CONCEPTS & SYNTHESIS

Transcription

1 CONCEPTS & SYNTHESIS EMPHASIZING NEW IDEAS TO STIMULATE RESEARCH IN ECOLOGY Ecology, 86(8), 2005, pp by the Ecological Society of America SEASONAL AND LATITUDINAL TRENDS IN CLUTCH SIZE: THERMAL CONSTRAINTS DURING LAYING AND INCUBATION CAREN B. COOPER, 1,3 WESLEY M. HOCHACHKA, 1 GREG BUTCHER, 2 AND ANDRÉ A. DHONDT 1 1 Laboratory of Ornithology, Cornell University, 159 Sapsucker Woods Rd., Ithaca, New York USA 2 Audubon Science Office, 545 Almshouse Rd., Ivyland, Pennsylvania USA Abstract. Explaining patterns of latitudinal and seasonal trends in clutch size are two of the oldest and most fundamental endeavors in avian life history research. Underlying the majority of studies regarding any type of clutch size variation (i.e., individual, seasonal, latitudinal) of altricial birds is the premise that the primary cost of reproduction stems from feeding offspring. However, both altricial and precocial species of birds display latitudinal and seasonal variation in clutch size. Additionally, individual variation in costs of laying and incubation, recently demonstrated, indicates that understanding latitudinal and seasonal clutch size trends will require increased attention to earlier phases of reproduction. Given the strength and ubiquity of the clutch size patterns, many environmental factors, such as food supply and predation, have been proposed to account for the patterns, but temperature has been largely overlooked. Gradients in many variables may be important because the primacy of selection pressures may also vary in space and time. Furthermore, physiological systems may constrain responses to selection pressures. Thus, it is possible that intraspecific geographic and seasonal patterns in clutch size are at least partially influenced by temperature-dependent physiological processes. Therefore, we suggest that it is important to examine physiological responses of birds (e.g., embryo development, incubation energetics) directly influenced by physical properties of the environment, which exhibit predictable types of spatial and temporal variation (e.g., temperature, humidity, day length). We review two recently proposed, complementary hypotheses that are excellent candidates for this approach. By one mechanism, the thermal inertia of large clutches makes them favorable in cooler weather (the clutch-cooling hypothesis of J. M. Reid et al.). By the other, the reduction in egg viability under warm temperatures favors small clutches (the egg-viability hypothesis of S. H. Stoleson and S. R. Beissinger). Using general linear mixed models, we found that large-scale nesting patterns of Eastern Bluebirds and Red-winged Blackbirds are consistent with the egg-viability hypothesis in that females appear to initiate incubation before clutch completion when they lay large clutches at low latitudes. Although attempts have been made to overcome the logistical obstacles associated with studying large-scale phenomena through meta-analyses and multiple small-scale study sites, we demonstrate the significant potential of new technologies combined with volunteer-based studies to validate these hypotheses as we outline directions for future research. Key words: clutch size variation; Cornell Nest Record Program; critical temperatures; egg viability; hatching failure; incubation period; latitude gradients; onset of incubation; seasonal constraints. INTRODUCTION In many species of birds, as in reptiles, fishes, and invertebrates, clutch size is highly variable (e.g., Wyngaard 1986, Fleming and Gross 1990, Leslie 1990, Sinervo 1990, Dugan et al. 1991, 1994, Iverson et al. Manuscript received 3 December 2003; revised 28 November 2004; accepted 22 December 2004; final version received 24 January Corresponding Editor: R. Greenberg. 3 cbc25@cornell.edu , Madec and Daguzan 1993). For over half a century, it has been argued that diverse selection pressures operate on clutch size in birds (reviewed in Ricklefs 2000) so as to optimize it, although the relative importance of each pressure remains unclear. A large number of hypotheses have been formulated to explain why clutch size varies with density, habitat quality, female age, condition, and quality, time in season, and geographic location. The variation in clutch size with geographic location, a general decline in clutch size

2 August 2005 TRENDS IN CLUTCH SIZE AND TEMPERATURE 2019 toward the tropics, is one of the oldest and most controversial topics in life history research (Lack 1968, Klomp 1970, Ricklefs 2000). Most hypotheses formulated to explain why clutch size is smaller in more tropical regions than at higher latitudes fall into two broad categories. The first is a reproduction survival trade-off in which females that allocate more energy toward survival must allocate less toward reproduction (Stresemann 1927 fidé Lack 1947, Moreau 1944). The second category is the Lack clutch hypotheses, in which clutch size is constrained by the amount of food that parents can bring to their nestlings, because of varying day length (Lack 1947, Royama 1969), varying seasonality (Ashmole 1963, Ricklefs 1980, Blondel 1985), or varying predation rates (Skutch 1949, Cody 1966, Slagvold 1982). These hypotheses, many of which are not mutually exclusive, have in common the assumption that clutch sizes per se are not constrained, but that clutch sizes are varied in response to constraints on other aspects of avian life histories. However, there is growing awareness (Ricklefs and Wikelski 2002) that physiologically based constraints on life histories do exist, including demonstrations that inter-individual variation in clutch size can be constrained during the laying and incubation phases (reviewed by Monaghan and Nager 1997), and that trends in clutch size and timing of breeding due to climate change may be constrained during the egg-laying phase (Stevenson and Bryant 2000). Specifically, the response of embryos to ambient temperatures may affect embryo development and subsequent viability of embryos and post-hatching fitness of offspring; cooling rates may vary with clutch size (Vleck and Hoyt 1980, Stoleson and Beissinger 1995, 1999, Reid et al. 2000, Gorman and Nager 2004). In this paper, we summarize information on physical features of incubation and the physiology of embryo development. We review two hypotheses that serve as examples of how events during egg-laying and incubation can constrain the clutch sizes that birds can successfully raise. These two hypotheses are not mutually exclusive, and both constraints can occur, but under different environmental conditions. The hypotheses revolve around the effects of ambient temperatures on embryos. Because ambient temperatures both vary geographically, as well as through the season