Opposing selective pressures on hatching asynchrony: egg viability, brood reduction, and nestling growth

Transcription

1 Behav Ecol Sociobiol (2000) 48: Springer-Verlag 2000 ORIGINAL ARTICLE Javier Viñuela Opposing selective pressures on hatching asynchrony: egg viability, brood reduction, and nestling growth Received: 25 October 1999 / Revised: 30 May 2000 / Accepted: 25 June 2000 Abstract At least 19 hypotheses have been proposed to explain the evolutionary significance of avian hatching asynchrony, and hatching patterns have been suggested to be the result of several simultaneous selective pressures. Hatching asynchrony was experimentally modified in the black kite Milvus migrans by manipulating the onset of incubation during the laying period. Delayed onset of incubation reduced egg viability of first-laid eggs, especially when ambient temperature during the laying period was high. Brood reduction (nestling mortality by starvation or siblicide) was more commonly observed in asynchronous nests. The growth rate was slower in synchronous broods, probably due to stronger sibling rivalry in broods with high size symmetry. Lasthatched chicks in synchronous broods fledged at a small size/mass, while in control broods, hatching order affected growth rates, but not final size. Brood reduction, variable growth rates, and the ability to face long periods of food scarcity are probably mechanisms to adjust productivity to stochastic food availability in a highly opportunistic predator. The natural pattern of hatching asynchrony may be the consequence of opposing selective forces. Extreme hatching synchrony is associated with slow growth rates, small final size of last-hatched chicks, and low viability of first-laid eggs, while extreme hatching asynchrony is associated with high mortality rates. Females seem to facultatively manipulate the degree of hatching asynchrony according to those pressures, because hatching asynchrony of control clutches was positively correlated with temperature during laying, and Communicated by W.A. Searcy J. Viñuela ( ) Departamento de Ecología Evolutiva, Museo Nacional de Ciencias Naturales (CSIC), José Gutierrez Abascal 2, Madrid, Spain Present address: J. Viñuela, Instituto de Investigación en Recursos Cinegéticos (CSIC), Libertad 7, Ciudad Real, Spain, jvinuela@irec.csic.es Tel.: , Fax: negatively correlated with the rate of rabbit consumption. Key words Hatching asynchrony Egg viability Brood reduction Nestling mortality Nestling growth Milvus migrans Introduction Hatching asynchrony, a widespread trait in birds (Clark and Wilson 1981; Stoleson and Beissinger 1995), promotes the establishment of intra-brood size hierarchies (Stokland and Amundsen 1988; Viñuela 1996) that may have an adaptive value: when food availability is not sufficient to raise all brood members, the smaller lasthatched chicks will starve quickly, while the larger siblings may survive (brood reduction hypothesis; Lack 1954). If these size hierarchies did not exist, more chicks per brood would die. However, few studies have successfully tested the brood reduction hypothesis (but see Magrath 1989; Hébert 1993), and brood reduction has been proposed to be, at least partially, a non-adaptive consequence of hatching asynchrony that might be maintained for other reasons (Amundsen and Slagsvold 1991a). As many as 18 other hypotheses have been proposed to explain asynchronous hatching (Stoleson and Beissinger 1995; see also Veiga 1993; Slagsvold et al. 1995). Hatching asynchrony could be an adaptive trait maximising breeding success due to the establishment of a size hierarchy, or the pattern of incubation initiation per se could be selectively favoured, rather than the consequent size hierarchy (Stoleson and Beissinger 1995). For example, hatching asynchrony may be a consequence of birds starting incubation before the end of laying to avoid loss of viability of first-laid eggs ( egg viability hypothesis ; Arnold et al. 1987; Veiga 1992). That loss of viability may be determined by the thermal conditions to which the eggs are exposed before incubation starts (Webb 1987). The exposure of unincubated eggs to tem-

2 334 peratures below the optimal incubation temperature (i.e C; Webb 1987), but above the physiological zero (the temperature at which embryo development starts, C; Webb 1987) may be more deleterious than exposure to temperatures below physiological zero (Webb 1987; Veiga 1992). Consequently, birds should advance the onset of incubation (and thus increase hatching asynchrony) when laying occurs during periods of high ambient temperatures (Veiga and Viñuela 1993; Stoleson and Beissinger 1999). To date, most studies have tested the brood reduction hypothesis, while the possible selective forces affecting the onset of incubation have rarely been considered (Clark and Wilson 1981; Amundsen and Slagsvold 1991a; Stoleson and Beissinger 1995). There is, however, increasing awareness that a single hypothesis cannot explain inter- or intra-specific variation in hatching asynchrony (Clark and Wilson 1981; Stoleson and Beissinger 1995; Viñuela and Carrascal 1999). Some authors have suggested that several selective pressures could act simultaneously to shape the optimal level of hatching asynchrony within a species (Bollinger et al. 1990; Amundsen and Slagsvold 1991b; Veiga and Viñuela 1993; Stoleson and Beissinger 1995), but the existence of those opposing selective forces acting simultaneously on the same population has rarely been reported. Furthermore, very few studies have examined the correlates of natural variation in hatching patterns, which may be a key factor to understanding its significance (Wiebe 1995). Hatching asynchrony could be facultatively manipulated according to food availability, when this can be predicted at the time of laying (increasing/ decreasing hatching asynchrony when food availability is low/high, respectively; Wiebe and Bortolotti 1994a). The onset of incubation could also be manipulated depending on weather conditions during the laying period (advancing/delaying the onset of incubation when weather is unfavourable/favourable to preserve egg viability of unincubated eggs; Veiga 1992; Veiga and Viñuela 1993). Siblicide, brood reduction and asynchronous hatching are common in raptors, and the brood reduction hypothesis is often assumed to explain hatching asynchrony in this group (Newton 1979, Stinson 1979), although there are not enough detailed studies to prove it (but see Wiebe and Bortolotti 1995). Hatching asynchrony is greater in raptor species with obligate siblicide than in those with facultative siblicide (Meyburg 1974; Edwards and Collopy 1983). By contrast, little is known for raptors about intraspecific variation in hatching asynchrony (Newton 1979). Wiebe and Bortolotti (1994a, 