INTRODUCTION

Nest success is a key factor in the dynamics and evolution of avian populations (Martin 2015). Because nest success can vary with traits of individuals, with the composition of populations, and with changes in the biotic and abiotic environment (Lebreton et al. 1992, Dinsmore et al. 2002, Arcese 2003, Germain et al. 2015), short-term studies of nest success may yield biased estimates of variable influence when population size or composition, or biotic and abiotic environmental factors, vary at temporal scales that exceed the study's duration (Wiens 1977, Franklin 1989, Norris et al. 2007, Clutton-Brock and Sheldon 2010, Martínez-Padilla et al. 2014, Blight et al. 2015). In such cases, long-term studies may be needed to reliably identify management actions most likely to influence nest success, population growth, or evolutionary response in the future (Arcese 2003, Benton et al. 2006, Lovett et al. 2007, Clutton-Brock and Sheldon 2010). We explored these possibilities by estimating the influence of 2 intrinsic factors (female age and inbreeding coefficient), 2 abiotic factors (rainfall and temperature), and 3 biotic factors (breeding densities, parasitism rates, and brood parasitism) on nest success (≥1 fledgling) in a 39 yr study of an island Song Sparrow (Melospiza melodia) population subject to marked environmental variation (Table 1). We then compared those estimates to results from 3 parallel models based on three 13 yr periods—representing the first, middle, and last thirds of the overall study—to test for temporal change in variable influence.

TABLE 1.

Factor classes, associated variables, usage rationale, and the expected relationships of each variable in predicting nest success in our study population of Song Sparrows on Mandarte Island, British Columbia, Canada, 1975–2014.

METHODS

Factor Definitions

We explored 2 intrinsic, 2 abiotic, and 3 biotic factors that potentially affected nest success (≥1 fledgling) from a total of 2,895 nests over a 39 yr period. Intrinsic factors included the age and f of nesting females. Female age was unknown in 1975 but was known for 97% of total nest records by 1979, and for nearly 100% of records thereafter (Appendix Figure 4). Because females may breed for ≤9 yr (median = 2; Table 2 and Appendix Figure 5), we included age and age² as fixed effects in models to accommodate quadratic fits between female age and performance (e.g., Tarwater and Arcese 2017a, 2017c). We didn't include male age in our analyses because males contribute to parental care only after young hatch and because prior results had indicated small average effects of male age on reproductive rate (Keller et al. 2008, M. D. Crombie and P. Arcese personal observation).

TABLE 2.

Range, median, and first and third quartiles (Q1:Q3) of the intrinsic, abiotic, and biotic variables used to predict nest success in our study population of Song Sparrows on Mandarte Island, British Columbia, Canada, 1975-2014. Annual rates of brood parasitism are reported here to show variation in the likelihood of parasitism but are modeled as a binary variable associated with each nest (see text).

Inbreeding (f) was estimated by merging a social (1975–1990) and genetic (1991–2014) pedigree to obtain precise estimates for >75% of birds hatched after 1993, based on the identity of all 4 social and/or genetic grandparents (Sardell et al. 2010, Reid et al. 2014). Prior to 1994, f was estimated for 0–88% of birds annually, based mainly on the social pedigree but with reduced precision due to extrapair paternity (Reid et al. 2014). Females with precise estimates of f initiated 1,819 nests (63% of 2,895 nests), mainly in the last half of the study (Appendix Figure 6). To include all females in analyses, we imputed the mean value (± SE) of f over the study (0.06 ± 0.001, n = 1,076; Table 2) for all birds that did not have precise estimates of f. Doing so allowed us to compare factors potentially influencing success over all years without excluding the earliest years, when populations were larger and fluctuations more extreme. The consequences of imputation were explored by comparing models with and without imputed data included, but we report imputed model estimates here because those results were similar to estimates from models based only on females with precise estimates of f (see Appendix Table 4 and Appendix Figure 7).

Abiotic factors included cumulative rainfall (mm) and the minimum temperature (°C) experienced during the nesting period, defined as the 31-day window from the initiation of nesting to fledging (e.g., Arcese et al. 2002). The earliest nest was built during the last week of February, and the latest during the third week of July. Weather was recorded 11 km west of Mandarte Island (Victoria International Airport;  http://climate.weather.gc.ca).

