2000; Tinbergen and Verhulst 2000; Nilsson 2002; Ardia 2007) and have not considered prey selection by parents. Studies considering prey quality have reported almost constant provisioning rates per nestling and prey size irrespective of brood size manipulation (Neuenschwander et al. 2003), or a tradeoff between provisioning rates per brood and prey size. In the latter case, parents often delivered larger prey items at lower provisioning rates (Stoehr et al. 2001; García-Navas and Sanz 2010), a pattern which is sometimes restricted to fathers (Siikamäki et al. 1998, but see also Wright and Cuthill 1990). Nonetheless, different prey types can show different nutritive values (Arnold et al. 2010; Wiesenborn 2012), so parents may also compensate for a decrease or increase in their provisioning rate by adjusting prey proportions rather than prey sizes (Magrath et al. 2004). Therefore, it is necessary to assess intrapopulation variability and the plasticity of individual prey selection in order to fully understand parental feeding behaviour and foraging strategies. For instance, previous research highlighted a link between different foraging strategies and productivity in herbivorous mammals (White 1983; Spalinger and Hobbs 1992), which display less sophisticated foraging skills than insectivorous birds. The high variability of parental foraging strategies within and among populations of many species suggests that the optimal foraging strategy may depend on the environmental conditions present (Stephens et al. 2007; Naef-Daenzer 2012). Parents would also modify their foraging strategies within populations, adjusting prey proportions and prey sizes to changing environmental conditions or levels of effort. Animals exploit diverse food resources by implementing different foraging strategies. Sometimes there are several profitable food types, so different species specialize in different food types according to their foraging and breeding strategies (Heithaus et al. 1975). On the other hand, generalist and specialist species may show divergent distribution patterns and population trends depending on how they exploit their habitat (Seamon and Adler 1996), and different populations of the same species may display either a generalist or a specialist strategy depending on the habitat (Quevedo et al. 2009). There is a lack of research, however, on the variability of these foraging strategies and their consequences for fitness within populations. Breeding
13
passerines use a wide variety of prey as food sources (e.g. Barba and Gil-Delgado 1990; Grundel 1990; Blondel et al. 1991; Banˇbura et al. 1994; Gilroy et al. 2009; Maziarz and Wesołowski 2010). However, in most cases it is also unclear whether parents may be considered generalists or whether different pairs specialize in different prey types within this wide range of variation (Estes et al. 2003; Bolnick et al. 2003; Woo et al. 2008; Newsome et al. 2009). In the Mediterranean area, bird populations rely more often on alternative prey than populations of the same species inhabiting northern latitudes. This variability in foraging strategies fits with the diverse and variable prey distribution that is characteristic of the Mediterranean area (Banˇbura et al. 1994; Blondel et al. 2010). In the case of the great tit Parus major, we recently reported that a population inhabiting a Mediterranean mixed forest presented high consistency in parental effort and prey selection over time (PaganiNúñez and Senar 2013) and consumed a wide variety of prey (Pagani-Núñez et al. 2011; Pagani-Núñez and Senar 2014). The great tit shows a wide geographical distribution across Europe and Asia (see e.g. Kvist et al. 2007), and it is possible to find viable populations in a wide variety of habitats (forest, agricultural and urban landscapes). Since different habitats may require different breeding and/or foraging strategies, characterizing the intrapopulation variability of parental foraging behaviour could represent a reliable approach to explaining and understanding animal distribution patterns (Naef-Daenzer 2012). In this study, our main objective was to estimate individual diet specialization with respect to the mean diet of the population in order to assess the variability in foraging strategies among breeding pairs. This allowed us to characterize each pair as more or less generalist or specialist. To experimentally evaluate the consistency of these strategies, we also evaluated the consistency of parental foraging behaviour by means of a brood size manipulation experiment. We hypothesized that, regardless of prey availability (and without underestimating its relevance in shaping population patterns), parents with a pure generalist foraging strategy should respond to brood size manipulation by adapting prey proportions or prey size to an increase or decrease in their provisioning rate (Grieco 2001; Stoehr et al. 2001; García-Navas and Sanz 2010). In reduced broods, we could expect an increase in provisioning rates per nestling and/or in the delivery of high-quality prey such as caterpillars and spiders, while we could expect the opposite response in increased broods. Conversely, parents with a specialist foraging strategy should not respond to our experimental manipulation (in relation to prey quality) because specialization would imply a loss—to some extent—of ability to shift individual dietary preferences (Trowbridge 1991; Nosil 2002), or may result in non-overlapping foraging niches between individuals (Bolnick et al.
