How and why avian reproductive success varies in response to precipitation: A global meta-analysis

Significance

The timing, intensity, and seasonality of weather events shape ecological phenomena in many systems. However, the relative importance of these abiotic factors relative to biotic factors for animal populations remains elusive. Mechanisms underlying animal responses to temperature are relatively well studied, but the reasons for animal responses to variation in precipitation are largely unknown. This global synthesis elucidates why precipitation leads to divergent responses in species living in different regions having different life histories and provides a framework for understanding mechanisms underlying bird population change under current and future precipitation regimes.

Abstract

Precipitation influences animal physiology, behavior, and ecological interactions, therefore affecting demography. However, the nature and strength of those effects vary widely, have been attributed to mechanisms operating over different time scales, and may depend on species-level attributes. We conducted a global meta-analysis accounting for shared evolutionary history to a) understand the nature of relationships between precipitation and avian reproductive success and b) test alternative hypotheses explaining where and why precipitation affects some bird species more than others based on species-level attributes or environment. The nature and magnitude of responses to precipitation were unrelated to species attributes but were dependent upon the timing of precipitation and geographic context. Birds responded negatively to precipitation that fell during the nesting period, while birds responded positively to precipitation that fell prior to the breeding season. Responses to precipitation varied with elevation and latitude; at high elevations at high latitudes, birds experienced lower reproductive success following precipitation, while at high elevations at low latitudes (i.e., in the tropics), the responses were opposite, though sample sizes at high elevations in the tropics were small. Negative responses in colder regions likely reflect direct thermogenic consequences of wetting. In warmer regions, because positive responses to precipitation were unrelated to diet or predation risk, responses to precipitation were likely indirect and mediated by vegetation. Ultimately, this study implicates both direct and indirect responses to precipitation that depend on the timing of precipitation and biogeographic context with large implications for avian populations under changing climate regimes.

Precipitation and temperature are the primary sources of climatic variation across the globe. The timing of hot and cold seasons, wet and dry seasons, and the absolute values of temperature and precipitation define biomes (1, 2), set species range limits (3), and shape animal phenology (4). The mechanisms underlying animal responses to temperature are well studied; excessive cold and heat increase energetic costs of thermoregulation and can be lethal near critical thresholds, ultimately affecting population growth rates and species distributions (2, 5, 6). However, we know less about how and why precipitation affects animal population growth.

Precipitation may affect demography via direct physiological responses, behavioral changes that mitigate physiological costs, or altered ecological interactions occurring due to precipitation that indirectly influence focal taxa via other species’ responses to rain (7). Those responses could thus be mediated by mechanisms potentially operating at different parts of a species’ distribution, annual cycle, or life cycle (8). Unsurprisingly, evidence regarding how animals respond to variation in precipitation therefore remains ambiguous and sometimes contradictory. Divergent associations between precipitation and reproductive success have been documented from birds having diverse life history strategies and living in different biomes. Those species-specific associations vary between positive and negative (9–12), with responses sometimes even varying from positive to negative within species, at different sites (13) or at different times throughout annual or nesting cycles (14–21).

Relationships between animal population growth and the timing of precipitation can help determine whether responses manifest via direct or indirect mechanisms. The direct effects of high precipitation amounts should manifest over short time periods and would be expected to elicit negative responses mediated by thermogenic costs of wetting (increased costs of maintaining stable internal temperatures), behavioral changes to avoid wetting, or physical harm as a consequence of storms. In birds especially, early life stages may be particularly susceptible to direct adverse consequences of precipitation as they are endotherms with external embryonic development. Bird eggs must be maintained within narrow species-specific ranges of environmental conditions to properly develop (22). Furthermore, eggs and nestlings are stationary throughout development, exposing them and their parents to higher risk of unfavorable environmental conditions (23).

By contrast, high or low precipitation amounts over relatively long time periods (e.g., multiple months, a year, or multiple years) can result in changes in other aspects of ecological communities that lead to indirect and either negative or positive associations between precipitation and population growth (24–26). For example, plant community composition and structure can respond to interannual variation in precipitation (27), and such changes to vegetation may influence reproduction via the availability of climatic microrefugia (28). Concurrent with changes in vegetation, prey, and/or predator populations may increase or decrease (26) resulting in either positive effects or negative effects on avian fecundity (29). Such indirect relationships between precipitation and animal populations must operate over periods of months to years because they result from complex interactions spanning multiple trophic levels. Importantly, indirect mechanisms linking precipitation to reproduction do not necessarily depend on temperature as the direct mechanisms do.