1995) found in the American kestrel (Falco sparverius) that synchronous hatching maximised nestling survival when food availability was high, while hatching asynchrony allowed earlier brood reduction, and thus reduced parental investment, when food availability was low (see also Gibbons 1987). In the same species, hatching asynchrony optimised nestling growth by reducing the energetic cost of competition between siblings (Wiebe and Bortolotti 1994b; see also Slagsvold et al. 1995). The egg viability hypothesis has not been tested in raptors, although hatching failures are an important source of offspring mortality in this group (Newton 1979). In this study, the relationships between hatching asynchrony and egg hatchability, brood reduction, and nestling growth were explored in a medium-sized raptor, the black kite (Milvus migrans). In the study area, temperatures higher than C are common during the breeding season, providing a good opportunity to test the egg viability hypothesis. Food-dependent brood reduction and hatching-order-dependent growth rates and mortality are common in black kites (Viñuela and Veiga 1992; Viñuela 1997a, 1999). Consequently, this is also a suitable species in which to study the effect of hatching asynchrony on brood reduction and growth. Black kites are a paradigm of a generalist predator (Brown and Amadon 1968; Cramp 1980), but breeding success in this population of black kites is affected by the rate of consumption of rabbits (Viñuela and Veiga 1992). This is in turn affected by the location of nesting territories with respect to rabbit warrens (Viñuela et al. 1994). Thus, rabbits are a key prey affecting breeding success, and their consumption rate may be at least partially predicted before laying. Both conditions seem necessary for the existence of facultative variation of hatching asynchrony in relation to prevailing food levels (Wiebe 1995). Methods The study was conducted in Matas Gordas, Northern Doñana National Park (south-west Spain, 37 N 6 5 W). This is an open and flat area of Mediterranean forest (cork oaks Quercus suber), scrublands and grasslands, close to a seasonally inundated marshland (see Viñuela and Veiga 1992; Viñuela 1993). The breeding area was visited almost daily from the start (mid March) to the end (end of July) of the breeding seasons Before laying, nests were visited every 2 8 days, depending on the building condition during the previous visit. Advanced nests with a well-built lining (Desai and Malhotra 1979; Viñuela 1993) were checked every 1 2 days. After the first egg appeared, nests were visited daily throughout laying, and the eggs were individually marked with felt pens. To minimise the exposure of eggs or chicks to extreme thermal conditions during visits, the nests were not visited during rainy weather, early in the morning or at midday. Laying order, hatching success and hatching asynchrony Laying was considered finished when the third egg was found or if 4 days had elapsed since the second was found. When already completed two-egg (n=19) or three-egg (n=5) clutches were found, the egg with a cleaner and brighter colour was considered the last one laid (Desai and Malhotra 1979; details in Viñuela 1997a). In analyses of hatching success, all eggs lost by causes other than hatching failure (predation, nest collapse or clutch desertion) were excluded. After laying, nests were not visited again until days after the laying date of the first egg. Nests were visited daily during hatching to record the hatching condition of each egg or recently hatched chick (see Viñuela 1996). Hatching asynchrony was estimated as the hours elapsed between the estimated hatching times for the first and last chicks in a brood. A variation of the method of Stokland and Amundsen (1988) was followed to estimate hatching asynchronies in hours from daily visits (details in Viñuela 1996).

3 335 During 1988 and 1989, hatching asynchrony of 52 randomly selected clutches was experimentally altered by removing the eggs as they were laid and replacing them with hen eggs artificially marked to mimic the natural pigmentation of kite eggs. From these nests, the first two eggs were taken on the days they were found, and were held until 3 days after the second egg was laid, when they were returned to the nest. Third-laid eggs were marked when found, but not removed. Experimental clutches were alternatively assigned to one of two treatments: (1) asynchronous clutches had eggs maintained during the removal period in incubators at C, with water containers below the eggs (Burnham 1978, 1983; Campbell and Flood 1977), and (2) synchronous clutches were not incubated, but maintained at ambient temperature during the removal period, outside the same building with the incubators (1 3 km from nests), and inside a cardboard box in the shade. Eggs in both treatments were turned 180 twice daily. Hereafter, experimental nests are designated asynchronous and synchronous, while nests with unaltered hatching asynchrony are designated controls. To explore the possible influence of weather during laying on hatching asynchrony of control nests, the averages of maximum daily temperatures (variable "temperature") and rainfall (variable "rainfall") during the time elapsed between, and including, the days of laying of the first and last eggs, were considered. The same variables were used to examine the possible effect of weather on hatching success in synchronous clutches, considering the period in which the eggs were kept unincubated. Meteorological data were obtained from an Instituto Nacional de Meteorología station located 2 5 km from the study nests. Maximum temperatures were selected because minimum spring temperatures in the study area are usually around the optimal temperature to preserve the viability of unincubated eggs (about 10 C; Olsen and Haynes 1948; Campbell and Flood 1977). In contrast, maximum daily temperatures are often higher than the physiological zero (24 28 C; Webb 1987). Nestling mortality and growth Nests were visited 2 3 days after the day of hatching of last chicks and every 4 7 days (mostly 5 days) thereafter until the end of the nestling period. During every visit, chicks were re-marked and mortality recorded. All nests in which there was nestling mortality due to causes other than starvation or siblicide (predation, abandonment, and nest collapse) were excluded from the analyses of nestling mortality. However, three cases in which broods were depredated near the end of the nestling period, well beyond the maximum age of death by starvation or siblicide recorded during the 3 study years, were considered as nests without starvation or siblicide mortality. Chicks of the genus Milvus have a high resistance to food deprivation, often suffering severe growth delays due to protracted