Oecologia
2003; Newsome et al. 2009; Vander Zanden et al. 2010; Quiroga et al. 2011). Finally, we ascertained how intra- and interpopulation variability in parental foraging strategies could help to explain the distribution patterns of great tits.
Materials and methods Fieldwork and diet recording During the spring of 2012, we monitored 182 nest boxes at Can Catà field station (45°27′N, 2°8′E; Catalonia, northeast Spain), a predominantly evergreen mixed forest. Nest boxes were checked twice weekly to determine laying and hatching dates and brood size. Habitat structure showed a marked gradient of variation in relation to tree composition. The proportion of oaks (Quercus spp.) ranged from 5 to 95 % and correlated negatively with the altitude above sea level (82–219 m above sea level; see Pagani-Núñez et al. 2011, 2014). The main tree species were Aleppo pine Pinus halepensis, holm oak Quercus ilex and the oak Q. cerrioides. Nestling diet was successfully recorded in 31 nests by means of Micro-D cameras attached to the covers of the nest boxes and focused on the entrance (see PaganiNúñez and Senar 2014 for more details on camera properties and filming procedures). We filmed for 48 h in each nest when nestling age was 10–14 days (Pagani-Núñez and Senar 2013). Cameras were installed and camouflaged the day before recording to minimize the impact of installing the device. Recording was performed for five hours each day, from 7 a.m. to 12 a.m. We used full clock hours to standardize for interindividual differences in the time at which recording began. The sex of each parent was easily identified from the recordings. In addition, most individuals had been previously marked with numbered PVC rings. We determined prey type and size for each feeding action of each parent within this time window. Prey size was determined according to a semi-quantitative scale with three values: 1 = small; 2 = medium; 3 = large (Barba et al. 1996). In this work, we define brood size as the number of nestlings in the nest at the time of recording. Diet overlap and trophic specialization We assessed between-pair trophic specialization in relation to population niche breadth by means of the proportional similarity index, PSi = 1 − 0.5∑j │pij − qij│ = ∑j min(pij, qij), and the mean pairwise overlap (see Bolnick et al. 2002 for more details). The former is a measure of the overlap between the diets of an individual and the population; the latter is the mean PSi between all pairwise combinations of individuals, or—in our case—pairs (Bolnick et al. 2002). Both indices were obtained through IndSpec, version 4.0
(Bolnick et al. 2002). These indices provide information on the use of trophic resources by individuals properly framed in the context of a concrete population, and range from 0 to 1, where 1 indicates that the individual’s diet is equal to the mean diet of the population and lower values indicate that the individual specializes with respect to the population mean. To perform this analysis, we used nestling diet proportions on control days (5 h of recordings from 7 a.m.; see Pagani-Núñez and Senar 2013), which represented the normal conditions for parents. We used mean values for each pair instead of individual data because our aim was to assess the diet overlap between territories (that is, between pairs or nests). In addition, we standardized prey proportions to the mean size of each kind of prey in each nest (following Barba et al. 1996, as previously described): we multiplied mean prey size by the number of prey items and calculated new percentages on this basis. We divided prey into fourteen classes, which can be listed as follows according to their functional groups and their importance in the nestling diet: caterpillars (Lepidoptera larvae), chrysalises (Lepidoptera pupae), moths and butterflies (Lepidoptera adults), other larvae, spiders (Araneae), spiders’ eggs, grasshoppers (Orthoptera), stick insects (Phasmida), mosquitoes (Diptera), bees and flying ants (Hymenoptera), beetles (Coleoptera), unidentified arthropods, fruits and formless remains. We used the most detailed classification to investigate diet specialization and how parents combine different prey classes, but we simplified it to facilitate the assessment of parental responses to our experimental manipulation. We