Biotic factors included local breeding density and brood parasitism rate in the vicinity of nests, and the parasitism of individual nests (presence–absence of one or more Brown-headed Cowbird eggs). Local density was estimated for each nest as the number of nests within a 50 m radius that were initiated by other females in the same year, irrespective of timing (an area of 7,854 m², representing 2–8 territories on average; Smith et al. 2006b; see also Germain and Arcese 2014). We estimated local parasitism similarly, but as the fraction of nests within the 50 m radius that were parasitized. Brood parasitism was recorded as presence or absence of one or more Brown-headed Cowbird eggs and treated as a binary variable to estimate direct effects of parasitism from the potential effects of parasite activity (Smith and Arcese 1994, Arcese et al. 1996). Other potential nest predators on the island include Northwestern Crows (Corvus caurinus), deer mice (Mus musculus), Glaucous-winged Gulls (Larus glaucescens), river otters (Lontra canadensis), Red-winged Blackbirds (Agelaius phoeniceus), and European Starlings (Sturnus vulgaris). However, in the absence of Brown-headed Cowbirds, nest depredation averages <20% annually and is relatively constant (Arcese and Marr 2006).

The Julian date of the first egg laid in each nest (“lay date”) was used as a covariate in models to account for seasonal variation in nest success (e.g., Appendix Figure 8) and was known by observing clutches or by backdating nestling age. Lay date was unknown for 149 nests (5% of 2,895 nests) that were found after young fledged, found after failing, or never found but known to exist on the basis of female behavior. We estimated lay date for these 149 nests as the mean lay date of nests active in the same part of the breeding season annually, and we report the consequences of doing so in Appendix Table 4 and Appendix Figure 7. Results reported below include imputed lay dates.

Female identity (female ID) and year of study were included as random effects in models. Female ID helped account for uneven sampling across females, and year of study accounted for annual variation in models not attributed to our main factors of interest (Table 1).

RESULTS

Of 2,895 nest attempts examined, 64% were successful (Appendix Figure 9). Annual variation in the 7 factors hypothesized to affect nest success was substantial (Figure 1 and Table 2). Female age and f varied from 1 yr to 9 yr (coefficient of variation [CV] = 0.62) and from 0 yr to 0.28 yr (CV = 0.91), respectively (Table 2). Cumulative rainfall and minimum temperature during nesting varied from 0 mm to 128 mm (CV = 0.59) and from −3.2°C to 10.2°C (CV = 0.91), respectively (Table 2). Local density, local parasitism, and the annual proportion of nests parasitized varied from zero nests to 46 nests (CV = 0.60), from zero nests to 12 nests (CV = 1.37), and from 0% to 44% (CV = 2.78) across years, respectively (Table 2).

FIGURE 1.

Temporal variation in nest success (gray lines) and 6 potential drivers of success (solid black lines) on Mandarte Island, British Columbia, Canada, 1975–2014. The black dashed line represents inbreeding (f) estimates largely based on social pedigree, which reduces the precision of those estimates via extrapair paternity (∼27% of young; Reid et al. 2014); the precision of f increased through time as genetic data were accumulated and lineages refined. Detailed monitoring did not occur in 1980.

Five of 7 hypotheses in Table 1 were supported in univariate tests of the 7 factors of interest. Nest success initially increased with female age (β = 0.76, 95% CI: 0.11–1.41, n = 2,706), then declined later in life (female age²; β = −0.82, 95% CI: −1.48 to −0.17, n = 2,706), revealing a quadratic relationship between nest success and female age overall (Figure 2). In addition, nests initiated by more inbred females (β = −0.23, 95% CI: −0.43 to −0.03, n = 2,895) and those exposed to more rainfall (β = −0.37, 95% CI: −0.60 to −0.13, n = 2,895), high local breeding density (β = −0.55, 95% CI: −0.83 to −0.27, n = 2,818), and high local parasitism rates (β = −0.57, 95% CI: −0.90 to −0.25, n = 2,812) were also less likely to succeed. By contrast, neither the minimum temperature during nesting nor an act of brood parasitism predicted nest success in single-factor models (β = 0.25, 95% CI: −0.20 to 0.69, n = 2,895 and β = −0.08, 95% CI: −0.36 to 0.21, n = 2,893, respectively).

FIGURE 2.

Probability of nest success in relation to the age of female Song Sparrows (solid line; dashed lines indicate 95% confidence intervals) on Mandarte Island, British Columbia, Canada, 1975–2014. Values are logit back-transformed partial effects of female age and female age² from the averaged model to predict nest success over the 39 yr of the study, while keeping all other intrinsic, abiotic, and biotic factors at their mean.

Of 128 possible models, 4 were within ΔAIC < 2 of the best model (hereafter “top-ranked models”; Table 3). The average of these top-ranked models included all 7 factors of interest, with a cumulative Akaike weight of 0.40 (hereafter “averaged model”; Table 3). In the averaged model (n = 2,639), brood parasitism tended to be a negative predictor of success (95% CI: −0.6 to 0.01; Table 3), whereas local parasitism became less influential on success (95% CI: −0.74 to 0.03; Table 3) compared to univariate models. Overall, based on the averaged model coefficients, nests were less likely to succeed when initiated by inbred females (95% CI: −0.42 to −0.03; Table 3), when subjected to high rainfall during the nesting period (95% CI: −0.57 to −0.12; Table 3), in years with high breeding densities (95% CI: −0.84 to −0.18; Table 3), and if they were parasitized by Brown-headed Cowbirds (95% CI: −0.6 to 0.01; Table 3). As in our univariate tests, nest success first increased (95% CI: 0.17–1.46; Table 3) then declined with female age (95% CI: −1.5 to −0.2; Table 3).

TABLE 3.

Parameter estimates for factors identified as predictors of nest success in top-ranked models for our Song Sparrow dataset from Mandarte Island, British Columbia, Canada, 1975–2014.

By contrast, none of the 7 factors we tested predicted nest success in all three 13 yr periods (Figure 3 and Appendix Table 6). Female age and female age² were included only in models based on the entire 39 yr study, but the direction and magnitude of their estimated effects on nest success were consistent across periods (Figure 3 and Appendix Table 6). Local density behaved similarly, predicting reduced nest success over the entire study but varying in magnitude across periods (Figure 3 and Appendix Table 6). By contrast, despite detecting negative effects of inbreeding, rainfall, and parasitism on nest success over the entire study, we found these effects to be highly variable in magnitude and direction across periods (Figure 3).

FIGURE 3.

Standardized coefficients (with 95% confidence intervals) of the averaged models to predict nest success from the entire (1975–2014), early (1975–1988), middle (1989–2001), and late (2002–2014) study periods on Mandarte Island, British Columbia, Canada (see Appendix Table 6).

DISCUSSION

We found strong support for the influence of 7 intrinsic, abiotic, and biotic factors previously found to affect nest success in birds in a 39 yr study of a single Song Sparrow population, but we also found that the magnitude and significance of those effects varied temporally (Table 1 and Figure 3). Such variation might arise because the factors we considered are linked to other latent variables not included in our models, or because reducing sample size (from 39 yr to 13 yr) also reduced the precision of estimates based on small or inconsistent effects. Below, we first explore potential causes of temporal variation in our estimates and then discuss their implications.

Nest success was similar in 1-yr-old and 5-yr-old female Song Sparrows, 5–8% higher in 3-yr-olds, but 9–38% lower in females ≥6 yr old (Figure 2). This pattern mirrors age-related changes in annual reproductive success reported from this population (Keller et al. 2008, Tarwater and Arcese 2017a, 2017c), indicating that variation in nest success contributes to that pattern. Age-related variation in reproductive success is widely reported in birds (Sæther 1990, Forslund and Pärt 1995, Martin 1995) and is typically attributed to initial increases in food acquisition and predator avoidance skills, followed by senescence in one or more traits (e.g., Desrochers and Magrath 1993, Robertson and Rendell 2001, Smith et al. 2006a, Keller et al. 2008). However, age-related variation in reproductive effort, performance, and survival also reflects individual life history (Reznick et al. 2000, Roach and Carey 2014). Thus, variation in nest success may also reflect a female's allocation of resources to her own survival vs. reproduction in the current year, which trade off strongly in the Mandarte Song Sparrow population (Tarwater and Arcese 2017a, 2017c). The fact that female age was retained in models based on the first, middle, and last 13 yr of the study (Table 1 and Figure 3), despite an expected increase in associated prediction intervals (Figure 3 and Appendix Table 6), indicates that age-related variation in nest success is a persistent feature of this population. By contrast, because few females reach senescent age classes in species with high average mortality (Nussey et al. 2008), the age-related declines in success that we documented will have very limited effects on demography overall (Brommer et al. 2007, Roach and Carey 2014).

Nest success also declined as inbreeding (f) increased in female Song Sparrows (Tables 1 and Figure 3), supporting Keller's (1998) finding that hatch success and nestling survival were lower in the nests of inbred females, and Taylor et al.'s (2010) finding that the survival of Song Sparrow embryos declined as homozygosity and f increased. However, the estimated effect of f on nest success was modest, being 16% lower in the most inbred vs. outbred females (f = 0.28 vs. 0.00, respectively; Table 2). In addition, f was retained as a predictor of success in only the middle third of our study, which coincided with a rapid increase in mean f (Figures 1 and 3). Temporal variation in the magnitude of inbreeding depression is expected, given interactions with episodic stressors such as extreme weather events (Table 1), and the potentially large effects are unlikely to be detected in short-term studies (Keller et al. 1994, Coltman et al. 1999, Meagher et al. 2000, Kristensen et al. 2005).

Nest success declined as rainfall increased during the nesting period over the entire study (Figure 3 and Table 1), but those effects varied temporally (Figure 3 and Appendix Table 6). Rainfall is a well-known stressor during nesting and a proximate cause of nest failure, with the potential for tonic and interactive effects (Ricklefs 1969, Frederick and Collopy 1989, Aguilar et al. 2000, McDonald et al. 2004, Sillett et al. 2004, Ouyang et al. 2012). In Great Tits (Parsus major), stress-induced increases in plasma corticosterone preceded nest failure and occurred more often during inclement weather (Ouyang et al. 2012). In Black-throated Blue Warblers (Setophaga caerulescens), reproductive success declined as breeding density increased, but more so in years with high rainfall (Sillett et al. 2004). The existence of such interactions and the irregularity of weather and climatic extremes emphasize the value of long-term studies, especially when combined in meta-analyses (Sæther and Bakke 2000, Sæther and Engen 2002). In contrast to rainfall, temperature during nesting was unrelated to nest success (Figure 3), despite the fact that temperature can affect the energetic costs of breeding (Bryan and Bryant 1999, Pérez et al. 2008, Ardia et al. 2009) and nest-site preference (Germain et al. 2015). Interestingly, Song Sparrows in coastal California produced more broods and more young in years with higher cumulative rainfall (Chase et al. 2005). Tarwater and Arcese (2017b) also reported that higher rainfall in late winter and summer led to longer breeding seasons on average. Future studies might therefore consider both cumulative and episodic rainfall, given that higher rainfall is expected to increase reproductive output but may also reduce nest success.

Nest success also declined as population density increased (Table 1 and Figure 3), with a similar effect in all periods examined, but with prediction intervals that included zero in the last 13-year period, when sample size was least (n = 778; Figure 3 and Appendix Table 6). Two potential mechanisms for this decline are (1) that Brown-headed Cowbirds visit the island more often and depredate more nests at higher host densities or (2) that high population density reduces nest success by increasing intraspecific strife or by reducing per capita food abundance (Table 1). Using 16 yr of data in which Brown-headed Cowbirds were present on Mandarte Island and 23 yr when they were absent, we found that nest success declined significantly with increasing density in years with Brown-headed Cowbirds present (β = −0.46, 95% CI: −0.77 to −0.14) but not in years when they were absent (β = −0.40, 95% CI: −0.91 to 0.11). This result confirms that Brown-headed Cowbirds have the potential to regulate host population growth at population and regional scales (Arcese et al. 1996, Jewell and Arcese 2008). The rate of brood parasitism near a focal nest also reduced its success, as did being parasitized (Figure 3). But these effects varied across periods because Brown-headed Cowbirds were absent in many years and rare later in our study, when Song Sparrow density was also low on average (Figure 1).

We detected the effects of several factors hypothesized previously to influence nest success in birds but found that their occurrence and magnitude varied greatly over 4 decades. In Song Sparrows, age-related variation in nest success and the negative effects of high breeding density were consistent predictors of success, whereas those linked to inbreeding, Brown-headed Cowbirds, and weather were episodic but sometimes large. These findings reemphasize the value of long-term studies, given that rare events can have large but uncertain effects on populations (Ludwig 1999) and may also interact with population structure and the biotic or abiotic environment to influence abundance and distribution (Wiens 1987, Franklin 1989, Norris et al. 2007, Clutton-Brock and Sheldon 2010, Martínez-Padilla et al. 2014, Blight et al. 2015). Meta-analyses of other long-term studies, especially those aimed at identifying factors with large and persistent effects on populations (e.g., Sæther and Engen 2002, Sutherland et al. 2017), could help managers prioritize actions and anticipate temporal variation in success as a consequence of co-occurring trends in climate, population structure, or community composition.