at an individual site, these hypotheses may help to explain both the geographic and the seasonal variation in clutch sizes that is so widely documented (Moreau 1944, Lack 1947, 1968, Klomp 1970, Toft et al. 1984, Hochachka 1990, Briggs 1993, Crick et al. 1993, Dhondt et al. 2002). Neither of these hypotheses has been thoroughly tested, and we present analyses of data that indirectly imply that both of these hypothetical constraints exist. We believe that the evidence is strong enough to warrant further and more detailed testing of the hypotheses, and we outline much of the required work. It is our view that constraints on clutch size during laying and incubation need to be viewed as a complement to the constraints imposed by other aspects of avian life histories. We suggest that these constraints during egglaying and incubation are most likely to be manifested under either very cold or very warm ambient temperatures. As such, we suggest that the selective forces that control clutch size vary across broad geographic areas, even within individual species. BIRD INCUBATION BEHAVIOR Birds lay one egg every day or every other day, and most bird species begin incubation at some point in the laying period (Hébert 2002). This behavior is most common among species with precocial young, which typically have mechanisms to synchronize the hatching event (reviewed in Brua 2002). Fewer species delay the onset of incubation until their clutch is complete. The majority of passerine birds, of those species for which incubation behavior is known, show limited variation in the onset of incubation, beginning with either the ultimate or penultimate egg of the clutch. However, passerines generally lack any mechanisms to synchronize the hatching event. Thus, initiating incubation during laying results in hatching asynchrony, with the youngest nestling frequently starving (Löhrl 1968). A late-season increase in the frequency of hatching asynchrony, due to changes in the onset of incubation, has also been documented (reviewed by Stoleson and Beissinger 1995). Incubation is energetically expensive (Mertens 1980). Studies using doubly labeled water have shown that parents spend as much energy incubating as when feeding nestlings (reviewed in Williams 1996), with the difference in field metabolic rate between incubation and nestling phases being least pronounced in small birds (Tinbergen and Williams 2002). Metabolic rates during incubation are 2 3 times higher than basal metabolic rate, with the value being strongly dependent on ambient temperature (Tinbergen and Williams 2002). In approximately half of all bird species, both parents incubate, which permits eggs to be attended 90% of the time (Deeming 2002). Only two avian orders (Passeriformes and Apodiformes) show lower levels ( 75%) of nest attentiveness (Deeming 2002). These low levels of attendance are caused not only by female-only incubation, but also by reliance on exogenous energy sources: incubating females must leave the nest intermittently to forage. Thus, passerine eggs experience frequent temperature fluctuations during incubation as the female balances her energy needs with the thermal requirements of the eggs. Embryos develop optimally between 36 and 40 C, and development is suspended at temperatures below C (Fig. 1). With these observations as a starting point, we outline two hypotheses regarding the potential role of temperature on latitudinal and seasonal trends in clutch size, and enumerate testable predictions generated by each. These hypotheses pertain particularly to birds

3 2020 CAREN B. COOPER ET AL. Ecology, Vol. 86, No. 8 FIG. 1. (a) At temperatures below physiological zero (PZ) (often assumed to be C, but may be highly variable), eggs left unattended during laying will remain viable for extended periods (Webb 1987). Prior to incubation, at temperatures between 26 and 36 C, embryos will develop slowly (dashed thick line), such that, especially embryos in first-laid eggs may undergo unsynchronized tissue growth, abnormal development, and mortality (reviewed in Deeming and Ferguson [1992]). (b) After incubation, frequent drops to temperatures between 26 and 36 C may slow development and extend the duration of the incubation period. Normal incubation temperatures are between 36 and 40.5 C, and embryos die if temperatures exceed 40.5 C (panel a). When eggs spend more time at optimal temperatures (solid thick lines), incubation proceeds quickly (but see Tieleman et al. 2004). Intermittent incubators probably let egg temperatures fluctuate between optimal and suboptimal temperatures many times each day. with small eggs and uni-parental, intermittent incubation behavior. PHYSICAL AND PHYSIOLOGICAL CONSTRAINTS ON CLUTCH SIZE Constraints from egg-viability hypothesis: Warm ambient temperature will select against laying a large clutch because large clutches must either be incubated during the laying phase (resulting in brood reduction through hatching asynchrony), or experience a higher probability of hatching failure. Veiga (1992) speculated that high rates of hatching failure in larger clutches could be due to the erosion of egg viability in eggs laid first in the sequence that have prolonged exposure to ambient environmental conditions prior to incubation. Stoleson and Beissinger (1999) hypothesized that the rapid decrease in egg viability under the climate conditions found in the tropics and late in the season in temperate regions may act as a constraint on clutch size. At temperatures below physiological zero (the temperature that triggers some embryonic development) unattended eggs in the preincubation phase will remain viable for extended periods (White and Kinney 1974, Webb 1987, Ewert 1992). Physiological zero is usually assumed to be between 24 and 26 C for most species, although most studies identifying physiological zero come from poultry (references in Webb [1987]). If eggs are exposed to extended periods of ambient temperatures between physiological zero and 36 C (the estimated start of normal incubation), the embryos will experience unsynchronized tissue growth, abnormal development, and mortality (Deeming and Ferguson 1992). Although the amount of exposure to unfavorable temperatures prior to incubation necessary to produce embryo mortality may vary among species, three days may be the minimum (Stoleson and Beissinger 1999). Therefore, eggs laid first in a large clutch ( 4 eggs) are most likely to be at risk of mortality. Females can minimize the potential for hatching failure by an early onset of incubation, particularly in environments with warm ambient temperatures (Table 1, EV1 and EV2). Constraints from clutch cooling rates hypothesis: Because large clutches retain heat longer, large clutches incur lower fitness