periods of food scarcity (Hiraldo et al. 1990; Veiga and Hiraldo 1990; Viñuela and Veiga 1992; Viñuela and Ferrer 1997). Thus, mortality due to starvation was easy to detect, as the chicks suffered evident retarded growth and were extremely emaciated before death. Sibling aggression was common in this population (Viñuela 1999), and in most chicks with starvation symptoms, evidence of aggression was also recorded (sibling fights and/or wounds on the chicks). In 6 out of the 17 cases assigned to siblicide, dead chicks had starvation symptoms during the previous nest check (Viñuela 1999). Thus, siblicide and starvation were often associated (see also Newton 1979). Because starvation as well as siblicide could be related to hatching asynchrony, data of both kinds of deaths have been pooled. On every visit after hatching, chicks were weighed with Pesola balances with an error of less than 1%, and tarsus length was measured with Vernier callipers to the nearest 0.1 mm. Growth rates (K) for each chick were estimated by fitting growth curves using a non-linear least-squares regression routine (see Viñuela and Bustamante 1992; Viñuela 1997a). Final mass and tarsus length were considered to have been reached when the seventh primary was 180 mm long (fully feathered chicks; Viñuela and Veiga 1992). As feather growth at this time is linear (Hiraldo et al. 1990), the age when the seventh primary reached 180 mm could be inferred by interpolation between preceding and subsequent measures (Viñuela and Bustamante 1992). Fledging weight at that age was also calculated by interpolation between measures. At this age, weight gain has finished, and only a slight and variable weight recession may occur (Hiraldo et al. 1990). Growth rate, the duration of the nestling period and the form of the growth curve may vary between sexes in raptors with sexual size dimorphism (Bortolotti 1986b). However, black kites show only slight reversed sexual size dimorphism (Cramp 1980), and nestlings cannot be sexed by morphological measurements (Viñuela and Bustamante 1992). In this kind of species, we should not expect to find strong effects of sexual dimorphism on sibling competition (see e.g. Schaadt and Bird 1993). Other variables Brood size at hatching varied between one and three chicks. Nests in which only one chick hatched (n=27) were not considered. In the analyses of mortality and growth, two-chick broods hatched from two- or three-egg clutches were distinguished. In these latter analyses, hatching order was considered instead of laying order (in most synchronous nests hatching occurred in reverse order to that of laying; Viñuela 1997b). Laying date is defined as the estimated day of laying of the first egg in a clutch, assuming that the laying interval between the first and second egg is 3 days (which is the usual period between second- and third-laid eggs; Viñuela 1999). The median of the frequency distribution of laying dates during the 3 study years (April 15) has been used to classify the nests as early (laying on or before the median date) or late (laying after median date). The difference between the mass of the first- and last-hatched chicks on the day hatching was completed ("chick mass hierarchy") was calculated for each brood. This difference was clearly correlated with hatching asynchrony (Viñuela 1996). The residuals from the regression of chick mass hierarchy on hatching asynchrony were in turn strongly correlated with mass increase of firsthatched chicks during the hatching period (Viñuela 1996). Consequently, these residuals (hereafter size hierarchy residuals ) are estimates of the growth rate of first chicks during the hatching period, independent of hatching asynchrony. These residuals have been classified as positive values, representing rapid growth of first-hatched chicks during the hatching period, or negative values, indicating first chicks that did not hold their full advantage. From 1987 to 1989, the population of black kites in the study area steadily increased due to the settlement of new pairs, apparently inexperienced individuals (see Viñuela 1993, 1997a; Viñuela et al. 1994). These new pairs appearing during the study period were considered inexperienced (first and second year in the area), as opposed to pairs with more than 2 years breeding in the local area. Prey remains in nests were recorded on every nest check as described in Viñuela and Veiga (1992) and Viñuela et al. (1994). The mean number of rabbits found per nest visit ("rabbit consumption") was calculated for each nest, and this variable was used as estimator of the relative importance of rabbit in the diet (Viñuela and Veiga 1992; Viñuela et al. 1994). Statistical analyses Statistical analyses were performed with SPSS (Norušis 1992). Results are presented as means±sd, unless otherwise indicated. To estimate the relative importance of brood size, year, parental experience, laying date, size hierarchy residuals, rabbit consumption and hatching asynchrony in determining the probability of brood reduction, logistic regression analyses were performed, using a forward stepwise selection procedure based on the likelihood-ratio test (Norušis 1992; Bollinger 1994; Viñuela 1997a).

4 336 Nestling mortality was included as the dependent variable, after pooling the three cases of two chicks dead in a nest with the onechick-dead cases. This grouping should not bias the results with respect to hatching asynchrony, because each category of hatching asynchrony had a single case (Fig. 2). Nevertheless, the analyses were also repeated after removing these three cases from the sample. All independent variables were considered as categorical, except rabbit consumption. The interaction between year and hatching asynchrony was considered (an important prediction of the brood reduction hypothesis is that synchronous nests should have higher mortality in "bad" years). To test the effect of the experimental treatment on growth, ANOVAs were performed, considering growth variables as dependent variables and treatment as the main factor. The effect of other variables potentially affecting growth (Hiraldo et al. 1990; Veiga and Hiraldo 1990; Viñuela and Veiga 1992) was also examined by means of linear regression (laying date) or ANOVA (brood size, parental experience, hatching order and year). When more than one variable showed a significant effect on growth variables, ANCOVA analyses were performed, including only the interactions with treatment, due to relatively low sample size (n=140 chicks). Non-significant interactions were removed to increase the power of the test. Averages for the treatment groups were considered instead of averages per brood (Amundsen and Slagsvold 1991a; Wiebe and Bortolotti 1995). To study the possible effects of clutch size, year, parental