Biogeographic and species-level attributes should modify responses of birds to precipitation via these direct and indirect mechanisms. Responses to precipitation may depend on elevation and latitude, due to latitudinal and elevational gradients in temperature. If so, relationships between precipitation and reproduction likely depend on underlying direct mechanisms. Cooler temperatures at higher elevations and higher latitudes could affect the relationship between animals and precipitation in several ways. Precipitation may cause stronger direct negative responses at higher elevations (and in particular, high elevations at high latitudes) because thermoregulatory costs are higher in cold temperatures and increase when an animal is wet (6, 30). To minimize those costs, animals behaviorally avoid wetting which can lead to reduced foraging (31), potentially influencing nest attentiveness and fledging success. Lower temperatures also decrease plant productivity, which can reduce prey availability (32). On the other hand, reduced evapotranspiration due to colder temperatures may mean that plants are less water-limited at higher elevations than lower elevations [i.e., plant biomass and structure at higher elevation can be maintained at lower moisture levels than at lower elevations (33)]. More rain (i.e., less water limitation) at higher elevations may then diminish the indirect effects of precipitation on nest success for birds living at high elevations. The strength of precipitation responses should vary among biomes (26) being strongest in those biomes where animals are more exposed to the elements and less buffered by forest canopies or urban structures, such as deserts and grasslands. Additionally, in more open biomes, precipitation has dramatic and relatively rapid consequences for the structure and function of plant communities (27) which may lead to differences in responses via indirect vegetation-mediated processes.

Avian responses to precipitation will also likely vary with species-level attributes. If food availability mediates reproductive responses to precipitation, the degree of dietary specialization should influence species’ sensitivity to fluctuations in food availability, with generalists being more buffered by those effects than specialists. Nest type and height should mediate direct consequences of high rainfall in at least two ways. Nests that are most exposed and those on or near the ground are most likely to experience physical destruction during storms. Furthermore, young birds in those exposed nest types will likely suffer the strongest adverse thermoregulatory consequences of rain (34–36). Additionally, modes of nestling development and parental care may mediate both direct and indirect response to precipitation. Species having the longest developmental times, altricial young, and fewest adults incubating and caring for young should be those least able to mitigate the effects of adverse environmental conditions on their young.

To date, no study has elucidated the causes of divergent reproductive responses to precipitation by quantitatively testing alternative mechanistic hypotheses. We filled this gap by synthesizing the global literature to explain how and why avian reproductive success varies in response to precipitation. We searched for reports of associations between precipitation and reproductive success in birds from around the world. We extracted data on the nature and magnitude of responses to precipitation corresponding to 1) daily metrics (e.g., daily nest survival rates), 2) metrics of reproductive success measured over the nesting period (e.g., nest success), or 3) productivity (e.g., number of fledged young). We analyzed each of these groups of response variables separately (see SI Appendix for individual models), but due to limited sample sizes, we also grouped metrics in a general assessment of reproductive success (Fig. 1). We assessed reproductive responses to precipitation that fell over three time frames: direct (i.e., rain that fell during the incubation and/or nestling periods), short-term indirect (i.e., measures spanning the whole reproductive season such as total breeding season rainfall and precipitation prior to nesting), and longer-term (i.e., lagged measures such as annual precipitation and rainfall lagged 2 y) (Fig. 1). We incorporated both biogeographic (e.g., elevation, latitude, biome) and species-specific (i.e., life history traits) covariates to elucidate the mechanisms underlying varied avian responses. We conducted both “vote-counting” comparative analyses that included the largest number of published relationships using multiple linear regression models, and phylogenetic meta-analyses that accounted for phylogenetic relatedness of taxa to account for evolutionary histories. These analyses elucidated the relative importance of each covariate in shaping avian responses to precipitation.

Fig. 1.

