ANIMAL BEHAVIOUR, 2000, 59, 989–999 doi:10.1006/anbe.1999.1380, available online at http://www.idealibrary.com on
Patch time allocation and patch sampling by foraging great and blue tits BEAT NAEF-DAENZER
Swiss Ornithological Institute (Received 8 March 1999; initial acceptance 29 April 1999; final acceptance 29 December 1999; MS. number: 6142R)
The rate at which parents deliver energy to their brood is an important factor in avian reproduction because poor condition caused by malnutrition may reduce the offspring’s survival to breeding. Models of central place foraging predict that nesting parents should optimize their prey delivery rate by minimizing travelling distances and by selecting patches where the gain per unit cost is high. I investigated the allocation of searching time amongst food patches in the home ranges of breeding great tits, Parus major, and blue tits P. caeruleus, by radiotracking. The density of locations in individual trees was positively correlated with prey biomass within trees and negatively with the distance of the trees from the nest. These two factors explained 52% of the variance in the allocation of the birds’ search time. In rich patches, food was reduced considerably within 20 m of the nests, and the birds’ travelling distances increased significantly during the nestling period. In parallel to foraging selectively in rich resources near the nest, the birds continually sampled the trees in their territory. The average surplus search time due to resource exploration was 1.52 times (range 1.25–1.99) the expected search time if the birds had exclusively used the most profitable patch. Despite considerable effort in patch sampling, the overall search time per unit prey was 30% better than expected by an equal use of trees. The results suggest that foraging tit parents come close to the maximum rate of prey delivery possible in a given patch distribution.
rate (as with ‘ad libitum’ food; Keller & van Noordwijk 1993; Naef-Daenzer & Keller 1999). On the other hand, collecting the large amounts of food required is costly and may affect the parents’ future condition and survival (Nur 1984a, b, 1988; Ho ˜ rak et al. 1999). This leads to a trade-off between maximizing chick growth and minimizing parental foraging costs. The timing of breeding is important in both these respects: when caterpillars are most abundant, the average search time per feeding trip is ca. 40% lower than before or after the peak (Naef-Daenzer & Keller 1999), allowing high delivery rates at relatively low costs. Therefore, there is a double premium in birds having the nestling period coincide with the peak food availability. Little is known about the foraging strategies of breeding tits in the field. The spatial distribution of prey is usually uneven and changes considerably with time (Gibb 1958; Royama 1970; Perrins 1991). Experiments in captivity have shown that foraging tits prefer rich resources and respond quickly to temporal changes in the distribution and size of prey (Smith & Dawkins 1971; Krebs et al. 1977; Ydenberg 1984; Gru ¨ nberger 1992). Field observations have shown great and blue tits return up to nine times to the same spot, often the same branch, and that
In many bird species, fledging weight affects juvenile survival (Magrath 1991 and references therein) and strongly influences the number of recruits to the breeding population (Tinbergen & Boerlijst 1990; Magrath 1991; Verboven & Visser 1998). Hence, parents should optimize their offspring’s growth rate by foraging efficiently and by breeding when food is abundant. For great tits, Parus major, and blue tits, P. caeruleus, nestling food is abundant for a very short period, and thus the timing of breeding in relation to food availability is of particular importance (e.g. Perrins 1970; van Balen 1973; Garnett 1981; Noordwijk et al. 1981; Verhulst & Tinbergen 1991). Tit parents are unable to maintain high prey delivery rates when food is scarce. Thus, the abundance of prey largely determines the energy flow to the nest and the resulting fledging weights (Gibb 1955; van Balen 1973; Gebhardt-Henrich & van Noordwijk 1991; Perrins 1991; Naef-Daenzer & Keller 1999). To optimize nestling growth and condition at fledging, the parents should achieve food delivery rates allowing nestlings to grow at about the maximum physiological Correspondence: B. Naef-Daenzer, Swiss Ornithological Institute, CH-6204 Sempach, Switzerland (email:
[email protected]). 0003–3472/00/050989+11 $35.00/0
2000 The Association for the Study of Animal Behaviour
989
2000 The Association for the Study of Animal Behaviour
990
ANIMAL BEHAVIOUR, 59, 5