experience, laying date, temperature, rainfall, and rabbit consumption rate on hatching asynchrony of control nests, one-way ANOVAs and linear regressions were used. When more than one variable showed significant effects, multiple-regression analyses were performed. Results Hatching asynchrony, clutch size, and experimental treatment Hatching asynchrony in three-egg control clutches (81.6±19.2 h, range =22 103, n=24) was larger (AN- OVA; F 1,45 =31.8, P<0.001) and less variable (F 22,23 =2.4, P<0.05) than in two-egg control clutches (44.7±25.3 h, range=8 92, n=23). Hatching asynchrony of control nests was intermediate between that of experimental synchronous and asynchronous nests, the differences being highly significant (two-chick broods: F 2,52 =15.77, P<0.001; three-chick broods: F 2,28 =15.6, P<0.001; Table 1). Hatching success To compare hatching success between control and experimental nests, data from 1987 were excluded, as there were no experimental nests that year, and there was some yearly variation in this variable (Viñuela and Sunyer 1992). Overall, proportions of unhatched eggs were not different between control and experimental nests (G 2 =2.26, P>0.30; Fig. 1). The percentage of unhatched first eggs was higher in synchronous than in control nests (G 1 =5.1, P=0.02), but there were no significant differences between asynchronous and control nests (G 1 =0.08, P=0.78; Fig. 1). There were no differences in hatching success between treatments for second- (G 1 =0.18, P>0.70) or third-laid eggs (G 1 =2.34, P>0.3; Fig. 1). Table 1 Hatching asynchrony (h) in black kite broods of two and three chicks. Variation between control nests (natural hatching patterns) and experimental nests (hatching asynchronised or synchronised by altering incubation pattern during the laying period). Sample sizes in parentheses Clutch size The averages of maximum daily temperatures during the period when the first eggs of synchronous clutches were maintained unincubated were higher for the eggs that did not hatch (21.2±2.2 C, n=10) than for those that hatched (19.2±2.1 C, n=12; t=2.1, P=0.05). Brood reduction Treatment Asynchronous Control Synchronous Two eggs 82.4±33.1 (11) 53.9±23.8 (35) 19.8±14.9 (9) Three eggs 108.7±13.7 (4) 89.7±20.9 (23) 34.5±21.4 (4) Fig. 1 Hatching failures during 1988 and 1989 in control and experimentally asynchronized and synchronized clutches, classified according to laying order. Percentages of unhatched eggs within each category (sample size above bars). Eggs lost due to causes other than hatching failure were excluded Brood reduction (mortality by starvation/siblicide) was the main source of mortality in this population (61% of lost chicks; Viñuela 1991). It affected mainly lasthatched chicks (78% of deaths; Table 2) and, in most cases, only one chick per brood (Fig. 2). Parental experience, brood size, rabbit consumption, size hierarchy residuals, and hatching asynchrony were included as significant variables in a forward stepwise logistic regression model with brood reduction as dependent variable (Fig. 2, Table 3). Brood reduction was more frequent in broods of inexperienced pairs relative to experienced pairs (Fig. 2). Two-chick broods coming from three-egg clutches had the lowest frequency of brood reduction, while three-chick broods had the highest (Fig. 2). The rate of consumption of rabbits and size hierarchy residuals were negatively correlated with brood reduction (Table 3). Brood reduction was more common in asynchronous broods (Fig. 2). This model classified 81% of the cases correctly (χ 2 7 =43.9, P<0.001). Laying date,

5 year and the interaction year hatching asynchrony did not significantly affect brood reduction (χ 2 1 =2.3, P=0.13; χ 2 2 =1.5, P=0.47; χ2 3 =2.2, P=0.52, respectively). This analysis was repeated after removing the three cases of broods where two chicks died (Fig. 2), and provided the same results, with only slight changes in P and R values (χ 2 7 =40.4, P<0.001; 80% of cases correctly classified). Excluding the data from 1987 (there were no experimental nests in that year), the results were similar, but in this case, the contribution of hatching asynchrony was highly significant (χ 2 2 =9.6, P=0.008, R=0.17). Table 2 Mortality by starvation or siblicide classified according to hatching order and brood size. Data for 41 dead chicks in 38 multiple broods (three cases of three-chick broods in which two chicks died) Brood size Hatching order 1st 2nd 3rd Two chicks 3 17 Three chicks Nestling growth 337 Except for tarsus growth rate, hatching order had the strongest effect on growth variables (Table 4). First chicks had the largest final size and the fastest growth, while third chicks were the lightest/smallest at fledging and had the slowest growth. In synchronous nests, hatching order affected final size of chicks [final mass (FM): F 2,19 =6.6, P=0.006; final tarsus (FT): F 2,19 =6.8, P=0.006], but not growth rates [mass growth rate (KM): F 2,19 =0.5, P=0.62; tarsus growth rate (KT): F 2,19 =1.2, P=0.32]. In contrast, in control nests, growth rates were more affected by hatching order (KM: F 2,98 =5.2, P=0.007; KT: F 2,98 =4.3, P=0.02) than final size (FM: F 2,98 =0.2, P=0.79; FT: F 2,98 =0.39, P=0.68). Hatching order did not affect any of the growth variables in asynchronous nests (KM: F 1,15 =2.9, P=0.11; KT: F 1,15 =0.5, P=0.48; FM: F 1,15 =0.8, P=0.39; FT: F 1,15 =1.6, P=0.23). However, no three-chick broods were raised in the asynchronous treatment group (due to the high incidence of brood reduction reported above), so third chicks (those suffering more strongly from growth delays/small final size; Table 4) are not represented in this treatment. Fig. 2 Brood reduction (starvation/siblicide mortality) in the black kite (white bars broods without mortality, hatched bars broods in which one chick died, black bars broods in which two chicks died). Broods classified according to parental experience (1st or 2nd breeding year in the area vs 3 or more years) (a), brood size (two chicks hatched from three eggs, two chicks from two eggs, and three chicks) (b), laying date (early vs late laying pairs, classified by the median of the distribution of laying dates during 3 years) (c), size hierarchy residuals (positive and negative, indicating a faster or slower growth of the first-hatched chick than that predicted by hatching asynchrony; see Methods) (d), year (e), and hatching asynchrony (experimental synchronous, experimental asynchronous, and control nests) (f)