Hypothesized relationships between precipitation and reproductive success and the mechanisms linking drivers to responses. The temporal scale of precipitation (read left-to-right, representing short-term during the nesting attempt, to medium term during the nesting season, to rainfall averaged over longer time scales or lagged relative to reproduction) is hypothesized to influence the nature of the response due to a shift from direct effect to indirect effects. Direct hypothesized mechanisms are the thermogenic costs of wetting and/or physical destruction of nests due to inundation of nesting substrate. Indirect hypothesized mechanisms include precipitation-driven changes in vegetation structure in ways that influence nest vulnerability (to weather or predation), changes in predator abundances, or changes in prey availability. Under each of the potential mechanisms (columns), filled cells indicate predicted relationships with either biogeographic or species-level covariates (rows). Whereas increased precipitation leads to a directional prediction (↓ or ↑ in the figure) via the direct and vegetation (more rain = more growth) mechanisms, increased precipitation may lead to idiosyncratic responses in other trophic-mediated responses (X in the figure). Bolded font indicates results from either or both sets of analyses that were consistent with the predicted relationships. Finally, different types of reproductive metrics may vary in their response to precipitation depending on temporal scale. We analyzed each of three groups of variables separately (see Supplemental Materials), but for most analyses, we report on the aggregated reproductive success response.

Results

Our searches produced 865 sources (SI Appendix, Fig. S1); after systematic screening for sources that described the nature of the relationship between precipitation, the time period over which weather was summarized, and a nest-related metric (Materials and Methods), our dataset included 175 sources for vote counting and 89 sources for phylogenetic meta-analyses of all combined metrics of reproductive success (SI Appendix, Fig. S1). The dataset included 141 species (SI Appendix, Table S1) from 60 families (SI Appendix, Fig. S2) representing a wide variety of biomes and biogeographic attributes (SI Appendix, Fig. S2 and Table S1). The sources yielded 73 studies of daily reproductive metrics (e.g., daily nest survival), 166 reports of full nesting period metrics (e.g., nest success), and 56 reports of productivity metrics (e.g., number of young fledged). In analyses that accounted for evolutionary histories, we detected minimal phylogenetic signal in predictors for daily survival rate (Blomberg’s k = 0.12), nest success (Blomberg’s k = 0.55), or total reproductive success (Blomberg’s k = 0.18). The SI Appendix include full model summaries (SI Appendix, Tables S3 and S4), species-level attributes of studies included in analyses (SI Appendix, Table S1), and correlation matrices for variables included in each dataset (SI Appendix, Figs. S6–S11). The funnel plot for studies included in the phylogenetic meta-analyses indicated little publication bias (SI Appendix, Fig. S12).

Results were relatively congruent among the vote-counting analyses (i.e., multiple linear regression [MLR] models) and phylogenetic meta-analyses (i.e., multivariate mixed effects models [MMEM]). The temporal scale of the precipitation metric was related to the nature of the relationship between reproduction and precipitation in ways we predicted in both vote counting (Figs. 2 and 3A) and phylogenetic meta-analyses (Fig. 3B). When short-term precipitation during the nesting attempt was correlated with nest success, more than half of those responses were negative and less than a quarter were positive. In correlations between precipitation lagged relative to reproduction, responses shifted to predominantly positive responses [χ² (4, N = 283) = 25.9, P < 0.001; SI Appendix, Fig. S4]. Parameter estimates from both model sets (vote-counting and phylogenetic meta-analyses) reflected this pattern; relative to the short-term precipitation metrics ($\widehat{\beta }$ = −0.171, SE = 0.128), reproductive success was more positive following lagged precipitation in MLR models ($\widehat{\beta }$ = 0.429, SE = 0.143; SI Appendix, Table S3) and in the MMEM analyses (short-term precipitation metrics, $\widehat{\beta }$ = −0.044, SE = 0.007; lagged metrics, $\widehat{\beta }$ = 0.050, SE = 0.007; SI Appendix, Table S4).

Fig. 2.

Reproductive responses to precipitation from original studies (i.e., full dataset used in vote counting and phylogenetic meta-analyses). Nature of reproductive responses to precipitation depends upon the time scale over which rainfall is measured. The majority of responses to rain that falls during reproductive events are negative (red) which likely reflects mechanisms involving direct costs. As rainfall metrics become less tightly coupled with reproduction, more responses become positive (green) reflecting indirect, likely trophically mediated mechanisms.

Fig. 3.

Parameter estimates (dots) and 95% CIs (lines) from vote counting analysis (A) and phylogenetic meta-analysis (B) assessing how the timing of precipitation (blue), biogeographic correlates (green), and species-level attributes (yellow) mediate relationships between precipitation and all metrics of avian reproductive success. A positive parameter estimate indicates an increase in reproductive success following precipitation, while a negative parameter estimate indicates a negative response to rainfall following precipitation. The reference level was the breeding season for precipitation timing, “aquatic” for biome, “generalist” for diet, “minimal” for nest shape, “altricial” for nestling development, and “biparental” for parental care. Asterisks indicate significance level: 0.05 (*), 0.01 (**), and 0.001 (***). Nestling development was excluded from the phylogenetic meta-analysis due to collinearity.