the rate of prey delivery strongly depends on the size and density of caterpillars (Smith & Sweatman 1974; Naef-Daenzer & Keller 1999; Naef-Daenzer et al., in press). However, the foraging strategies and range use of tits in relation to the spatial distribution and quality of food patches have not been investigated in the field. Andersson (1978, 1981) modelled how a predator should exploit an initially uniform prey distribution when the prey is brought to a central place. The model predicts that searching effort per unit area should decrease linearly with increasing distance from the central place, the slope being steeper the higher the initial density of food. This effect is caused by the gradual depletion of the food with time. As prey availability close to the nest declines, further points offer equally good returns because the higher prey density there offsets the greater travel time. Over a fixed period of time, this process results in searching effort per unit area decreasing with increasing distance from the central place. In a patchy resource distribution, higher prey density or more profitable prey types (Lessells & Stephens 1983) compensate for greater travel costs. Under these conditions an optimal forager must decide how long to stay in a patch and where to continue its search if it leaves the patch. The marginal rate model (Charnov 1976) provides hypotheses for optimal patch residence time. It shows that an optimal forager is expected to stay in a given patch (or to return to it in the case of titmice taking single prey) until the capture rate reaches a threshold value. This threshold is equal to the average capture rate over the patches available. Since the threshold is reached less frequently in rich patches, and where travelling effort is low, the forager will allocate more of its search time to patches offering higher returns than the average (for a detailed discussion see McNair 1982; Roche 1996). The marginal value approach has also been used to model giving-up rules for different patch patterns. McNair (1982) proposed models that predict more persistent search, that is, increasing giving-up thresholds, where prey density is high. Krebs et al. (1977) showed experimentally that as well as absolute prey availability, the quality of the patch relative to others also affects the giving-up decisions and thus the time allocated to a site. Both ‘residence time’ and ‘giving-up’ models predict that a major proportion of search time should be allocated to the most profitable patches whereas poor patches should be avoided. Titmice forage in unstable patch distributions (example given in Naef-Daenzer & Keller 1999) on very cryptic prey. Thus, the quality of trees in the home range must be continually sampled. However, the more time the bird invests in sampling (poor) patches the less is its gain per unit cost. Aspects of optimal sampling of patchy distributions have been analysed theoretically (Stephens 1987) and experimentally (e.g. Kramer & Weary 1991; Kohlmann & Risenhoover 1997). However, the proportion of search time devoted to patch sampling and the consequences for an animal’s overall foraging efficiency have not been investigated in the field. To quantify the birds’ response to the patchy prey distribution, I investigated, using radiotracking, how tit
parents allocated their search time to individual trees and estimated food availability in each tree around the nests. I consider two levels of analysis: (1) how great and blue tits allocate their search time in relation to food availability and travelling distance; and (2) the costs of resource exploration to tit parents in relation to the patchiness of resources within the home ranges. METHODS I did the study in an oak-rich deciduous woodland near Basel, Switzerland. The main tree species in the ‘Birsfelder Hard’ are oak, Quercus sp. (30% of trees in the upper canopy), ash, Fraxinus excelsior (36%), beech, Fagus silvatica (18%), and hornbeam, Carpinus betulus (10%). About 300 nestboxes were placed along forest roads. For detailed descriptions of the study area see Mosimann et al. (1987) and Gebhardt-Henrich & van Noordwijk (1991).
Radiotracking Tit Parents To locate foraging great and blue tits, I used a miniature radiotransmitter. The radiotags weighed 900 and 620 mg for great and blue tits, respectively, which is ca. 5% of the birds’ body mass. For quick location of the signals, the pulse frequency of the transmitters was tuned to 4–6 Hz. The birds were caught with mist nets near the nestboxes when the nestlings were 5–7 days old. Transmitters were attached on the bird’s back using 40–50 mg of cyanoacrylate glue (Loctite Ltd, Herb, U.K.). Details of the technique and transmitter design are given in NaefDaenzer (1993a, b, 1994). All birds were released within 15 min of capture. The tits were tracked by triangulation from two fixed antenna stations placed 60–100 m from each other and from the nestbox. Null peak bearings (Kenward 1987) were taken once a minute and were denoted as successful if the bearing angle could be determined within 5 s. This resulted in 20–40 valid locations per hour. Triangulation from fixed stations requires that a bird’s position does not change for 5–10 s. Hence, flying birds could not be located and data exclude transit movements. The average accuracy of location was 1.4, that is 2.4 m at a distance of 100 m (Naef-Daenzer 1993b). Since the average diameter of tree crowns in the upper canopySD was 7.22.4 m, the spatial resolution was sufficient to estimate location densities for individual trees in the home ranges. During the 1988–1990 breeding seasons, I collected data from 12 great and seven blue tits. Because of the effort involved in capturing birds and installing receiving equipment, the tagged birds were observed as intensively as possible. Therefore, the sample is limited to a relatively few individuals from which a large number of locations was collected (for a methodological discussion see Aebischer et al. 1993). Data collection began after the marked birds restarted feeding (2–6 h after tagging) and continued as long as the transmitters remained attached (median 6 days, range 1–20 days). Birds were tracked for 4–10 h/day, the observation
NAEF-DAENZER: PATCH TIME ALLOCATION IN TITS
sessions being distributed equally over the daylight period. The departures and arrivals of the tagged birds at the nestboxes were recorded by an industrial inductive switch (i.e. an electronic device activated by the metal components of the transmitter; Baumer Electric, Frauenfeld, Switzerland). The switch was placed outside the box above the hole and triggered an acoustic signal at one of the antenna stations. After catching a prey, great and blue tits returned directly to the nest (Smith & Sweatman 1974; personal observation). Hence, the last radiolocation preceding a visit at the nest was considered as the site where a prey was found (see also Naef-Daenzer & Keller 1999). I observed no adverse effects of radiotagging on the adult birds, except extensive comfort behaviour directed to the transmitter for about a day after tagging. This resulted in the loss or destruction of the transmitter in eight tagging attempts. No injuries or skin irritations were observed in six recaptured birds. Qualitative comparisons of tagged birds with their mates did not reveal any changes in flight characteristics (e.g. speed, angle of body axis, hovering), travelling distances or feeding frequencies (details in Naef-Daenzer 1993b). Radiotags of about 5% of body mass do not significantly affect mortality, range use or behaviour of juvenile tits (B. Naef-Daenzer, F. Widmer & M. Nuber, unpublished data) and probably not of adult birds either. In three birds tracked during nest building and incubation, we found no significant relationship between the average density of locations per unit area and the distance from the nest (Y=28.51–0.08X, r29 = 0.09 NS). This indicates that the concentration of locations near the nest observed after hatching was due to the duties at the nest rather than an effect of the transmitter load on the bird’s patch choice. One male changed quickly from a homogeneous distribution of locations before hatching to the pattern shown in Fig. 2 in the Results thereafter (for details see Naef-Daenzer 1994). The radiotags fell off within 1–20 days and the median duration of the radiomarks was 6 days (Naef-Daenzer 1993b). Triangulation from remote fixed stations allowed the birds to be located frequently without observers moving across the home ranges. Since all tagged birds resumed feeding within a few hours no effect on the growth of the broods was observed. Capture and marking of the birds was approved by the Swiss Federal Department of Forestry. Transmitters and receiving equipment were licensed by the Swiss Federal PTT.
Prey Size and Distribution I estimated the availability of prey (lepidopteran and sawfly larvae, referred to as ‘caterpillars’) by collecting frass droppings with funnels of 39 cm diameter (ca. 0.125 m2, see Gibb & Betts 1963; Liebhold & Elkinton 1988; Perrins 1991). These collectors were placed below each tree of the upper canopy. Frass droppings were collected continuously from 4 April to 30 May 1989 and from 20 March to 3 June 1990. Periods of 2–3 days were sampled, the duration of each sample measured to the
nearest 1 h. In a parallel study in the same forest, the abundance of herbivorous insect larvae in selected trees was estimated from branch samples. Fischbacher et al. (1998) showed a good correlation of caterpillar biomass estimates taken from branch samples with data from frass collection below the same trees. I calculated caterpillar biomass from the equation given in Fischbacher et al. (1998). Tree crowns were assumed to be circular and I estimated their diameter in the field. To map the exact positions of trees, funnels, nestboxes and antenna stations, I used a Zeiss Elta theodolite.
Analysis The analysis of patch time allocation refers to the individual trees that were visited at least once during the observation period. Because the range use of an animal is nonrandom, the application of parametric statistics has been criticized (White & Garrott 1990 and references therein). Independence of data in a strict sense would require that only one location per individual is included, which would make it impossible to estimate any home range dimensions or resource selection. There was significant autocorrelation of consecutive locations of the tracked birds at the scale of minutes (Naef-Daenzer 1994). Despite this, interval samples provide an estimate of the proportion of activity time allocated to different places within the home range for the observation period (Altmann 1974). From the statistical point of view, it is more important that the observation period and therefore the total number of locations, is sufficient to represent the full range used by an animal (for further discussion see De Solla et al. 1999). The kernel techniques I used for estimating the density of locations in trees are nonparametric. Because the data for each individual include several days and hundreds of locations, the resulting estimates of search time allocated to individual trees are considered to be free of bias from short-term autocorrelation and were treated as independent estimates. I obtained 6305 radiolocations of great tits and 8559 of blue tits. The software Grid (Naef-Daenzer 1993a) was used to calculate the intensity of use of each individual tree in the home ranges by a bivariate kernel procedure. Reviews of kernel techniques are given in White & Garrott (1990) and Worton (1989); here I give only a brief summary of the method. To analyse the intensity of use of a site within the home range, the pattern of locations (points) must be transformed into an estimate of density at that site. Kernel-based methods return such density estimates for each intersection of a user-defined grid. The estimate is calculated by an algorithm (the kernel function) that defines which locations are included and how they are weighted. The most common technique is the bivariate normal kernel which weights locations according to a bivariate Gaussian distribution that has its maximum at the grid point being processed. The resulting matrix of density estimates can be used in the presentation of contour plots or to determine use densities at any particular location by interpolation. Based on the accuracy of bird locations, the standard deviation of the kernel function was set to 2.4 m and the estimates
991
992
ANIMAL BEHAVIOUR, 59, 5
Table 1. Summary of radiotracking data collected on great and blue tits in 1989–1990 Observation period
Locations
Brood size
Average nestling age (days)
Species
Sex
Box number
1989 Great Great Great Great Great Blue Blue
M+F M M M+F M+F M M
012 012 091 345 345 373 373
24–28 May 29 May–3 June 3–7 May 16–18 May 20–22 May 30 April–5 May 6–11 May
700 866 610 427 285 1144 620
6 6 8 10 10 12 12
6 11 9 8 12 9 16
1990 Blue Great Blue Blue Great Great Blue Blue
M M M M M M F F
016 017 371 371 372 372 373 373
29 April–3 May 10–16 May 4–10 May 11–16 May 16–20 May 21–26 May 21–25 May 26–31 May
709 587 919 879 533 629 863 727
4 9 7 7 9 8 8 8
3 11 8 13 9 14 8 14
To account for the great seasonal changes in food availability, data from six broods covering more than 7 tracking days were split into two tracking periods each.
were computed with a 1-m2 grid of 150150 cells. The resulting use densities at a point represent the weighted number of locations recorded within a radius of 2.4 m, or approximately 20 m2. To account for different sample sizes for individual birds, location densities in individual trees were expressed as a percentage of the kernel density recorded at the nest site. Although the home ranges of neighbours overlapped to some extent, the sites of intense foraging were used exclusively by one pair (Naef-Daenzer 1994). Thus, the analyses assume no significant competition by direct interference between birds (neither within nor between species). Because caterpillar density and size change quickly during the season, long series of tracking data were split into subsamples (tracking periods) of 3–6 days, each consisting of 285–1144 locations. To avoid bias by individuals observed for relatively long periods, only two tracking periods per individual were included in the analysis. A total of 15 tracking periods including more than 250 locations each were analysed (Table 1). I determined patch quality from 2990 frass samples from 601 individual trees. Naef-Daenzer & Keller (1999) have shown that the search time per feeding trip in relation to caterpillar biomass follows a power function. Accordingly, I expressed the profitability of the trees as the expected search time per feeding trip, using the equation Y=9.5X 0.27, where X is the caterpillar biomass in the respective tree in mg/m of branches (Figure 4 in Naef-Daenzer & Keller 1999). I excluded 22 outliers with frass dropping rates exceeding 500% of the daily average (1.7% of the sample), as the previous/ following samples suggested errors. Additional data on brood size, age of nestlings and growth were provided by Gebhardt-Henrich & van Noordwijk (1991).
RESULTS
Prey Distribution Figure 1 gives an example of the distribution of prey biomass and the use of trees by tit parents. Caterpillars were patchily distributed in all home ranges, although the between-tree variation was not high compared with the seasonal changes. The interquartile range of food availability in individual trees was between 49.1 and 130.0% of the average caterpillar density at a given date. Thus, about 25% of trees offered prey densities that were markedly above the average. The seasonal changes in the biomass available were considerably larger. During the caterpillar peak (first half of May), the average prey biomass was 110 mg/m of branches, which is about six times higher than late in the season (15 mg/m, mid-June). Prey density in oaks was considerably higher than in other tree species (ash, beech, hornbeam, see below). The average number of treesSD visited at least once per tracking period was 6013 (N=15 tracking periods), which indicates the total number of trees that were available in the home ranges. However, only 197 trees had more than five visits per tracking period. Therefore, only a selection of the available trees was used for foraging. Within 10 m of the nest, trees of all species had high location densities. At 10–50 m from the nest, however, the average location density on oak trees was significantly higher than in beech, ash or hornbeam (ANOVA: F3,781 =9.68, P