6 338 Table 3 Forward stepwise logistic regression model with nestling mortality as dependent variable (0=no mortality, 1=one or two chicks dead from brood reduction). All the variables in Fig. 2 were included as categorical independent variables and rabbit consumption (mean number of rabbits found per nest visit) as a continuos variable. Laying date, year and the interaction year hatching asynchrony did not enter the model. Hatching asynchrony and brood size have two parameters in the equation, because there are three brood categories (asynchronous=a, control=c, and synchronous=s; two chicks from two-egg clutches, two chicks from threeegg clutches, and three chicks) (Parameter coding: Parental experience experienced=1, inexperienced= 1; Brood size (1) two chicks from two eggs=1, two chicks from three eggs=0, three chicks= 1; (2) two chicks from two eggs=0, two chicks from three eggs= 1, three chicks=1; Hierarchy residuals positive=1, negative= 1; Hatching asynchrony (1) A=1, C=0, S= 1; (2) A=0, C=1, S= 1 Variable Coefficient±SE χ 2 df P R Parental experience 1.51± < Brood size (1) 0.76± (2) 1.65± Rabbit consumption 1.07± Hierarchy residuals 0.82± Hatching asynchrony (1) 1.61± (2) 0.31± Constant 2.06±0.78 Table 4 Final size and growth rates (K) of mass and tarsus in chicks of black kites (only broods where more than one chick hatched). Averages with different superscripts were significantly different (ANOVAs and Bonferroni post hoc test: *P<0.01, **P<0.001) (A asynchronous nests, S synchronous nests, C control nests) Growth variable Mass (g) Tarsus (mm) Final K Final K Hatching order * ** ** 1 (70) 718±78 a 0.181±0.03 a 57.1±1.7 a 0.142± (58) 723±69 ab 0.165±0.03 b 56.5±1.7 ab 0.136± (12) 676±98 b 0.161±0.02 ab 55.2±1.4 b 0.135±0.01 Hatching asynchrony * A (17) 716.3± ±0.01 ab 57.4± ±0.02 C (101) 714.3± ±0.03 a 56.5± ±0.02 S (22) 726.4± ±0.03 b 56.9± ±0.01 Chicks from synchronous nests had slower mass growth than chicks from control nests (ANOVA: F 2,137 =4, P=0.02; Table 4). Mass growth rate was also negatively correlated with laying date (linear regression: F 1,138 =12, P<0.001). ANCOVA analyses revealed significant and independent effects of laying date (F 1,134 =23.4, P<0.001), hatching order (F 2,134 =10.8, P<0.001) and treatment (F 2,134 =5.4, P=0.006). Other growth variables were not significantly affected by the experimental treatment (Table 4). Natural variation in hatching asynchrony Given the striking variation in hatching asynchrony with clutch size, analyses were performed distinguishing between two- and three-egg clutches. Hatching asynchrony of two-egg control clutches was positively correlated with temperature during laying (Fig. 3), and marginally correlated with laying date (r=0.4, n=23, P=0.07). No other factor significantly affected hatching asynchrony of two-egg clutches (year: F 2,20 =0.5, P=0.63; parental experience: F 1,21 =1.4, P=0.24; rainfall during laying: r= 0.3, n=23, P=0.14; rabbit consumption: r= 0.01, n=23, P=0.96). Multiple regression revealed a significant effect of temperature during laying on hatching asynchrony of two-egg clutches (t=2.6, P=0.02), while the effect of laying date was not significant (t=1, P=0.34). Thus, the effect of laying date must be a consequence of the temperature increase as the breeding season advances. Hatching asynchrony of three-egg clutches was negatively correlated with rabbit consumption (Fig. 3). No other factor showed a significant effect on hatching asynchrony of three-egg clutches (year: F 2,21 =0.3, P=0.72; parental experience: F 1,22 =0.1, P=0.81; temperature during laying: r=0.4, n=24, P=0.08; rainfall during laying: r= 0.2, n=24, P=0.25; laying date: r= 0.03, n=24, P=0.9). Discussion Hatching asynchrony and hatching success First-laid eggs in experimental synchronous clutches, but not second-laid ones, had low hatching success. Those experimental unsuccessful first-laid eggs were exposed

7 339 to higher ambient temperatures than successful ones during the laying period. A lower hatching success of firstlaid eggs in three-egg control clutches has already been reported in this species (Viñuela 1997a). These results suggest that there is a physiological limit to the time an egg may be maintained unincubated, because first synchronous eggs remained unincubated 1 4 days longer than second ones. Furthermore, these results support the conclusion that the time an egg may remain unincubated depends, in turn, on environmental conditions, temperatures higher than physiological zero being especially deleterious, as predicted by the egg viability hypothesis (Arnold et al. 1987; Veiga 1992; Veiga and Viñuela 1993; Stoleson and Beissinger 1999). It could be argued that the first synchronous eggs, removed from the nests and maintained unincubated for 4 7 days, had already started embryological development and, consequently, were more sensitive to the effects of exposure (Webb 1987). However, patterns of hatching asynchrony and incubation in control nests indicated that incubation does not start with laying of first eggs (Viñuela 1991). Furthermore, hatching success of second eggs in synchronous clutches was not different from that of second eggs in control or asynchronous clutches, although second eggs had probably started embryonic development when they were removed from nests, since incubation seems to start shortly before or on the day of laying of second eggs (Viñuela 1991). Similar results have been found for wildfowl (Arnold et al. 1987), house sparrows (Passer domesticus; Veiga 1992), and green-rumped parrotlets (Forpus passerinus; Stoleson and Beissinger 1999). That this relationship between hatching asynchrony and egg viability has been noted for species in such different avian groups is interesting and suggests that this may be a widespread selective pressure in birds (Viñuela and Carrascal 1999). Factors related to brood reduction Fig. 3 Linear regressions of hatching asynchrony in two-egg clutches on average maximum temperatures during the laying period (a) and hatching asynchrony in three-egg clutches on rate of rabbit consumption during the nestling period (b) As observed for most of the variables related to breeding performance in the black kite (Viñuela 1991, 1993, 1997a; Viñuela and Sunyer 1992), and as reported for other species of raptors (e.g. Newton 1979, 1986; Poole 1989), parental experience had an important effect on nestling mortality. Late-laying pairs also experienced high mortality rates (Fig. 2), but the effect of laying date disappeared when parental experience was simultaneously considered (inexperienced pairs were usually late breeders; Viñuela 1991, 1993). Brood reduction was more common in broods where the growth rate of first chicks during the hatching period was relatively slow (negative size hierarchy residuals; Fig. 2). At least three factors could explain this result: (1) hungry chicks are more aggressive, and hunger may increase the probability of siblicide (Newton 1977; Mock 1984; Mock et al. 1987; Viñuela 1999); (2) slow early growth of senior chicks may indicate a low prey delivery rate by the male (the