The consequences of breeding season precipitation for reproductive success consistently depended on elevation and latitude but not biome. Reproductive success decreased following precipitation at higher elevations relative to lower elevation (MLR: $\widehat{\beta }$ = −0.176, SE = 0.060; MMEM: $\widehat{\beta }$ = -0.022, SE = 0.045). In the phylogenetic meta-analyses (i.e., MMEM), responses across latitudes were relatively strongly negative ($\widehat{\beta }$ = −1.724, SE = 0.103), while in the vote-counting (i.e., MLR) models, there was no relationship between latitude and reproductive responses to precipitation. Importantly, however, the relationship between elevation and reproductive responses to precipitation depended on latitude; at higher latitudes, responses to precipitation became more positive relative to low latitudes (interaction term, MLR: $\widehat{\beta }$ = 0.219, SE = 0.110; MMEM: $\widehat{\beta }$ = −1.015, SE = 0.061). At low latitudes, responses to precipitation became more positive at higher elevation, but the reverse was true at high (absolute) latitudes (Fig. 4B). Few responses to precipitation depended upon biome; in the MMEM analysis of daily nest survival, responses were contrary to predicted patterns; birds nesting in more open biomes experienced higher daily nest survival following rainfall than birds in other biomes ($\widehat{\beta }$ = 1.686, SE = 0.848), but this relationship was not present in other models.

Fig. 4.

(A) Distribution of elevations and latitudes represented in phylogenetic meta-analyses (i.e., MMEMs). Each dot is an original analysis, and dots are transparent (i.e., darker dots indicate multiple studies at that latitude/elevation, while lighter dots indicate fewer studies at that latitude/elevation). Bluer dots indicate studies occurred at higher latitudes, and pinker dots indicate lower latitudes as indicated in the color scale to the R of the figure. (B) Predicted reproductive success from phylogenetic meta-analyses (i.e., MMEM) following precipitation along elevational and latitudinal gradients. Elevation is measured in meters above sea level, and latitude is measured in absolute decimal degrees. At low elevations, birds respond similarly to precipitation (both positive and negative) regardless of latitude, but responses to rain diverge at higher elevations. At low latitudes (reddish line), reproductive responses are more positive following precipitation, whereas at high latitudes (blue line), reproductive responses are more negative following precipitation.

In neither analytic framework did reproductive responses to precipitation depend on the species-level attributes we included. Additionally, results of separate analyses of the three reproductive metric types (SI Appendix) were generally congruent with the analysis of aggregated reproductive success metrics. The only exceptions to these two generalities were that relationships between daily nest survival and precipitation varied by nest type in the MLR analysis of daily nest survival. However, results were contrary to those predicted; protected nests experienced lower daily survival following rain than those that were those more exposed to the elements ($\widehat{\beta }$ = −1.161, SE = 0.512). Parameter estimates for all models appear in SI Appendix, Tables S3 and S4.

Discussion

The results of this study imply that precipitation influences avian reproductive success via both direct and indirect mechanisms, and the nature of the response depends on how tightly matched the timing of precipitation metric is to nesting phenology. Short-term precipitation events that occur during a nesting attempt mainly have adverse consequences for avian reproduction, whereas precipitation that falls well before reproduction or integrates precipitation over longer time periods mainly has positive consequences for avian reproduction. Precipitation metrics summarized throughout the breeding season were mixed in their associations with nest success, either because climatic predictors at coarse resolutions have low predictive accuracy in areas where weather is temporally variable (37), or these time scales represent a mix of direct and indirect mechanisms. The fact that responses changed sign with time scale is consistent with responses reflecting different mechanistic links between rain and reproduction. Short-term responses measured during a breeding attempt such as precipitation during the incubation, nestling, or total nesting periods, likely represent direct, negative consequences of rain on avian energetics and/or physical destruction of nests due to storms. Indeed, other authors have postulated that extreme precipitation may better predict changes in reproductive vital rates than averages (38). In contrast, longer-term responses such as precipitation in the spring or prior year likely primarily represent vegetation-mediated positive responses to increased precipitation. Most studies linking precipitation to animal demography, distribution, or behavioral ecology either assume that underlying drivers are food mediated or do not explicitly explain the basis of predicted links between climate and response variables (19, 39–41). However, if food availability was the primary route by which precipitation influenced reproduction over longer time scales, then we would have expected responses to be stronger in dietary specialists than generalists who are more able to take advantage of altered prey availability. The divergent responses to precipitation evident in this study need not be related to food; they are consistent with animals having hygric niches (8) and focal studies occurring in different locations along the gradient of rainfall conditions under which each species can thrive; this study provides strong evidence consistent with key predictions of that framework.