costs than small clutches in cool, but not warm, climates. Therefore cold ambient temperatures will select for larger clutches, all other things being equal. Reid et al. (2000) showed that due to the heating and cooling properties of clutches, optimal clutch size might depend on ambient temperatures. They suggested that cooling rates are affected by several physical properties that vary with clutch size, including the amount of space that the clutch occupies, the proportion of an egg s surface in contact with other eggs, the amount of air trapped among eggs, and the amount of air circulating around the clutch. During on-bouts, incubating females expend energy to either increase or decrease the temperature of a clutch in order to optimize embryo development. On-bouts produce two types of energy demands: demands arising from re-warming eggs (or, in some circumstances, cooling eggs; Downs and Ward 1997) after an off-bout, and the presumably lower energy demands arising from maintaining eggs at an optimal temperature. Even though large and small clutches differ in their rates of re-warming, large clutches cool much more slowly than small clutches (Frost and Siegfried 1977, Reid et al. 2002b). Therefore, based on clutch warming and cooling demands, the energy requirements of on-bouts slightly favor smaller clutches and off-bouts greatly favor large clutches. During off-bouts, while females gain energy by foraging, un-

4 August 2005 TRENDS IN CLUTCH SIZE AND TEMPERATURE 2021 TABLE 1. Predictions arising from the egg-viability hypothesis (EV) and the clutch-cooling hypothesis (CC). Egg-viability hypothesis (EV) EV1. If circumstances cause eggs to be exposed frequently to temperatures between 26 and 36 C, eggs will be less likely to hatch; OR, in such temperatures, females will initiate incubation earlier in the laying sequence, resulting in more successful, yet asynchronous, hatching. Initiating incubation during laying may also shorten the nesting cycle by decreasing the time between clutch completion and hatching (Hanssen et al. 2002). EV2. Unhatched eggs should primarily be those laid earlier in the laying sequence. EV3. The proportion of time that nests are exposed to temperatures that erode egg viability can be high, and varies predictably with latitude and date. EV4. Species without mechanisms to synchronize the hatching event, such as altricial birds, are more likely to have small clutches at low latitudes. Clutch-cooling hypothesis (CC) CC1. Egg hatchability is not related to laying sequence, but to a wide array of conditions during incubation, such as humidity, microbes (Cook et al. 2003), or female attentiveness. CC2. The costs (i.e., measured in terms of energy, offspring fitness, or time currency) of incubation will vary with clutch size and temperature such that, even though costs may be inversely related to temperature, larger clutches will be proportionately less costly to incubate under cool conditions. CC3. Actual duration of incubation will vary with clutch size and temperature in a counterintuitive direction; under cool conditions, incubation will be longer for small clutches than large clutches. CC4. Birds respond rapidly to temperature conditions during incubation by adaptively adjusting daily incubation rhythms. Specifically, in cool temperatures, individuals with large clutches will decrease the duration of recesses to a lesser degree as temperatures fall; individuals with small clutches will adopt this incubation strategy under warm temperatures, all else being equal. attended eggs leave the range of optimal embryonic development, with large clutches leaving that range more slowly. The fitness costs associated with the demands of incubating small clutches in cool weather could arise through several mechanisms (Table 1, CC2). The most direct cost of smaller clutches in cold environments is the increased energy demand on the female to incubate small clutches in cold weather. Energetic limitations during incubation of small clutches in cool weather could result because the rapid cooling of the clutch restricts foraging time (Moreno and Hillstrom 1992, Reid et al. 2002b), rather than because the absolute energy requirements are too high. The amount of time that might be gained during each off-bout is small, but is summed over days that vary greatly in length across latitude, with generally longer days in the north than the south, and summed over approximately two weeks of incubation. Energy saved from lower incubation costs can be allocated to other phases of reproduction. For example, females with lower incubation costs might have shorter intervals between successive nest attempts. Alternatively, the fitness costs of incubating small clutches in cool weather could arise from the negative consequences of poor quality incubation. According to the clutch-cooling hypothesis, large clutches are favored in cooler weather because their thermal inertia permits females to take longer, and thus fewer, recesses (than would be permitted with a small clutch) without compromising embryo development. For a given session recess cycle in cool weather, embryos in small clutches may experience longer exposure to suboptimal temperatures (Table 1, CC4). A growing body of evidence suggests that poor developmental conditions can affect offspring phenotype and performance (see reviews by Lindström [1999], Metcalfe and Monaghan [2001], for more recent examples, see Gorman and Nager [2004]). Finally, the fitness costs of incubating small clutches in cool weather could arise from the negative consequences of a longer incubation period (Table 1, CC3). The higher thermal inertia of large clutches could select for large clutches in cooler weather because, for a given session recess cycle, embryos in a larger clutch will experience a higher average temperature (Reid et al. 2002b) and therefore a higher developmental rate. When eggs spend more time exposed to optimal developmental temperatures, the incubation period should be shorter (Conway and Martin 2000b, Joyce et al. 2001, Cresswell and McCleery 2003; but see Tieleman et al. 2004) (Fig. 1b). Maximizing the rate of embryonic development will reduce the number of days that eggs are exposed to predation (Cody 1966, Ricklefs 1969, Perrins 1977, Clark and Wilson 1981, Bosque and Bosque 1995, Conway and Martin 2000a) and, hence, will increase nest success. In multiplebrooded species, maximized embryonic development will increase the likelihood of subsequent nest attempts (Arcese and Smith 1988). Furthermore, embryos may deplete yolk resources during extended incubation periods, resulting in hatchlings of poor condition (Vleck and Hoyt 1980). The clutch-cooling hypothesis creates a counterintuitive prediction that small clutches can take longer to incubate. If fitness costs of incubation are related to the duration of the incubation period, small clutches are more costly than large clutches to incubate in cool weather. These two hypotheses ask whether there is evidence that thermal environments have shaped clutch size decisions, within the phylogenetic constrains that have determined the range of possible values for clutches