main provider of food during the early nestling period; Cramp 1980), and this may force long female absences from the nests facilitating siblicide (Newton 1977), and (3) slow growth of senior chicks at an age of low food demand would indicate conditions of extreme food scarcity. This result suggests that brood reduction in the black kite may function to adjust brood size to food availability (Lack 1954, 1968; Stinson 1980; Hagan 1986; Hiraldo et al. 1990; Veiga and Hiraldo 1990; Viñuela and Veiga 1992). Two-chick broods coming from three-egg clutches had the lowest frequency of brood reduction, while three-chick broods had the highest. Four-chick broods are very rare in this species (Fiuczynski and Wendland 1968), and three chicks is the maximum brood size at fledging recorded in my study area (Veiga and Hiraldo 1990; Viñuela 1991). An increase in nestling mortality with brood size has been observed in several species of birds (e.g. Klomp 1970; Greigh-Smith 1985; Shaw 1985; Skagen 1987; Steidl and Griffin 1991), but not in others (e.g. Richter 1984; Lessells 1986; Moreno 1987; Briskie and Sealy 1989; Veiga 1990). These contrasting results may reflect the existence of a continuum from species with high variability in clutch size, which can be adjusted to the individual ability to gather food or to environmental conditions (e.g. Bryant 1978; Moreno 1987;

8 340 Slagsvold and Lifjeld 1988), to species in which clutch size is less variable and brood reduction is the main mechanism for adjusting brood size to food availability (e.g. O'Connor 1978; Poole 1982; Pettifor et al. 1988; Briskie and Sealy 1989; Stouffer and Powers 1990). The rate of consumption of the main prey for black kites in my study area, the European rabbit, had an important effect on nestling mortality, as previously reported (Viñuela and Veiga 1992). Year did not have a clear effect on brood reduction when all variables were simultaneously considered in a multivariate analysis. If yearto-year variation in breeding performance is mainly explained by changes in rabbit consumption (Viñuela and Veiga 1992), after removing the effect of rabbit consumption on mortality the year would no longer be expected to have an effect, as has been found. Hatching asynchrony and brood reduction Experimental alteration of hatching asynchrony affected the frequency of brood reduction. The results suggest that high hatching asynchrony increased mortality rates, while no increased mortality was detected in synchronous nests. Most studies testing the brood reduction hypothesis either have not detected an effect of hatching asynchrony on productivity or, more frequently, have reported lower productivity of asynchronous broods (Amundsen and Slagsvold 1991a; Stoleson and Beissinger 1995). Lack (1954) argued that when food availability is insufficient to raise all the brood, early mortality of chicks would be favourable, because it would reduce the waste of resources (Gibbons 1987; Wiebe and Bortolotti 1995). However, this argument could not be valid for a species like the black kite. Black kites are opportunistic predators that exploit temporary situations of overabundance of easy prey (Viñuela and Veiga 1992; Viñuela 1999). As a consequence of this feeding strategy, food delivery rates to the nest are very irregular, periods of severe food shortage alternating with others of overabundance of prey (accumulated unconsumed prey rooting in the nests). Thus, a period of food scarcity may not be indicative of the potential for raising all the brood to fledging. In fact, 24 chicks with starvation symptoms and delayed growth were able to survive and fledge after growing at a very low rate or reaching a low fledging mass (Viñuela 1991; see also Hiraldo et al. 1990; and Veiga and Hiraldo 1990). Furthermore, the corpses of dead chicks quickly disappeared from nests, and, n nine occasions, these were confirmed to have been consumed by the surviving chicks. This consumption of dead chicks would allow a partial recovery of past parental investment, which consequently would not be completely wasted (see Alexander 1974). Hatching asynchrony and growth Last-hatched chicks fledged at a small size and mass, especially in synchronous broods. Hatching asynchrony facilitates the development of size hierarchies, but these may also develop through differential growth even in the absence of initial size differences (e.g. Pinkowski 1975; Poole 1982; Bancroft 1985; Moreno 1987; Skagen 1987; Smith et al. 1989; Amundsen and Slagsvold 1991b). For the establishment of size hierarchies, early fights between siblings may be decisive (Poole 1982; Anderson 1989) or, as argued by Amundsen and Slagsvold (1991b), slight differences in physical condition or behaviour, or simply chance, may also promote the establishment of hierarchies in synchronous broods. Mass growth was slower in synchronous than in control broods. In addition, last-hatched chicks in synchronous broods fledged with small size/mass, while in control broods, hatching order affected growth rates, but not final size. In synchronous broods, the absence of clear size hierarchies just after hatching promotes the occurrence of long and intense fights (Viñuela 1999), and the energy expenses of this competition would explain the slower nestling growth recorded in these broods (Seddon and Van Heezik 1991; Wiebe and Bortolotti 1994b). Interestingly, some degree of asynchrony may be favourable for last-hatched chicks, because under such conditions, although their growth rates are slowed, their final size is not clearly affected (see Bortolotti 1986a). Summarising, some degree of asynchronous hatching appears to be favourable to maximise growth rates, and to improve the "quality" of last-hatched chicks in large broods (see Slagsvold et al. 1995). Opposing selective pressures, optimal hatching asynchrony and natural variation in hatching patterns The intermediate hatching asynchrony observed in control nests may be a compromise resulting from opposing selective forces: extreme hatching synchronies increase sibling aggression (Viñuela 1999), are associated with slower growth rates and small size of last-hatched chicks, and may reduce the viability of first-laid eggs; but extreme hatching asynchronies are associated with high mortality rates, and promote brood reduction unrelated to food availability ("cainism"; Viñuela 1999). Hatching asynchrony in two-egg control clutches was positively correlated with temperature during laying, a result previously found in some species of Passeriformes, and that has been explained by some authors by the proximal constraints hypothesis (Slagsvold 1986; Slagsvold and Lifjeld 1989; Nilsson 1993): low temperatures during laying could reduce food availability and increase the energy demand of incubation and thermoregulation, thus inducing a conflict between laying and incubating eggs. Thus, at low temperatures, birds would start incubation later. However, this hypothesis is probably not valid for raptors, especially for black kites in my study area, because: (1) food before and during laying is mainly provided by males (Newton 1979; Cramp 1980; Wiebe and Bortolotti 1994a); (2) raptors gather the resources needed for egg formation before laying, while