recorded a total of 2362 feedings. The mean number ± SE of total feedings per pair across 5 h of recording was 76.19 ± 7.00, while the mean number of feeding trips per nestling was 15.81 ± 1.41. These analyses included a total of 62 parents from 31 nests. Brood size manipulation experiment In addition to the degree of diet specialization among pairs, we assessed the consistency of their foraging behaviour under different levels of effort. We performed a brood size manipulation experiment in 18 nests when nestlings were 10–14 days old. Our sample consisted of 36 parents and 94 nestlings (a mean of 5.22 nestlings per nest), 18 of which were moved between nests to experimentally modify brood size. We used pairs of nests with the closest hatching dates (up to 2 days) and with brood sizes that were the same or similar, thus decreasing the number of available nests from the initial 31 which we successfully filmed. That is, some nests remained unpaired due to these limitations and were not used in the experiment. Although there was great interpair variability, we collected our sample across the whole breeding season (we expect that this minimized the impact of reducing the initial sample). We randomly selected and
13
Oecologia
moved two nestlings from one nest to the other for 24 h, so we obtained a simultaneous experimental increase and decrease in brood size in the two complementary nests (“reduced” = −2 or “increased” = +2). For instance, a pair of nests with five nestlings each were adjusted to have three nestlings in one and seven in the other. The experiment was restricted to two nestlings to keep brood size within the natural range of variation for this species (Orell et al. 1996). If the nests did not have the same brood size, we reduced the larger nest and increased the smaller nest of the experimental pair. We recorded the “normal” and the “experimental” day for each nest. Thus, the control for each pair was their own “normal” foraging behaviour. Moreover, we performed the experiment before or after recording the normal day in successive pairs of nests. After the recording, we placed both of the moved nestlings back in their original nests. Provisioning rates per hour and provisioning rates per nestling, prey proportions and mean prey size were determined for each individual. In this case, prey were classified as caterpillars, spiders or others in order to focus on the most relevant prey types, following our previous procedures (Pagani-Núñez and Senar 2014; PaganiNúñez et al. 2014). Statistical analysis Once we had characterized the trophic structure of our study population, we aimed to identify which extrinsic and/or intrinsic factors determined this structure. We used an Akaike’s information criterion (AIC) approach to assess which intrinsic (brood size and foraging strategy) and extrinsic (date and habitat structure) factors were more relevant in shaping PSi values. The AIC provides information on how well a model based on a specific variable (or set of variables) fits to a particular dataset; it is a measure of the uncertainty (or entropy) in the data which the model does not account for (Burnham and Anderson 2002). The smaller the value of the AIC, the better the fit of the model to the data. Models that differ by less than two in their AIC scores (ΔAIC ≤2) fit the data similarly well. Moreover, we computed the AIC weights (which can be interpreted as conditional probabilities) for each model (Wagenmakers and Farrell 2004). Our analysis was based on a normal distribution with a log link function. The dependent variable was PSi value (after being arcsine square-root transformed), and independent variables included the number of prey types in the nestling diet, brood size (number of nestlings during recording on control days), date (number of days from 1st April), proportion of the trees surrounding the nest boxes that were Quercus spp. (to control for habitat quality, see Pagani-Núñez and Senar 2014), provisioning rate per nestling, and mean prey size. We included variables
13