5 2022 CAREN B. COOPER ET AL. Ecology, Vol. 86, No. 8 FIG. 2. Relative duration of the apparent incubation periods predicted by the egg-viability and clutch-cooling hypotheses for large and small clutches in hot and cool environments. within a given species. Environmental temperatures will not always determine clutch size, but particular temperature regimes (either at the coldest or warmest parts of a species range) may be an important force influencing clutch size. In summary, although the clutch-cooling hypothesis accounts for large clutches in the north (higher latitudes) and early in the season, the egg-viability hypothesis accounts for small clutches in the south (lower latitudes) and late in the season. These hypotheses do not replace, but are additional to, the previous thinking about trade-offs between survival and reproduction and the importance of food availability during the nestling stage. In particular, these hypotheses do not directly address interspecific variation along geographic gradients, but may contribute to a better understanding of the more constrained evolutionary options within a single species. CORRELATIVE SUPPORT FOR THE PHYSICAL/PHYSIOLOGICAL HYPOTHESES Ideally, these two hypotheses are best tested by intensive studies under controlled conditions, or detailed field studies of incubation behavior as a function of ambient temperature. However, in the current absence of such data, we will present less direct tests of these two hypotheses, using latitude and season as surrogates for actual ambient temperatures. Using these data, we were able to examine two predictions regarding the duration of the incubation period (Fig. 2). Results consistent with either the egg-viability or the clutch-cooling hypothesis would indicate that more detailed confirmatory research was warranted. METHODS Volunteers monitoring nesting birds have submitted information to the Cornell Nest Record Program (CNRP), from which we used data for the years Most of the volunteers monitored multiple nest boxes, and volunteers were located throughout the United States and southern Canada. Thus, we have data on nesting of Eastern Bluebirds (hereafter abbreviated EABL) from throughout their entire range of the eastern half of the United States and adjacent parts of Canada, a range that spans over 20 of latitude. Many volunteers also submitted data to the CNRP for cup-nesting species, including two species of uniparental, intermittent incubators that exhibit latitudinal trends in clutch size: the Red-winged Blackbird (Agelaius phoeniceus, RWBL) and American Robin (Turdus migratorius, AMRO). They also provided data for one species with biparental, continuous incubation and a nonvariable clutch size of two eggs: the Mourning Dove (Zenaida macroura, MODO). With these large data sets, we expected the patterns in incubation period and hatching failure for the uniparental, intermittent incubators to meet one or more predictions (Fig. 2; Table 1, EV1 and CC3). The clutch-cooling hypothesis predicts specific trends in incubation period among intermittent incubators (Table 1, CC3), and the egg-viability hypothesis predicts specific trends in incubation period and hatching failure among species that can lay large clutches (Table 1, EV1). Patterns of latitudinal or seasonal variation in incubation period or hatching failure for the biparental, continuous incubator (MODO) were analyzed as a control, because MODOs have neither intermittent incubation nor large clutches. From the information provided by volunteers, we computed an apparent incubation period as the number of days between clutch completion and hatching (Cresswell and McCleery 2003). We term this apparent because we do not know when females actually began incubation. The two hypotheses predict different patterns of temporal and spatial variation in the duration of the apparent incubation period. According to the clutch-cooling hypothesis, the apparent incubation period should be shorter when conditions cause embryos to develop more quickly, such that incubation periods for larger clutches will appear shorter than for small clutches as ambient temperatures decrease (Fig. 2). Conversely, according to the egg-viability hypothesis, the apparent incubation period should be shorter when conditions cause females to begin incubation before clutch completion, such that incubation periods for small clutches will appear longer than for large clutches as temperatures increase (Fig. 2). In estimating the apparent incubation period, we included all breeding attempts in which birds laid more than one egg, produced at least one nestling, and for which the dates for first or final egg and hatching date were reported (EABL, n 5021 nests; RWBL, n 717; AMRO, n 318; MODO, n 176). When the date of the first egg laid was reported, we estimated the date when the final egg was laid by assuming that one egg was laid per day, with a similar rule being used

6 August 2005 TRENDS IN CLUTCH SIZE AND TEMPERATURE 2023 when the date of the last egg and the clutch size were reported (Hébert 2002). Volunteers submitting data to CNRP reported when nestlings were first observed, but rarely reported the age of nestlings when first observed. We assumed that nestlings were, on average, two days old when first observed by CNRP volunteers, because the two-day value resulted in the best match between CNPR means and averages reported in the literature for all four species. The incubation period of Eastern Bluebirds varies from 11 to 19 days (Gowaty and Plissner 1998), Red-winged Blackbirds from 11 to 13 days (Yasukawa and Searcy 1995), and American Robins from 12 to 14 days (Sallabanks and James 1999). We excluded from analyses those records for which the duration between estimated clutch completion and hatching fell more than one day outside the range reported in the appropriate BNA (Birds of North America) account. Mourning Dove incubation was reported to take 14 days (Mirarchi and Baskett 1994), but there are gaps up to 48 hours long between the laying of the first and second egg, so we excluded from analyses those records of 12 and 16 days of incubation. We acknowledge that observer errors created some variation in apparent incubation period, but we have no reason to assume that the bias from observer errors varies systematically with latitude or day of year. Volunteers also reported the presence of unhatched eggs, from which we estimated the per egg probability of hatching failure across latitude, clutch size, and day of year, because the egg-viability hypothesis predicts spatial and temporal variation in hatching failure of large clutches. For all analyses, year was a class variable with years grouped as (years when DDT was present) or (years after DDT was banned), because organochlorine pesticides increase hatching failure in most birds, including Eastern Bluebirds (Bishop et al and references therein). In order to determine whether the apparent incubation period of large and small clutches varied systematically with latitude and/or date of clutch initiation, we used general linear mixed models (PROC MIXED; SAS Institute 2001). General linear mixed models with observer identity as a random variable permitted us to partially control for the non-independence among nest attempts within years and across years. For each species, the model predicting apparent incubation period included the fixed effects of DDT-year (categorical), clutch size (continuous variable), day of clutch initiation, latitude, longitude, the interaction between latitude and longitude, and the interactions of interest: clutch size by latitude and clutch size by day of clutch initiation. We had to treat clutch size as a continuous variable because sample sizes for the largest and smallest clutches were relatively small, and preliminary analyses suggested that treating clutch size as a continuous variable gave more precise estimates of clutch size effects than did treating clutch size as a categorical variable. To analyze the Boolean response variable, the per egg probability of hatching failure, we used the GLIMMIX macro in SAS with PROC MIXED. For each species, the model predicting per egg probability of hatching failure included the same fixed effects as did models of apparent incubation period, but with clutch size as a categorical variable, because sample sizes were larger (with multiple eggs per nest) for these analyses than for analyses of apparent incubation period. The random effect in analyses of hatching failure was nest, with hatching failure of all eggs from the same nest being assumed to be partially correlated. Due to computational constraints, we performed the analysis of per egg probability of hatching failure in Eastern Bluebirds using only data from 1964 to 1973 and 1979 to 1988, inclusive (n 4038 nests). Because the start of the breeding season varies with latitude, the measurement of clutch initiation as day of year is confounded by latitude. Consequently, we developed a measure of clutch initiation that controls for latitudinal variation in the start of the breeding season. First we divided the CNRP data among 2 latitudinal bands, corresponding to even-numbered lines of latitude. Then, from all Eastern Bluebird nest attempts (n ), we used the average date of clutch initiation for the earliest 1% of nests at each latitudinal interval to develop a regression equation to define the start of the breeding season at any latitude (see Cooper et al. 2005). We performed the same computations for RWBL (n ), AMRO (n 5414), and MODO (n 9368). We compared Akaike Information Criterion (AIC) values of models containing the covariate clutch initiation either measured as calendar date or measured as the day of breeding season. For all species, models containing day of breeding season were better supported by the data (in almost all cases the AIC was greater than 4). Therefore, we report only analyses using the day of breeding season. RESULTS Data from each species showed varying levels of support for both hypotheses. The egg-viability hypothesis and the clutch-cooling hypothesis both predict that the number of days between clutch completion and hatching will vary with clutch size and temperature for each species except Mourning Doves. Mourning Doves exhibited a nonsignificant trend toward longer incubation periods later (rather than earlier) in the season and in the north. Although the sample size for MODO was smaller than those for the other species considered, even if these trends were biologically real, they represent a smaller amount of variation in incubation period (and in the opposite direction for seasonal trends) than that observed in Eastern Bluebirds, Red-winged Blackbirds, and American Robins (Fig. 3). Thus, there was no evidence from the MODO data that shorter incubation periods at southern locations and later in the season were inherently present for reasons unconnected with variation in clutch size. Furthermore, both

7 2024 CAREN B. COOPER ET AL. Ecology, Vol. 86, No. 8 FIG. 3. The egg-viability hypothesis predicted early onset of incubation in the South and late in the season, which is consistent with estimates for Eastern Bluebirds and Red-winged Blackbirds. This suggests that incubation period, measured as time from last egg laid to first egg hatched, simply appears shorter because females started incubation during laying. Day 1 of clutch initiation is at the start of the breeding season at the latitude of the nest. Lines are mean apparent incubation periods with 95% confidence intervals for 3- and 5-egg clutches (solid and dashed lines, respectively) for Eastern Bluebirds, 3- and 4- egg clutches for Red-winged Blackbirds and American Robins, and 2-egg clutches for Mourning Doves. Means and confidence intervals are estimated from general linear mixed models of Red-winged Blackbirds (df 565), Eastern Bluebirds (df 4492), American Robins (df 94), and Mourning Doves (df 91). hypotheses predicted that the interval between clutch completion and hatching would be shorter for larger clutches, under either cooler (clutch-cooling hypothesis) or warmer (egg-viability hypothesis) temperatures. Data from most species examined were consistent with the egg-viability hypothesis, with supporting patterns being most consistent for Red-winged Blackbirds. Data from Red-winged Blackbirds showed the apparent incubation period to decline with greater clutch size at more northern (higher) latitudes (P 0.05), and for larger clutches and later in the nesting season (P 0.003; Fig. 3). Also consistent with the egg-viability hypothesis were trends in the per egg probability of hatching failure in Red-winged Blackbirds, which were highest for large clutches at low latitudes (P 0.05) and for large clutches late in the season (P 0.06) (Fig. 4). With Red-winged Blackbirds, hatching failure of small clutches increased with more northern latitude, whereas hatching failure of large clutches was highest in the south (Fig. 4). Hatching failure in Eastern Bluebirds increased later in the season for both large and small clutches (P ), and decreased at more

8 August 2005 TRENDS IN CLUTCH SIZE AND TEMPERATURE 2025 FIG. 4. The egg-viability hypothesis predicted higher hatching failure in warm weather, particularly for large clutches. The lines are calculated per-egg probabilities of hatching failure for clutches for Eastern Bluebirds (df 4027) and for 3- egg (thick solid lines) and 4-egg (thick dashed lines) clutches for Red-winged Blackbird clutches (df 1711). Lighter solid and dashed lines indicate 95% confidence intervals. northern latitudes (P 0.06) (Fig. 4). Apparent incubation periods were shorter for large clutches at low latitudes than for small clutches (P 0.02), and showed a nonsignificant shortening trend as the season progressed. All of these patterns are consistent with the egg-viability hypothesis (Fig. 3). There were no significant trends in the apparent incubation period or hatching failure of American Robins, although the trend of shorter incubation periods late in the season is consistent with egg-viability hypothesis, and the trends across latitude are consistent with the clutchcooling hypothesis (Fig. 3). Unfortunately, the nest record cards had virtually no data on American Robins at low latitudes, which limited our ability to detect geographic patterns. An increase in hatching failure as the breeding season progresses appears to be a widespread phenomenon (Koenig 1982). Reduced egg viability with warmer temperatures late in the breeding season may be the cause of this pattern of hatching failure (Table 1, EV3). Our evidence supports this possibility: the late-season hatching failure occurred more often in large clutches. FIG. 5. During the 2002 breeding season, we asked volunteers to place time temperature recorders under their nest boxes (ambient temperature, in the shade). Data from these ambient temperature recorders show that at lower latitudes, temperatures are commonly between 26 and 36 C. Ambient temperatures recorded by volunteers show that a high percentage of the time, temperatures exceeded 26 C in a given 24- h period (here averaged over 2-week and 2 latitude intervals), which suggests that constraints on egg viability are widespread enough to influence seasonal and latitudinal trends in clutch size. The value within each circle (and the relative size) indicates the average percentage of each day over 26 C; blank circles indicate 25% of the day. By late April, these critical temperatures occur between 20% and 30% of each day in the South ( 30 N latitude), but hardly at all in the North ( 40 N latitude). By June, they occur almost 60% of each day in the South and only 20% of each day in the North.

9 2026 CAREN B. COOPER ET AL. Ecology, Vol. 86, No. 8 Thus, hatching failure could be a selective force for the seasonal decline of clutch size. In summer and at low latitudes, ambient temperatures that are likely to erode egg viability often persist for days (Fig. 5). The shorter incubation period observed later in the season may indicate that females began incubating prior to clutch completion in an attempt to maintain egg viability. Nevertheless, higher hatching failure could mean that females were not able to maintain egg viability, despite the frequent occurrence of early onset of incubation. A small sample of data on Eastern Bluebird nest box temperatures from 2002 showed no temperature during the laying periods that exceeded the 40.5 C considered lethal to embryos (C. Cooper, unpublished data). These preliminary data suggest that effects of temperature on egg viability may be the result of subtle alterations in embryo development, rather than direct embryo mortality due to high temperatures. Although three species showed several trends consistent with the egg-viability hypothesis, fewer trends indicate support for the clutch-cooling hypothesis. Eastern Bluebirds and Red-winged Blackbirds exhibited the general trend of large clutches having shorter apparent incubation periods, but the clutch-cooling hypothesis predicts this primarily in the north or early in the season. Tieleman et al. (2004) recently found that average egg temperature did not predict the duration of the incubation period in tropical birds. Many metabolic processes, such as somatic growth rate, are latitude dependent, and this type of latitudinal compensation has been observed in other taxa (see Bullock 1955, Dehnel 1955, Levinton 1983, Yamahira and Conover 2002). Consequently, embryos from populations at high latitudes may develop faster at a given temperature, whereas embryos from populations at low latitudes may develop more slowly at the same temperatures. If other factors, such as latitudinal differences in the rate of protein turnover for embryonic growth, determine incubation period, then an examination of the mechanisms of the clutch-cooling hypothesis should focus on female energetics. Even though our evidence for shorter apparent incubation periods is consistent with the egg-viability hypothesis, and slightly consistent with the clutchcooling hypothesis, other explanations are possible. The shorter incubation period of large clutches may be a result of determinant incubation; that is, birds may begin incubation after laying a given number of eggs, irrespective of the final clutch size (e.g., Smith 1988). Determinant onset of incubation does not conform to the hypothesis that the concentration of progesterone peaks with the laying of the final egg, initiating incubation (Mead and Morton 1985). However, it does conform to observations of slow increases in attentiveness during laying (e.g., Haftorn 1981, Cresswell and McCleery 2003). Nevertheless, determinant incubation could be a proximate mechanism, selected for ultimately because of egg viability. Another source of variation in the apparent incubation period may be gaps in laying. We assumed that females laid one egg per day, but if females tended to skip a day of laying during bouts of cool weather (see Cresswell and McCleery 2003), then apparent incubation periods would be longer when we estimate clutch completion based on first-egg dates. For Eastern Bluebirds, the laying date of small clutches was more likely to be estimated based on early-laid eggs (logistic regression, P ). When nests are detected early, estimates of apparent incubation period, which are based on the presumed date of the last-laid egg, are more likely to be lengthened by laying gaps. Thus, laying gaps could produce the same patterns that are predicted by the egg-viability hypothesis. However, the frequency of laying gaps appears to be small, because the average apparent incubation period only varied by a fraction of a day ( day, mean SE), depending on whether the first record of egg-laying was from early or late in the sequence. The effect was essentially invariant across latitude and season. This finding was from a general linear mixed model including the