9 passerines do it as laying proceeds (Newton 1979; Perrins and Birkhead 1983; Viñuela 1997a); (3) the relative energy demand of egg formation and incubation is much lower in large birds than in passerines (Ricklefs 1974; Gessaman and Findell 1979; Tatner and Bryant 1993); (4) energy constraints in passerines become apparent when temperatures drop below 5 10 C (Järvinen 1983), but in my study area, during the laying period of kites, the temperatures very rarely drop below 5 10 C, and (5) when passerines are exposed to thermal constraints during laying, they often delay or even stop laying, but in the Black kite, the laying interval showed slight variation and laying interruption was not observed (Viñuela 1991, 1997b, 1999). Alternatively, hatching asynchrony would increase with environmental temperature if the female starts incubating to avoid the loss of viability of unincubated eggs as temperatures approach the physiological zero, as has been suggested for house sparrows (Veiga 1992; Veiga and Viñuela 1993). A significant negative correlation was found between hatching asynchrony of three-egg control clutches and the rate of rabbit consumption. Given that the rate of rabbit consumption is mainly determined by territory location (Viñuela et al. 1994), female black kites could facultatively vary hatching asynchrony according to predicted food levels, as suggested for American kestrels (Wiebe and Bortolotti 1994a; Wiebe 1995). Intra-clutch egg size variation in the black kite has been suggested as a mechanism to enhance brood reduction when food availability is low, and a relationship between territory quality (in terms of rabbit availability) and intra-clutch egg size variation has been found (Viñuela 1997a). A similar phenomenon seems to occur for hatching asynchrony. The relationship between hatching asynchrony and temperature was more marked in two-egg clutches, while the correlation with rate of rabbit consumption was only found in three-egg clutches. Two-egg clutches are mainly late clutches in this population (Viñuela 1991), and temperatures in the study area typically rise as spring advances, so late breeders, most laying two eggs, are more exposed to high deleterious temperatures than early breeders, many of them laying three eggs. Furthermore, in two-egg clutches, where size hierarchies imposed by hatching asynchrony are not so marked (Viñuela 1996), the effect of hatching asynchrony on the survival prospects of last-hatched chicks is lower than in more asynchronous three-egg clutches (Viñuela 1997a), so adjustment of hatching asynchrony to food levels should be less important in two- than in three-egg clutches. The hatching asynchrony of three-egg clutches was less variable than in two-egg clutches, although in larger clutches, the possible range of hatching asynchronies is wider. At least two factors may explain this result: (1) third-laid eggs had a shorter incubation period, reducing expected hatching asynchrony (Viñuela 1997b), (2) the opposing selective pressures could be restraining the optimal range of asynchronies to a larger extent in three-egg clutches, because lower hatching success of first-laid eggs in control three-egg clutches has been reported (Viñuela 1997a), and size hierarchies imposed by hatching asynchrony are much larger in these clutches (Viñuela 1996). In summary, natural variation in hatching asynchrony also supports the hypothesis of opposing selective pressures on hatching asynchrony. In the case of black kites, this natural variation may be a mechanism to adjust hatching asynchrony to prevailing food levels, as reported in American kestrels (Wiebe and Bortolotti 1994a), but also to adjust the onset of incubation to prevailing environmental conditions during the laying period, as reported in house sparrows (Veiga and Viñuela 1993). Relatively few non-precocial species have completely synchronous or asynchronous hatching, most showing some intermediate pattern of hatching asynchrony (Stoleson and Beissinger 1995; Viñuela and Carrascal 1999). The effect of opposing, simultaneous selective forces may explain that widespread pattern (Viñuela and Carrascal 1999), and future studies on hatching asynchrony should consider this possibility. Acknowledgements This is a contribution to the Research Project PB of the Dirección General de Investigación Científica y Técnica. During this work I was supported by a predoctoral fellowship from P.F.P.I. of Ministerio de Educación y Ciencia. J.P. Veiga provided invaluable support and advice during all the work. He, Steven R. Beissinger, Gary R. Bortolotti, Douglas W. Mock, Tore Slagsvold, Scott H. Stoleson, and an anonymous referee provided useful criticisms on previous drafts. During the last analyses and writing, support was provided by a Fulbright Postdoctoral Fellowship at Yale University and by NSF grant no. IBN to Steven R. Beissinger. I declare that all the manipulations included in this paper were done with the necessary permissions and comply with the Spanish laws on scientific research and nature conservation. References 341 Alexander RD (1974) The evolution of social behaviour nn Rev Ecol Syst 5: Amundsen T, Slagsvold T (1991a) Hatching asynchrony: facilitating adaptive or maladaptive brood reduction? 20th Int Ornithol Congr, pp Amundsen T, Slagsvold T (1991b) Asynchronous hatching in the pied flycatcher: an experiment. Ecology 72: Anderson DJ (1989) The role of hatching asynchrony in siblicidal brood reduction of two booby species. Behav Ecol Sociobiol 25: Arnold TW, Rohwer FC, Armstrong T (1987) Egg viability, nest predation, and the adaptive significance of clutch size in prairie ducks. Am Nat 130: Bancroft GT (1985) The influence of total nest failures and partial losses on the evolution of asynchronous hatching. Am Nat 126: Bollinger PB (1994) Relative effects of hatching order, egg size variation, and parental quality on chick survival in common terns. Auk 111: Bollinger PB, Bollinger EK, Malecki RA (1990) Test of three hypothesis of hatching asynchrony in the common tern. Auk 107: Bortolotti GR (1986a) Evolution of growth rates in eagles: sibling competition vs. energy considerations. Ecology 67: Bortolotti GR (1986b) Influence of sibling competition on nestling sex ratios of sexually dimorphic birds. Am Nat 127: Briskie JW, Sealy SG (1989) Nest-failure and the evolution of hatching asynchrony in the least flycatcher. J Anim Ecol 58:

10 342 Brown LH, Amadon D (1968) Eagles, hawks and falcons of the world. Country Life Books, London Bryant DM (1978) Establishment of weight hierarchies in the broods of house martins Delichon urbica. Ibis 120:16 26 Burnham W (1978) Incubation, hatching and breeding equipment at the Peregrine Fund's Western Project. Hawk Chalk 17:47 55 Burnham W (1983) Artificial incubation of falcon eggs. J Wildl Manage 47: Campbell J, Flood F (1977) Incubation in raptor breeding. Hawk Chalk 16:38 50 Clark AB, Wilson DS (1981) Avian breeding adaptations: hatching asynchrony, brood reduction and nest failure. Q Rev Biol 56: Cramp S (ed) (1980) The birds of the Western Palearctic, vol I. Oxford University Press, Oxford Desai JH, Malhotra AK (1979) Breeding biology of the pariah kite Milvus migrans at Delhi zoological park. Ibis 121: Edwards TC, Collopy MW (1983) Obligate and facultative brood reduction in eagles: an examination of factors that influence fratricide. Auk 100: Fiuczynski D, Wendland P (1968) Zur Populationsdynamik des schwarzen Milans (Milvus migrans) in Berlin: Beobachtungen J Ornithol 109: Gessaman JA, Findell PR (1979) Energy cost of incubation in the American kestrel. Comp Biochem Physiol A 63:57 62 Gibbons DW (1987) Hatching asynchrony reduces parental investment in the jackdaw. J Anim Ecol 56: Greigh-Smith PW (1985) Weight differences, brood reduction, and sibling competition among nestling Saxicola torquata. J Zool (Lond) 205: Hagan JM (1986) Temporal patterns of pre-fledgling survival and brood reduction in an osprey colony. Condor 88: Hébert PN (1993) An experimental study of brood reduction and hatching asynchrony in yellow warblers. Condor 95: Hiraldo F, Veiga JP, Mañez M (1990) Growth of nestling black kites Milvus migrans: effects of hatching order, weather and time of season. J Zool (Lond) 222: Järvinen A (1983) Breeding strategies of hole-nesting passerines in northern Lapland. Ann Zool Fenn 20: Klomp H (1970) The determination of clutch size in birds. Ardea 58:1 124 Lack D (1954) The natural regulation of animal numbers. Clarendon Press, Oxford Lack D (1968) Ecological adaptations for breeding in birds. Methuen, London Lessells CM (1986) Brood size in Canada geese: a manipulation experiment. J Anim Ecol 55: Magrath RD (1989) Hatching asynchrony and reproductive success in the blackbird. Nature 339: Meyburg B-U (1974) Sibling aggression and mortality among nestling eagles. Ibis 116: Mock DW (1984) Infanticide, siblicide and avian nesting mortality. In: Hausfaster G, Hrdy SB (eds) Infanticide: comparative and evolutionary perspectives. Aldine, New York, pp 3 30 Mock DW, Lamey TC, Ploger BJ (1987) Proximate and ultimate roles of food amount in regulating egret and heron sibling aggression. Ecology 68: Moreno J (1987) Nestling growth and brood reduction in the wheatear Oenanthe oenanthe. Ornis Scand 18: Newton I (1977) Breeding strategies in birds of prey. Living Bird 16:51 82 Newton I (1979) Population ecology of raptors. Poyser, London Newton I (1986) The Sparrowhawk. Poyser, London Nilsson J-A (1993) Energetic constraints on hatching asynchrony. Am Nat 141: Norušis MJ (1992) SPSS for Windows: advanced statistics, release 5. SPSS, Chicago O'Connor RJ (1978) Growth strategies in nestling passerines. Living Bird 16: Olsen MW, Haynes SK (1948) The effect of different holding temperatures on the hatchability of hen's eggs. Poultry Sci 27: Perrins CM, Birkhead TM (1983) Avian ecology. Blackie, Glasgow Pettifor RA, Perrins CM, McCleery RH (1988) Individual optimization of clutch size in great tits. Nature 336: Pinkowski BC (1975) Growth and development of eastern bluebirds. Bird-Banding 46: Poole AF (1982) Brood reduction in temperate and sub-tropical ospreys. Oecologia 53: Poole AF (1989) Ospreys: a natural and unnatural history. Cambridge University Press, Cambridge, UK Richter W (1984) Nestling survival and growth in the yellowheaded blackbird Xantocephalus xantocephalus. Ecology 65: Ricklefs RE (1974) Energetics of reproduction in birds. In: Paynter RA (ed) Publ Nuttall Ornithol Club No 15, pp Schaadt CP, Bird DM (1993) Sex-specific growth in ospreys: the role of sexual size dimorphism. Auk 110: Seddon OJ, Van Heezik YM (1991) Hatching asynchrony and brood reduction in the jackass penguin: an experimental study. Anim Behav 42: Shaw P (1985) Brood reduction in the blue-eyed shag Phalacrocorax atriceps. Ibis 127: Skagen SK (1987) Hatching asynchrony in American goldfinches: an experimental study. Ecology 68: Slagsvold T (1986) Asynchronous versus synchronous hatching in birds: experiments with the pied flycatcher. J Anim Ecol 55: Slagsvold T, Lifjeld JT (1988) Ultimate adjustment of clutch size to parental feeding capacity in a passerine bird. Ecology 69: Slagsvold T, Lifjeld JT (1989) Constraints on hatching asynchrony and egg size in pied flycatchers. J Anim Ecol 58: Slagsvold T, Amundsen T, Dale S (1995) Costs and benefits of hatching asynchrony in blue tits (Parus caeruleus). J Anim Ecol 64: Smith HG, Kallander H, Nilsson J-A (1989) The trade-off between offspring number and quality in the great tit Parus major. J Anim Ecol 58: Steidl RJ, Griffin CR (1991) Growth and brood reduction of mid- Atlantic coast ospreys. Auk 108: Stinson CH (1979) On the selective advantage of fratricide in raptors. Evolution 33: Stinson CH (1980) Weather-dependent foraging success and sibling aggression in red-tailed hawks in central Washington. Condor 82:76 80 Stokland JN, Amundsen T (1988) Initial size hierarchy in broods of the shag: relative significance of egg size and hatching asynchrony. Auk 105: Stoleson SH, Beissinger SR (1995) Hatching asynchrony and the onset of incubation in birds, revisited: when is the critical period? Curr Ornithol 12: Stoleson SH, Beissinger SR (1999) Egg viability as a constraint on hatching asynchrony at high ambient temperatures. J Anim Ecol 68: Stouffer PC, Powers HW (1990) Density effects of asynchronous hatching and brood reduction in European starlings. Auk 107: Tatner P, Bryant DM (1993) Interspecific variation in daily energy expenditure during avian incubation. J Zool (Lond) 231: Veiga JP (1990) A comparative study of reproductive adaptations in house and tree sparrows. Auk 107:45 59 Veiga JP (1992) Hatching asynchrony in the house sparrow: a test of the egg viability hypothesis. Am Nat 139: Veiga JP (1993) Does brood heat loss influence seasonal patterns of brood size and hatching asynchrony in the house sparrow? Ardeola 40: Veiga JP, Hiraldo F (1990) Food habits and the survival and growth of nestlings in two simpatric kites (Milvus milvus and Milvus migrans). Holartic Ecol 13:62 71 Veiga JP, Viñuela J (1993) Hatching asynchrony and hatching success in the house sparrow: evidence for the egg viability hypothesis. Ornis Scand 24: