Predation by sparrowhawks decreases with increased breeding density in a songbird, the great tit
- First Online:
- Cite this article as:
- Götmark, F. & Andersson, M. Oecologia (2005) 142: 177. doi:10.1007/s00442-004-1715-z
Predators may regulate prey populations if predation rate increases with prey density. Alternatively, if space-limited (e.g. territorial) predators become ‘satiated’ when prey exceed a certain density, increased prey abundance may lead to reduced predation rate. These alternatives have been difficult to test experimentally for mobile prey in the wild. We present such a test, manipulating the density of great tits (Parus major) by adding nest boxes in territories of sparrowhawks (Accipiter nisus). Predation rate was measured for young tits after they left the nests. Although the great tit is an important prey, there was no evidence for regulation during the breeding season: the rate of hawk predation declined with increasing density of tits. This result was not confounded by changes in breeding density of alternative prey species (other songbirds). Hawk predation can therefore favour dense breeding in a territorial (solitary) bird, and conspecific attraction and aggregation reported in several territorial species may partly result from predation pressure. This result also has potential implications for conservation work.
KeywordsConservationDensity dependenceField experimentFledgling preyPredator satiation
Some predators can regulate the density of their prey, but the general importance of such regulation is not clear, alone or relative to other forms of regulation based for instance on availability of food and nest sites (Crawley 1992; Sinclair and Pech 1996; Newton 1998). One reason is that for mobile prey, predator regulation is difficult to study experimentally in the wild.
Predation has favoured many different forms of defence in prey (Crawley 1992), their dispersion behaviour being one important aspect. Dispersion usually makes detection of (cryptic) prey more difficult, and may reduce predation risk (Curio 1976). Alternatively, groups or loose aggregations of prey may satiate predators that are limited in numbers by territoriality, nest sites or other resources (Krebs and Davies 1993). Therefore, predation rate may decrease with increasing prey aggregation size, a ‘dilution effect’ (e.g. Foster and Treherne 1981).
Many songbirds and other terrestrial prey species are dispersed during the reproductive season, and such spreading out may reduce predation risk. Several studies suggest, however, that songbirds form loose aggregations through conspecific attraction when they settle in territories at the start of the breeding season (Ward and Schlossberg 2004, and references therein). Attraction between settling territory owners was described as long ago as 1937 and has been much studied since (see Stamps 1988). Several explanations for conspecific attraction, such as mating advantages, have been proposed (Stamps 1988; Ward and Schlossberg 2004), but it remains far from well understood. Reduction in local predation risk is another possible contributing reason for conspecific attraction (also mentioned in Stamps 1988). For predators such as Accipiter hawks that rely on surprise to catch prey, Hamilton’s (1971) two-dimensional selfish herd model may help explain the initial evolution of grouping in randomly dispersed prey.
In the avian literature (especially regarding songbirds), predation on fledglings and adults is often viewed as relatively unimportant for regulation of breeding populations, compared to, e.g., food availability (Newton et al. 1997; Newton 1998). However, no study of breeding songbirds that we are aware of has manipulated densities of adults or fledged juveniles experimentally, and subsequently measured predation rates. Food availability is probably important in regulating breeding songbird populations, but there may be interactions between food and predation (Martin 1992). Predation mortality of adults and fledglings may increase when food is scarce, because foraging effort then increases (for evidence in the chaffinch (Fringilla coelebs), see Götmark et al. 1997). Many studies of territorial songbirds implicitly assume that food and nest sites are the primary determinants of spacing patterns, whereas conspecific attraction and predation on adults or fledglings are rarely mentioned.
Here, we use observational data in combination with a field experiment to explore the importance of predation for population regulation and prey spacing in a songbird in the breeding season. There are two alternative research hypotheses: (1) predation rate increases with prey density, so predation contributes to prey population regulation, or (2) predators become ‘satiated’ when prey density exceeds some level, above which predation rate declines with increasing prey density. In this situation, predators are unlikely to regulate the prey during the breeding season.
We test these alternatives in one of the most often studied birds, the great tit (Parus major), and its main predator (with respect to fledglings and adults), the Eurasian sparrowhawk (Accipiter nisus). Instead of removing predators, the main method used in earlier experiments, we increased the density of prey and measured the predator response, using sparrowhawk pairs (territories) as sample units. As predators of songbirds have large hunting ranges in the breeding season, predation rate and prey density need to be studied in a sufficiently large area. We used predator pairs as sample units to avoid pseudoreplication, which might arise if prey were used as sample units, with only few predators involved. As we used hawk territories for sampling, conclusions for predator and prey apply at the landscape level (Forman 1999).
High predator density and high predation rate might contribute to extinction of prey under certain conditions (Newton 1998; Sinclair et al. 1998). If high prey density satiates predators, the highest predation rates may occur at low prey densities, potentially leading to a ‘predation trap’ from which prey may be unable to increase (Newton 1993, 1998). As small, sparse and declining populations are of great conservation interest, our predation experiment also has relevance for important applied aspects.
We test the null hypothesis that predation rate is independent of prey density, i.e. that sparrowhawks take similar proportions of great tits in each hawk territory, irrespective of tit density. We first analysed observational data (great tit prey remains) from 43 different hawk territories. The results indicate that sparrowhawk predation rate decreases with increasing density of great tits. For a controlled test of the hypothesis that higher tit density leads to lower predation by sparrowhawks, we then carried out a field experiment, which corroborates the hypothesis generated by the previous observational results. For both data sets, we also examined covariation in the density of alternative prey (based on census of potential prey in each hawk territory).
Materials and methods
Study area and species
The study was carried out over 5 years (1994–1998) near Göteborg, SW Sweden. The study area of about 500 km2 includes a large nature reserve (Sandsjöbacka) south of the city, where most (about 15) sparrowhawk pairs nested. Several pairs nested relatively close to urban areas and farms outside the city and a few pairs nested in forest parks in the city. Forest covered 60–70% of the study area, deciduous and coniferous forest about equally; other habitat types were fields and meadows, urban habitats, exposed bedrock, lakes, and mires. The study area has an altitude <100 m and is relatively productive, with a diverse fauna of breeding birds (Götmark and Post 1996).
In Europe and large parts of Asia, the sparrowhawk is a major predator of woodland songbirds (Newton 1986). The hawk brings prey to the nest site, and often removes feathers at plucking posts, usually less than 30 m from the nest. The male provides most of the prey, feeding the female during nest-building and incubation. After hatching, she feeds the chicks with food delivered by the male. The female usually starts hunting when the young are half-grown and need more food (Newton 1986). In our study area, sparrowhawks mainly nest in 25- to 50-year-old stands of planted Norway spruce Picea abies. Each spring, we visited sites where hawks had nested, and also searched for new sites. Twenty-five to 30 hawk pairs started breeding each year, but a few failed and normally 17–24 pairs bred successfully. Pairs that failed relatively late (end of June or later) were included in the analyses if we had large samples of prey from them. In total, 43 hawk territories were included in the analyses; of these, 4 were occupied in all 5 years, 5 in 4 years, 8 in 3 years, 11 in 2 years, and 15 in 1 year.
The great tit, a non-migratory passerine bird with an all-purpose breeding territory, nests in natural holes or nest boxes in forest (Cramp and Perrins 1993; Gosler 1993). It is a common prey of sparrowhawks (Tinbergen 1946; Geer 1978, 1982; Perrins and Geer 1980; Gray 1987; McCleery and Perrins 1991; Gosler 1993; Gosler et al. 1995; reviews in Newton 1986, 1993, 1998). In our study area, the great tit was the most common species in prey remains of the hawks (Götmark and Post 1996; Götmark 2002). The breeding density of great tits varies considerably among hawk territories (see below), making this species suitable for study of density-dependent predation. The majority of great tits taken by breeding sparrowhawks are fledglings (see references above). Great tit fledglings are fed and cared for by parents for about 10 days after they have left the nest, and they are especially vulnerable to predation at this time (e.g. Newton 1986; Götmark and Olsson 1997; Götmark 2002).
Prey remains are often used to study prey choice in sparrowhawks. This method does not give a fully accurate picture of prey choice (review in Newton 1986), but for the present purpose the limitations are not serious or likely to lead to bias. Each year, we collected prey remains from about 10 April to early August (when the hawks begin to leave the nest area). We used wing and tail feathers to identify prey (in a few cases, heads and legs were used). Feathers collected at each visit to a hawk nest were lumped, and for each visit we later determined the minimum number of individuals of each prey species (Reynolds and Meslow 1984). This was done by matching wing and tail feathers, especially the outermost tail feathers that are distinctive in many species. When we found many feathers of a species, we divided total numbers of wing (or tail) feathers with the actual (mean) number for an individual (based on the literature) to estimate the number of prey. Fledglings were identified from feathers that were not fully grown (with sheaths), but a few weeks after fledging they become difficult to identify. We pooled adult and fledgling great tits (in observational data) to increase sample sizes and avoid misidentification of older fledglings. Nest sites and plucking posts were visited at 8- to 15-day intervals for collection of feathers. At least 30 collected prey from the breeding season were required for inclusion of a hawk pair (territory) in analyses.
Forty-six prey species were studied (see Götmark and Post 1996). The total number of prey collected during the 5 years was 8,440; the average annual sample of prey per hawk pair varied between 73 and 98 (1994–1998). The great tit was the most common prey (20% of all prey; adults and fledglings). Identification of prey feathers was done by one person only (Jan Olsson).
Breeding great tits and other potential prey were censused by 2-km line transects near each hawk nest, along a square with sides of 500 m. For great tits, which are territorial early in the season, we used census data from April; for other species, we used census data either from April or May depending on the start of male territorial activity (i.e. mainly singing). A census method that corrects for variation in detectability of species was used (for details, see Götmark and Post 1996). One of us (F.G.) censused all transects in all years, recording singing males and other individuals that were judged to represent breeding pairs. Each census day, two hawk territories were censused for prey, from about 0500 to 1000 hours (local time). We found no significant diurnal or seasonal change in the number of recorded great tits, so the raw data were used without corrections; for other species, we corrected as described by Götmark and Post (1996).
A total of 17,427 pairs (potential prey) were recorded in the line transects. Based on means for the 5 years, and correcting for detectability bias, the three most common songbirds were willow warbler (Phylloscopus trochilus) (14% of all recorded pairs), robin (Erithacus rubecula) (11%), and chaffinch (10%). The great tit was fourth commonest in the line transects (9%). Its proportion was higher in the prey remains (20%), suggesting that the great tit is a vulnerable prey, subject to relatively high predation risk (see also Götmark and Post 1996, with value of relative predation risk for great tit).
In the analyses, each hawk territory was only included once (one randomly chosen season if the territory was studied for several years). Observational data from 36 hawk territories were from 3 years (1994–1996); in addition we used data from seven territories that were not manipulated in the experiment in 1997–1998 (in total, n=43).
The density of breeding great tits might be correlated with the density of other breeding songbirds that are potential prey of sparrowhawks. We used our census data (n=43 territories) to analyse the correlation (Pearson’s r) between great tit density and density of six major prey species (Götmark and Post 1996). The r values were as follows; willow warbler −0.24, robin 0.17, chaffinch 0.01, house/tree sparrow (Passerdomesticus/montanus) 0.02, blue tit (Paruscaeruleus) 0.34, and blackbird (Turdus merula) −0.24; pooling these seven species gave an r value of 0.03, and for all 46 prey species pooled the r value was 0.06. Only the correlation between great tit and blue tit was significant (P=0.03). The blue tit was less common (5.8% of recorded pairs) than the great tit, and less ‘preferred’ as prey by sparrowhawks (see relative predation risk value in Götmark and Post 1996).
In the field experiment 1997–1998, we set up nest boxes at two different densities near hawk nests (within 500 m), recording predation by recovering rings of great tit fledglings taken by hawks (see below). Great tits nested in about 65% of the nest boxes; the remaining ones were empty (about half) or contained nests of blue tits or pied flycatchers (Ficedula hypoleuca). Sparrowhawks usually prefer to hunt close to their nest (Geer 1978), as shown by radio-tracking studies (Newton 1986; Selås and Rafoss 1999). A review of the effects of provisioning of extra nest boxes in bird populations (Newton 1994) concluded that such boxes in almost all cases increase the density of species nesting in natural holes and nest boxes, which also seemed to be the case for great tits in our experiment (see below).
In 1997, before the breeding season, we set up either 14 nest boxes (in each of 12 hawk territories) or 40–50 nest boxes (nine hawk territories) in or near the area where we counted prey in line transects (about 0.8 km2). Hawk territories could not be selected randomly in two groups, but were matched in several respects (e.g. in habitat composition) and we had no indication that non-random selection created any bias. Some of the territories were not used by hawks in 1997, or they nested far from the nest boxes. In 1997 and 1998, we studied in total 15 experimental territories (2 studied in both years; a mean value for each territory was used in the analyses below). The nest boxes were generally set up in all directions from the hawk nest used in the preceding breeding season, with relatively even spacing in a circle with radius of 500 m (hawk nest in centre). Within the circle area of 78 ha, the increase in number of breeding great tits ranged from 5 to 21 pairs producing young (see below, n=15 hawk territories), corresponding to addition of 0.06–0.27 pairs/ha. The added density falls within the natural range of variation in breeding density of great tits.
The response of sparrowhawks to tit density might be ‘functional’, each predator increasing its consumption of tits, and ‘numerical’, the number of predators increasing (Crawley 1992; Sinclair and Pech 1996; Newton 1998). We studied both responses, the numerical one by census of nesting hawks within 2 km of each experimental hawk pair (nest). Sparrowhawk nests were sometimes close to each other (e.g. 500 m), and hawks might respond to increased great tit density in the experiment by nesting more densely. Sparrowhawks have partly overlapping hunting ranges (Newton 1986; Selås and Rafoss 1999), so we also studied predation by neighbours of experimental pairs (those nesting within 2 km; range 1–4 neighbours, mean 1.7, n=15). For each experimental hawk territory, the estimate of tit predation rate includes prey taken by the hawk territory owners and by the neighbours. Each year, less than half of the hawk pairs studied were experimental, and at least one of their neighbours lacked nest boxes put up for great tits.
Predation on fledglings was studied by ringing all young in the nest boxes, later recovering rings from fledglings eaten at the hawk nests (Perrins and Geer 1980; Götmark and Olsson 1997; Götmark 2002). Nestlings were ringed at 7–15 days old, except a few small and weak ones (likely to die before fledging). Each nestling had two numbered aluminium rings, one on each leg, to increase the chance of recovering rings (fledglings). After the young hawks fledge, from early July onwards, they stay for some weeks near their nest (Newton 1986). One to two weeks after their fledging, we used a metal detector to find rings in the hawk nest (which was taken down), under the nest tree, at plucking posts, and under perch trees. Most rings were in or under nests (82% in 1997 and 80% in 1998), usually in pellets shed by the hawks. We standardised comparisons among hawk pairs by only including rings found within and under the hawk nests. Rings from plucking posts (and perch trees) were only found for some pairs, partly reflecting the use of these sites by individual hawk pairs, and the extent to which we could find prey remains and rings there. Pellets with rings within and under the nest were probably mainly produced by the female (possibly the young), from prey delivered by the male. After the breeding seasons, we checked for remaining dead, ringed nestlings in the nest boxes and omitted these, as we only studied fledglings. Of 1,159 ringed fledglings, 7.3% were recovered by rings in or under hawk nests (i.e. rings of a total of 85 tits recovered in the 15 experimental territories).
As the number of ringed great tit fledglings varied widely among experimental territories in both of the two nest box density groups (14 and 40–50 nest boxes), we analysed predation rates using the number of ringed fledglings per territory as a continuous explanatory variable, rather than using two groups of territories.
There was a positive correlation between the number of great tit pairs in nest boxes and in line transects in hawk territories (r=0.55, P=0.035, n=15), suggesting that nest boxes increased tit density. We also used the census data (n=15 hawk territories) to analyse the correlation (Pearson’s r) between great tit pairs (nest boxes) and that of the other major prey species. The r values were as follows: willow warbler 0.05, robin 0.52, chaffinch 0.17, house/tree sparrow −0.16, blue tit 0.13, and blackbird 0.06; pooling all these species gave an r value of 0.03, and pooling of all 46 prey species an r value of 0.24. Only the correlation between the great tit and robin was significant (P=0.05). Compared to tits and house/tree sparrows, robins were relatively unimportant as prey in June, when most young tits leave their nests (Götmark and Post 1996, and discussion below).
Results and discussion
The negative (inverse) density-dependent predation on great tits suggests that sparrowhawks do not regulate the tit population in the breeding season. The functional response of sparrowhawks was weak, to judge from both observational (Fig. 1a) and experimental results (Fig. 2a). The neighbours of experimental pairs took on average only 22% of the recovered ringed fledglings killed by hawks in experimental territories (SD=22%, range 0–60%, n=15). The number of breeding neighbours found within 2 km of experimental hawk pairs was not related to the number of ringed tit fledglings in the experimental territory (Spearman rs=−0.15, P=0.88, n=15). Thus, the numerical response (including production of young by experimental hawk pairs, and nesting neighbours and their young) was weak. We have no data for hawks breeding >2 km from experimental pairs, or for non-breeding hawks (Newton 1986), but their impact was probably small, as they nested relatively far away, or did not feed young (non-breeders).
Absence of hawk regulation of tit fledgling numbers is consistent with lack of response in numbers of breeding tits when sparrowhawks almost disappeared in the late 1950s (due to pesticides) and later returned in the 1970s (Perrins and Geer 1980; Newton 1986; McCleery and Perrins 1991; Newton et al. 1997; Thomson et al. 1998).
Although sparrowhawks are the major predators of fledgling and adult great tits in our study area, other predatory birds may also take great tits. One potentially important predator is the pygmy owl (Glaucidium passerinum), but it does not nest in the study area. Tawny owls (Strix aluco) nest there, but great tits were relatively rare prey in their diet in a population study of tawny owls in the area (K. Wallin, personal communication).
Inverse density-dependent predation and its consequences
Dilution of predation risk for the fledglings is a likely contributing explanation for the inverse density-dependent predation (Hamilton 1971; Foster and Treherne 1981). Sparrowhawks settle more densely in areas with higher overall prey density (Newton 1986), but territorial behaviour in spring apparently imposes a lower limit to nearest nest distances (Newton 1986, Fig. 14 on p. 65) and to the total food intake, constraining the hawks’ numerical and functional response to increased prey density. Although the great tit was the most common species and a preferred prey of the hawks (Götmark and Post 1996), on average 80% of all prey were other species. A striking feature of the songbird prey community is that the species (males) arrive and establish breeding territories over a long period (almost 3 months in southern Sweden), but their young leave the nests more synchronised, during less than a month (about 25 May–20 June in the study area). This is probably mainly caused by the seasonal peak in food abundance (mainly insect larvae), and the corresponding peak in the fledgling abundance may dilute hawk predation rates. In addition, hawk predation seems to favour synchronised nesting in great tits (Götmark 2002). Another explanation for inverse density-dependent predation in the present case may be that predators at high songbird density are more likely to be detected by prey individuals, which warn other songbirds by alarm calls (Marler 1955) and thereby may limit hawk hunting success.
In our observational and experimental data sets, great tit density in hawk territories was not correlated with the density of other major prey species. Therefore, such co-variation was not a major confounding variable. We found correlations with some species [great tits and blue tits (observational data) and robins (experimental data)]. Under natural conditions, density correlations between pairs or groups of species of nesting songbirds probably occur, and total elimination of such correlations would be unrealistic (and unnatural). We do not have data on fledgling numbers of the other songbird species in hawk territories. Total fledgling numbers are almost impossible to measure in songbird community studies, especially for such large study areas as ours (500 km2).
The novel finding that sparrowhawk predation rate on a common songbird declines with increasing prey density in the breeding season lends support to several interesting hypotheses. It may help explain why songbirds in spring often settle in a clumped territorial pattern (Slagsvold 1980; Stamps 1988; Muller et al. 1997; Mönkkönen et al. 1999). Many songbirds settle close to conspecifics (Stamps 1988; Muller et al. 1997; Etterson 2003; Ward and Schlossberg 2004), and different species of songbirds associate with each other, probably partly in response to predation (Slagsvold 1980; Mönkkönen et al. 1999). Conspecific attraction in great tits establishing territories has not been investigated experimentally, but deserves more study in the light of our results. There are several alternative explanations for conspecific attraction in solitary songbirds (see Stamps 1988; Etterson 2003; Ward and Schlossberg 2004)
In a given area, increased great tit density might raise intraspecific competition for food, reducing reproductive output per breeding pair (reviewed in Gosler 1993). If so, there may be a trade-off between foraging disadvantages and anti-predator benefits for pairs that choose to breed in aggregations.
Implications for conservation
The results of this study and some others suggesting that predation rate is highest for sparse populations (Hudson 1992; Newton 1993, 1998), and for uncommon or rare species (Götmark and Post 1996; Spiller and Schoener 1998; Sinclair et al. 1998), point to the existence of a ‘predation trap’ (Newton 1993, 1998). This has implications for conservation work (see also Sinclair et al. 1998; Macdonald et al. 1999). For practical reasons, most predation studies in conservation biology have dealt with common prey species (Macdonald et al. 1999, p. 495). Our results suggest that further studies of predation in sparse populations and less common species, including endangered ones, may be crucial for clarifying the relations between prey density, predation and extinction risk. With decreasing population size, extinction probability will usually increase for several reasons (Soulé 1987). In prey populations where predation rate increases as prey density declines, this risk factor may add to the already known mechanisms threatening small, sparse populations.
We thank Martin Bergström, Jan Bergqvist, Anders Enemar, Peter Post and especially Jan Olsson for valuable help in the field, C. Askenmo, T. Bohlin, E. Korpimäki, I. Krams, I. Newton, P. Post, V. Selås, A.R.E. Sinclair and anonymous referees for comments and suggestions on the manuscript, and Dr P. Johannesson and Dr K. Wiklander (Department of Mathematical Statistics, Göteborg University) for confirming our statistical methods. The study was funded mainly by grants from the Swedish Research Council (NFR/VR) to F.G. and conducted in accordance with national laws for scientific research, including permit for ringing of birds issued by the Natural History Museum, Stockholm. We wish to thank the great tits, Irma Johansson, the land owners and Västkuststiftelsen for cooperation and help during our study.