early late categorical predictor, as well as its interaction with both latitude and day of season. For other species, most nests were found before three eggs were laid, and we could detect no effects of laying gaps. Another source of variation in the apparent incubation period could be hatching failure of first-laid eggs, which would make the incubation period longer by extending it to the hatch of the second egg. Although we do not know the laying sequence of unhatched eggs, the failure of first eggs to hatch most likely would be attributed to loss of egg viability; we observed shorter, not longer, apparent incubation periods under conditions where this type of hatching failure was expected. Given the weakness in alternative explanations for our observed patterns, our correlative findings regarding approximate incubation period and hatching failure lay a foundation to pursue both the clutch-cooling and the egg-viability hypotheses in more detail. We will outline research priorities to explore the potential of these hypotheses in accounting for latitudinal and seasonal trends in avian clutch size. DIRECTIONS FOR FUTURE RESEARCH The analyses that we have presented are only capable of suggesting patterns of incubation period and hatching failure that are consistent with the role of environmental temperature as a constraint on clutch sizes in birds. The following lines of research need to be pursued in order to directly establish the importance of environmental temperature as a determinant of clutch size. Further testing the egg-viability hypothesis Details of embryo thermal tolerance prior to incubation. The temperature above which development begins, and below which development suspends, is fre-

10 August 2005 TRENDS IN CLUTCH SIZE AND TEMPERATURE 2027 quently termed physiological zero. Although frequently cited for best estimates of physiological zero, Webb (1987) actually cautioned against the use of a single physiological zero for all species. Stoleson and Beissinger (1999) found that 3 5 days of exposure to ambient temperatures were necessary to erode the viability of wild parrot eggs. However, even minimal contact with eggs by the female during laying may result in even longer durations before egg viability is substantially eroded. If many days are necessary to erode viability, then egg viability is unlikely to be a mechanism for clutch size selection in species with small or medium (e.g., 6 eggs) clutch sizes. Therefore, research priorities should be to (1) determine physiological zero for different species, (2) determine the number of hours/days of exposure above physiological zero prior to incubation needed to erode egg viability, and (3) determine whether temperatures that erode egg viability occur across latitude and season in such as way that, if temperature is a selective force, large-scale and seasonal temperature patterns would create observed clutch size trends. Large-scale patterns of onset of incubation and hatching failure. The egg-viability hypothesis specifically predicts that, when females are unable to prevent hatching failure, it will be the first-laid eggs that fail to hatch because first-laid eggs have the potential to receive the most exposure to temperatures that can erode viability (Table 1, EV2). Eggs can also fail to hatch when embryos exceed lethal temperatures. This type of overheating is probably not a factor, based on our nest box temperature recordings (C. Cooper, unpublished data). Even though weather station temperatures, particularly in the South, can exceed 40 C in the summer, the microclimate around nests may not. Birds that nest in temperatures that exceed 40 C, such as shorebirds on sandy beaches, are known to cool their eggs during incubation (Downs and Ward 1997). Although we are unaware of observations of birds cooling their eggs prior to incubation, this possibility clearly deserves attention. Experimental work, such as desensitizing the brood patch with anesthetic (White and Kinney 1974), has demonstrated that birds can monitor egg temperatures closely during incubation through their brood patch. However, manipulative experiments, such as increasing or decreasing the temperature of clutches prior to incubation, can establish whether birds respond to temperature with a change in the onset of incubation, and whether the effects of temperature can operate independently of seasonal changes in other factors (such as day length). Therefore, research priorities include (1) determining the frequency of hatching failure of first-laid eggs vs. other eggs in the sequence, across latitude and season, and determining whether loss of egg viability is the cause; and (2) determining whether birds vary the onset of incubation in adaptive ways, based on temperature. Further testing the clutch-cooling hypothesis The fitness costs of small clutches in cool weather could manifest as energetic constraints, negative consequences of poor incubation conditions, or negative consequences of a longer incubation period. Each of these possible mechanisms calls for a specific research agenda. Energetic constraints. As previously mentioned, although most studies have shown that the energy required for incubation decreases with higher ambient temperature and increases for larger clutch size, this is not always the case. One reason for the equivocal results is an interaction between temperature and clutch size on incubation energy demands, as proposed by the clutch-cooling hypothesis. Thus, research priorities include examining incubation energetics of different clutch sizes under different ambient temperatures through manipulations of energy expenditure, energy intake, and energy reserves (described in Reid et al. 2002a), while controlling for female age and experience. When the total influence of clutch size on cooling rates is computed, the variation in total time incubating per day (as a function of day length) will influence the magnitude of the costs and benefits. Poor incubation. Given that ambient temperature and clutch size affect incubation rhythms, it is likely that the quality of care received by embryos can vary greatly among clutches. Growing evidence suggests that conditions during nestling development influence the subsequent performance of the bird as an adult, such as its clutch size (Haywood and Perrins 1992), and subcutaneous fat during autumn migration (Merilä and Svensson 1997). Little is known about the specific consequences of conditions during embryo development on subsequent offspring quality and performance. Vleck and Hoyt (1980) found that embryos deplete yolk resources during extended incubation periods, resulting in hatchlings of poor condition. Also, the temperature that defines physiological zero, and thus the effects of temperature fluctuations on embryo development, probably varies with the age of the embryo (Webb 1987). Therefore, research priorities include (1) determining the amount of exposure below physiological zero during incubation needed to cause embryo mortality or developmental problems; and (2) determining how different session recess cycles affect offspring phenotype and performance, using field observations and manipulations of incubation under artificial conditions. Longer incubation and trade-offs. The negative consequences of longer incubation periods may be avoided by adaptations other than adjustments in clutch size. For example, increasing provisions in egg yolk may be a strategy to accommodate longer incubation periods without affecting offspring quality (Tinbergen and Williams 2002). There are other ways in which fitness costs of small clutches in cool weather could

11 2028 CAREN B. COOPER ET AL. Ecology, Vol. 86, No. 8 be masked by trade-offs and the reallocation of resources. For example, even if small clutches were energetically more expensive to incubate, females could allocate energy that would have been used to feed nestlings toward incubating the eggs more quickly (e.g., Reid et al. 2000). Furthermore, patterns could be obscured by latitudinal compensation, as has been observed for other traits in other taxa (Bullock 1955, Dehnel 1955, Levinton 1983, Yamahira and Conover 2002). Therefore, research priorities include (1) gathering measures of geographic and seasonal variation in the actual (rather than apparent) incubation period for varying clutch sizes; and (2) examining the relationships among daily incubation rhythms, egg temperatures, development rates, egg volume, and final incubation period, in order to identify some of the potential trade-offs in the role of temperature influencing clutch size via the duration of the incubation period. Studying large-scale patterns Comparative analyses. Over 50 years ago, Lack (1947) compiled data to suggest that birds in the orders Passeriformes, near-passerine, Strigiformes, Falconiformes, Ciconiiformes, Ralliformes, and Galliformes exhibit latitudinal variation in clutch size, whereas birds in Alcidae do not. Other orders contain birds with ranges too restricted to exhibit latitudinal variation (e.g., Stercorariidae), or birds with either insufficient data or nonvariable clutch sizes (e.g., Pelecanifomres, Columbiformes, Laridae, Anseriformes, Podicipediformes, Limicolae, Clareolidae, and Gruidae). More studies (e.g., Kulesza 1990, Young 1994) are still needed to clarify the overall pattern of clutch size trends in different orders. Selection for small clutches in warm weather, according to the egg-viability hypothesis, should only pertain to species without mechanisms to synchronize hatching (most precocial species have synchronization mechanism, Table 1, EV4). If the clutch-cooling hypothesis alone were to account for latitudinal and seasonal trends in clutch size, then we would see clutch size trends most strongly in uniparental, intermittent incubators; perhaps less strongly in biparental, intermittent incubators; and not at all in biparental, continuous incubators. Life history traits commonly occur in certain combinations; for example, precocial species are often those with continuous incubation, larger eggs, relying on endogenous reserves, and not multi-brooded. Similarly, many continuous incubators are frequently those with precocial young (e.g., waterfowl), whereas intermittent incubators are those with altricial young (e.g., songbirds). These life history correlations make it difficult to distinguish the influence of incubation behavior from constraints of brood rearing (Lack s hypothesis) on clutch size, although exceptions exist (e.g., doves and pigeons are biparental, continuous incubators with altricial young). Comparative analyses are needed to determine whether latitudinal and seasonal trends in clutch size, and the likelihood that egg viability or clutch cooling are selective forces, are associated with species that lack the ability to synchronize hatching, or have uniparental, intermittent incubation, respectively. Volunteer resources. The coordination of data collection by volunteer bird enthusiasts is one useful technique to overcome the logistical obstacles of conducting research over large geographic scales (e.g., Wells et al. 1996, Rodrigues et al. 2000, Dhondt and Hochachka 2001, Noble and Freeman 2001, Steenhof et al. 2001, Benton et al. 2002, Robinson et al. 2003, Sauer et al. 2003). To address the hypotheses discussed in this paper, we began pilot-testing an approach during the 2002 and 2003 breeding seasons, recruiting volunteers across the country from The Birdhouse Network (TBN). Volunteers in TBN, administered by the Cornell Laboratory of Ornithology, Ithaca, New York, USA, monitor nest boxes to gather information on the habitat characteristics surrounding the box and details of the nesting attempt. They record when and how many eggs, nestlings, and fledglings occur (information available online). 4 Temperature fluctuations in the nest can be used to infer incubation behavior (Conway and Martin 2000b, Joyce et al. 2001, Badyeav et al. 2003). We provided volunteers with reusable, dime-sized, selfcontained data loggers that record date, time, and temperature at specified intervals, have a programmable start delay, and 512 bytes of memory (Thermochron ibutton, Maxim, Dallas, Texas, USA). Participants used these data loggers to measure three temperatures in relation to the nesting of Eastern Bluebirds: ambient temperature (in the shade), temperature inside the nest box, and temperature among the eggs in the nest cup. More than 50 participants have successfully used data loggers to record nest cup temperatures and 10 participants have programmed data loggers themselves and submitted their recordings electronically, whereas the others received pre-programmed data loggers through the mail. Recordings of temperatures from nest cups reveal the onset of incubation and permit us to assess variation in incubation rhythms (Fig. 6). That temperature fluctuations recorded with ibutton data loggers in the nest cup correspond to these behaviors has been verified with videotaping at nests of Tree Swallows (D. Ardia, personal communication) and House Finches (Badyaev et al. 2003). In the figure, a drop in nest cup temperature indicates the departure of the female from the nest (Fig. 6). We emphasize that these data indicate temperature fluctuations in the nest cup and serve to reveal female behavior, but do not necessarily correspond to egg temperatures. As we gather more information from this data collection technique, we will be able to test more predictions of the egg-viability and clutch-cooling hypotheses. 4