that characterize parental foraging strategies (provisioning rates and prey size) and brood size as intrinsic factors to ascertain the energetic costs and the consequences for fitness of different foraging strategies. The number of prey types selected by each pair helped us to quantitatively interpret the significance of our PSi values; in other words, whether being more specialist directly implied the selection of fewer prey types. We included the date of breeding and the proportion of the trees surrounding the nest boxes that were Quercus spp. as extrinsic factors, since these features strongly affect prey composition (see e.g. Pagani-Núñez and Senar 2014). Moreover, if these environmental features are correlated with PSi values, it could indicate a strong effect of prey availability on parental foraging strategies. Breeding date and habitat structure may determine brood size (see e.g. Perrins and McCleery 1989; Lambrechts et al. 2004), so we assessed the relationships among these variables beforehand. We found that pairs inhabiting territories with lower proportions of Quercus spp. showed a tendency to raise larger brood sizes (n = 31, r = −0.32, t = −1.80, p = 0.08) and to breed earlier in the season (n = 31, r = 0.32, t = −1.91, p = 0.07). Brood size did not correlate with date (n = 31, r = −0.07, t = −0.40, p = 0.69). There was considerable time overlap among nests (see Appendix A1). In order to investigate which prey types were most commonly captured together, we used a principal components and classification analysis (PCCA). Using a batch of variables, PCCA performs a principal component analysis and then shows how these variables are interrelated. We used the numbers of each prey type in each nest as variables for analysis, but we excluded the four least common prey types (Hymenoptera, Diptera, chrysalises and unidentified insects; each of these represent around 1 % of the nestling diet) to increase the statistical power of the PCCA. This approach allowed us to probe how and why parents combine several food sources. Based on our previous research, we would not expect caterpillars and spiders to be associated with each other, or with other relevant prey types such as grasshoppers or moths. Caterpillars and spiders have been described as key food sources for many bird species, and this finding would justify considering three functional food sources (caterpillars, spiders and others) when investigating diet composition. Moreover, the approach helped us to ascertain whether parents combine their prey according to availability patterns, or whether they select prey based on their characteristics (e.g. if they simultaneously focus on these similar prey types when foraging). Finally, we analyzed the consistency of parental foraging behaviour within our population. To analyze parental responses to our experiment, we used a repeated measures analysis of variance approach. We used a control and an experimental day for each pair; that is, the control for
Oecologia
each pair was their own natural behaviour. The provisioning rate per brood in the control and the experimental day were our dependent variables. Our within-subject factor was the comparison between these variables before and after the experiment, which we referred to as “time”, while the experimental group (experiment: “reduced” or “increased” if the dependent variable differed between experimental groups) and the sex (male or female if the sexes differed a priori) were the between-subject factors. We assessed their interactions to ascertain if the parents responded to our manipulation (time × experiment) and if there was a sex-related response (sex × experiment and time × sex × experiment) or some bias in our experimental procedure. We did the same for each of the following dependent variables: provisioning rate per nestling; proportion of caterpillars, spiders and others; mean prey size. A chi-square normality test showed that provisioning rate per brood [control day: χ2= 2.81, df = 4 (adj.), p = 0.59; experimental day: χ2 = 4.10, df = 3 (adj.), p = 0.25], provisioning rate per nestling [control day: χ2 = 2.41, df = 2 (adj.), p = 0.30; experimental day: χ2 = 2.61, df = 2 (adj.), p = 0.27], and mean prey size [control day: χ2 = 3.25, df = 3 (adj.), p = 0.35; experimental day: χ2 = 4.30, df = 3 (adj.), p = 0.23] were normally distributed. Moreover, prey proportions were arcsine square-root transformed. We used Statistica 6.0 to perform our analyses (StatSoft 2001), except for the PSi values, which were computed with R (R Development Core Team 2014).
Results Caterpillars were the main prey type, representing half of all food (55 %, Appendix A2). Spiders were the second most common prey type (14 %, Appendix A2), and the proportions of the food represented by the remaining prey types ranged from 1 to 7 % (Appendix A2). However, nestling diet was extremely variable among territories. For instance, caterpillars ranged from 16 to 93 % of the food, and not all prey types were present in all nests (Appendix A2). That is, different pairs selected different combinations of prey types. The similarity index score (SI, population mean of PSi) was 0.72, with PSi values for pairs ranging from 0.53 to 0.84 (Fig. 1). The mean pairwise overlap was 0.63. In certain cases, pairs of nests which overlapped in time and space displayed dissimilar PSi values (Appendix A1). Our AIC approach showed that the best model explaining the variability in PSi values included only brood size (Table 1), although the top seven models fitted the data similarly well (ΔAIC ≤2, Table 1). In the top model, the confidence interval of the coefficient included zero (lower CL 95 %: −0.05, upper CL 95 %: 0.001). Indeed, PSi values were negatively related to brood size (Fig. 2), suggesting that breeding
Fig. 1 Histogram showing the distribution of proportional similarity index values within our study population. Data correspond to 31 pairs of Mediterranean great tits Parus major and are based on natural brood sizes
pairs with more specialized diets than the overall population more frequently had larger broods. The second model included brood size and the proportion of the trees surrounding nest boxes that were Quercus spp. (Table 1), and the third included the proportion of the trees surrounding nest boxes that were Quercus spp. (Table 1). Models 4–7 included brood size and, respectively, provisioning rate per nestling (Table 1), mean prey size (Table 1), date (Table 1) and number of prey classes (Table 1). The first two axes of the principal component and classification analysis explained the most variance: 50.73 % of the variation in nestling diet (Fig. 3). The first component showed positive associations among the majority of prey types (Table 2). Stick insects were collected independently of those types, while moths and butterflies were selected in opposition to that majority of prey types (Table 2). The second component showed positive associations among caterpillars, fruits and other larvae (Table 2). On the other hand, spiders and their eggs seemed to be collected independently (Table 2). Finally, grasshoppers, stick insects, moths and butterflies, beetles and formless remains were collected in opposition to the first group (Table 2). The third component provided further confirmation that stick insects were often collected independently (Table 2), while the fourth and fifth components showed secondary associations between moths and butterflies and other larvae, and between caterpillars and grasshoppers in opposition to stick insects (Table 2). Thus, there were three main assemblages among the prey types, although two of them (stick insects and moths and butterflies) could be segregated into a fourth group and showed secondary associations (Fig. 3). Our repeated measures approach showed that parents reduced the number of feeding trips after we reduced brood
13
Oecologia Table 1 Results of our Akaike’s information criterion (AIC)-based method of assessing which intrinsic (brood size and foraging strategy) and extrinsic (date and habitat structure) factors were more relModel no.
Variable 1
Variable 2
1 2
Brood size Brood size
evant in shaping proportional similarity (an index of diet specialization) values among 31 breeding pairs of great tits Parus major df
AIC
ΔAIC
wi(ΔAIC)
% Quercus spp.
1 2
−55.80 −54.73
0.00 1.06
0.23 0.13
Provisioning rate per nestling Mean prey size Date Brood size Mean prey size
2 2 2 2 2
3
% Quercus spp.
4 5 6 7 8
Brood size Brood size Brood size No. of prey types % Quercus spp.
9
% Quercus spp.
Provisioning rate per nestling
10
Brood size
Date
Variable 3
1
2 % Quercus spp.
3
−54.59
−53.83 −53.82 −53.81 −53.80 −53.31
−52.96
−52.96
1.21
0.12
1.97 1.98 1.99 2.00 2.49
0.09 0.08 0.08 0.08 0.07
2.83
0.06
2.84
0.06
We included variables that characterize parental foraging strategies (provisioning rate per nestling and mean prey size) and brood size as intrinsic factors. We included the date of breeding, a measure of habitat structure (% of trees surrounding nest boxes that were Quercus spp.) and the number of prey types as extrinsic factors. We show the top ten models in order of statistical power
Fig. 2 Relationships between proportional similarity index (y-axis) and natural brood size (x-axis) during recordings of 31 pairs of great tits Parus major
size, and vice versa (reduced broods: from 37.95 ± 5.51 to 27.35 ± 2.90 feeding trips vs. increased broods: from 34.12 ± 5.57 to 48.62 ± 6.59 feeding trips; Table 3; Fig. 4). Experimental groups did not differ a priori in provisioning rate (Table 3). Provisioning rate per brood did not differ between the sexes and did not vary in relation to sex after our experimental procedure (all p > 0.24; Table 3). Additionally, provisioning rate per nestling did not differ a priori between experimental groups, nor between the sexes, and remained constant after we modified brood size (reduced broods: from 6.90 ± 5.91 to 7.54 ± 4.42 feeding trips per nestling vs. increased broods: from 7.62 ± 4.50 to 7.64 ± 3.96 feeding trips per nestling, all p > 0.24; Table 3). The proportion of caterpillars differed between experimental groups prior to experimental procedures
13
Fig. 3 Relationships among the ten most important prey types delivered to nestlings by 31 Mediterranean great tit Parus major pairs in natural conditions based on principal component and classification analysis. Prey types that are closer on the diagram were combined across the whole study period more often, while prey types that are farther apart were combined less often. Prey types close to the centre of the diagram were delivered independently
(reduced broods: 37.95 ± 5.51 % vs. increased broods: 52.37 ± 0.06 %; Table 4). The proportion of caterpillars delivered by parents did not differ between the sexes and remained constant after our experimental manipulation of brood size (all p > 0.18; Table 4). Females delivered more spiders than males (females: 18.80 % ± 0.03 vs. males: 12.03 % ± 0.02; Table 4), but the proportion of spiders
Oecologia Table 2 Results of the principal component and classification analysis of the ten most common prey classes within our population Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 % Var explained 27.55 No. caterpillars 0.50 No. spiders 0.79 No. fruits 0.38 No. grasshoppers 0.46 No. stick insects −0.01 No. moths and but- −0.43 terflies
23.17 0.46 0.01 0.74 −0.62 −0.34 −0.17
13.50 0.30 −0.17 0.09 −0.09 0.74 −0.63
10.09 0.27 0.03 −0.38 −0.18 −0.21 −0.46
8.29 −0.46 0.24 −0.07 −0.47 0.42 0.00
No. other larvae No. beetles No. spider eggs
0.39 0.28 0.75
0.09 −0.19 0.29
0.74
−0.02 0.37 −0.21
−0.55 −0.37 0.26
No. formless remains
0.62 −0.65 −0.03
0.00
0.12
−0.47
−0.30
We used the number of prey of each type delivered by 31 pairs of great tits Parus major to their nestlings (during 5 h of recording) as variables for analysis, to show how parents combined the various food sources
Table 3 Repeated measures analysis of variance using provisioning rate per brood or nestling for 19 pairs of great tit Parus major parents as the dependent variable, experimental group (reduced or increased) as the within-subject factor, and sex (male or female) as the betweensubject factor
Provisioning rate per brood Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment Provisioning rate per nestling Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment
SS
F1,32
p level
1159.21 1353.34 67.60 127.21 0.54 2800.04 20.54
1.43 1.68 0.39 0.16 0.16; Table 4). Finally, mean prey size did not differ a priori between experimental groups, nor between the sexes, and also remained constant after we modified brood size (all p > 0.09; Table 4).
Discussion Our findings support the idea that Mediterranean great tits can be considered diet generalists, while we recorded a tendency for diet specialization among breeding pairs. Interestingly, pairs raising larger broods showed more specialized diets, and diet specialization did not correlate with habitat structure or breeding date. However, pairs showed a tendency to raise larger brood sizes in territories with lower proportions of Quercus spp., and they occupied territories with lower proportions of Quercus spp. earlier in the season. These correlations with environmental features may help us to interpret our results. Previous research has stressed the key roles of habitat structure (Schneider and McNally 1993; Pienkowski et al. 1998) and the timing of breeding (Beja 1996; Eide et al. 2004; Rutz and Bijlsma 2006) in shaping prey availability and breeding performance of wild animals. This pattern is particularly marked in insectivorous birds due to their dependence on insect prey, which vary in their peaks of abundance (NaefDaenzer et al. 2000; Tremblay et al. 2003; Blondel 2007; Eeva et al. 2009). The great tit and its close relative the blue tit Cyanistes caeruleus (hereinafter “tits”), though usually considered diet generalists, mainly rely on caterpillars to feed their offspring (Naef-Daenzer et al. 2000; Tremblay et al. 2003; Blondel 2007; Eeva et al. 2009). Accordingly, it has been suggested that there is a lack of fit between their life-history strategies and environmental conditions at both extremes of their distribution ranges (Sanz 1998; Rytkonen and Orell 2001), and especially within evergreen forests (Sanz 1998; Fargallo 2004; Lambrechts et al. 2004). As has been shown, we could expect suboptimal conditions early or late in the season (before or after the peak in caterpillar abundance) and in territories with a higher proportion of Aleppo pines (where food is usually scarce). In this regard, the lack of alternative, profitable prey types to caterpillars strongly constrains tit breeding performance (Rytkonen and Krams 2003), while the presence of such alternative prey types is what guarantees the viability of southern marginal populations (Blondel et al. 1991; Banˇbura et al. 1994; Ziane et al. 2006; Pagani-Nuñez et al. 2011). In line with this idea, our findings hint that diet specialization could operate as the main mechanism through which great tits avoid these habitat and temporal constraints, allowing them
13
Oecologia Table 4 Repeated measures analysis of variance using the proportion of caterpillars or spiders or other prey delivered by 19 pairs of great tits Parus major to their nestlings or the mean prey size as the dependent variable, experimental group (reduced or increased) as the within-subject factor, and sex (male or female) as the between-subject factor
Fig. 4 Provisioning rates per brood in natural conditions and after the experimental procedure of increasing or decreasing the number of nestlings in the nests of 18 pairs of Mediterranean great tits Parus major. Filled circles correspond to reduced broods while empty squares correspond to increased broods
to increase or maintain their numbers under suboptimal conditions. In other words, in spite of the limiting effect of habitat structure and date on brood size, which is particularly evident in the Mediterranean area, more specialized pairs kept broods of comparable sizes to central European populations. On the other hand, increased interpair variability would be expected if parents raising larger broods have less time to search for food and have to exploit narrower foraging niches. Moreover, within species, niche partitioning may be promoted by spatial and temporal segregation (Cherel et al. 2009), and individuals may specialize in the exploitation of certain habitats (Paterson et al. 2012; Rousseau et al. 2012). In any case, great tit parents displayed a gradient from generalist to specialist foraging strategies (a pattern also found in other systems, e.g. Cherel et al. 2009), which was mainly determined by parental investment in reproduction. This dynamic pattern of diet specialization fits the diverse and variable prey distribution that is characteristic of the Mediterranean area (Blondel et al. 2010), highlighting that diversity of prey may also promote diversity of foraging strategies within species and populations (Sinclair et al. 2003; Duffy et al. 2007). The lack of plasticity of parental foraging behaviour in the short term may hint at why the relationship between brood size and foraging niche is so relevant. Several studies performed in different environments and populations of great and blue tits have reported contrasting results in relation to the capacity of parents to adjust their investment in reproduction to environmental conditions (Sanz 1998; Rytkonen and Orell 2001; Tremblay et al. 2003; Mägi et al. 2009; Atiénzar et al. 2010). We found that great tit parents, irrespectively of sex, rapidly modified their behaviour
13
% Caterpillars Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment % Spiders Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment % Others Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment Mean prey size Sex Experimental group Time Sex × experiment Time × sex Time × experiment Time × sex × experiment
SS
F1,32
p level
0.16 0.36