Responses to rain depended on biogeographic context in ways that implicate thermogenic mechanisms—i.e., the processes involved in maintaining endothermy—as the most important in mediating reproductive responses. Reproductive success declined more strongly following precipitation at higher (colder) elevations or latitudes in vote counting and phylogenetic meta-analyses, respectively. Species-specific studies exemplify such interactions at local scales; Horned Larks (Eremophila alpestris) breeding in Canadian mountains experienced 8 to 9 times higher nest failure during storms that occurred during cold weather but little consequences of storms during warm weather (42). Furthermore, in both cases, the nature and magnitude of reproductive responses depended on the interaction between elevation and latitude; no other mechanistic link predicted such biogeographic patterns.

The interactive nature of latitude and elevation in mediating rainfall responses reveals the spatial complexity in the effects of precipitation and temperature on endotherms. At high latitudes, seasonality is defined by dramatic changes in temperature far more so than by fluctuation in precipitation. Furthermore, lower temperatures are exaggerated at higher elevations, and birds breeding at high latitudes are phenologically, behaviorally, and physiologically constrained by temperature (43–45). Our results are consistent with precipitation exacerbating thermal constraints; at high latitudes, the consequences of precipitation became more negative with increasing elevation (Fig. 4B). Relationships between elevation and precipitation responses were opposite in the tropics, however. At low latitudes, temperature is relatively consistent throughout the year, and seasonality is characterized by variation in rainfall (46, 47). While we predicted more negative responses to precipitation at high elevations in the tropics as well as in temperate-breeding species, the reverse was true. It is possible that this result is mediated by indirect mechanisms, such as increased food availability or top–down consequences of rain-related predator responses. Alternatively, it is possible that long-term exposure to challenging thermal conditions has resulted in selection on traits that mitigate costs. For example, birds living at high elevations have evolved feathers having proportionately more downy plumes that provide greater insulative capacity relative to their low elevation counterparts (48). We encourage future studies to test these hypotheses in experimental settings, however. In particular, we urge more studies from the tropical regions, which was an underrepresented region in our dataset (Fig. 4A). As more studies are published from low latitudes, interactions between latitude and elevation can be interpreted with more certainty.

We did not explore interactions between latitude, elevation, and the various species-level covariates in our analyses due to lack of clear a priori hypotheses and sample size limitations. Similarly, the complexities surrounding differences in consequences of snow vs. rain, the ways in which wind may exacerbate precipitation responses, and nonlinear effects of rainfall reflecting threshold responses were beyond the scope of this study. However, these avenues of research all merit further investigation, particularly in light of the variation in responses reported here. For example, the net negative responses to short-term rainfall is consistent with the possibility that wind associated with storm events may help explain aspects of avian reproductive success either via increasing thermogenic costs or by reducing foraging success (49). These results go a long way toward understanding why we often see divergent responses among sites or taxa to temporal variation in rainfall, ruling out many of the potential mechanisms that could explain such differences. The next steps are to build upon this work in studies explicitly designed to reveal the nuances of rainfall responses and clarify complex or interactive effects involving abiotic drivers. If the different responses to precipitation in high-elevation taxa at different latitudes reported here hold true in future studies, results would be congruent with theories about indirect, biotic selective pressures constraining organisms at lower latitudes and direct, abiotic factors exerting strong selective pressures at higher latitudes (43, 50).

Across analyses, birds’ responses to precipitation were inconsistently dependent upon biome and nest structure, both factors that contribute to nest microhabitat. In one analysis (phylogenetic meta-analysis, daily nest survival), those living in relatively exposed contexts (“open” relative to “forest”) responded more positively to precipitation, while in another analysis (vote counting, daily nest survival), species that occupy more protected, or sheltered, nests (relative to limited or no nest structure) responded more negatively to precipitation. While these results contradict fundamental ideas regarding the function and importance of nest microhabitat in protecting young, we caution the inferences drawn from this particular result as relationships involving nest type and biome were not evident in most of our analyses. Contrary to many studies, we did not find support for food availability underlying relationships between rainfall and reproduction; there were no consistent differences among dietary guilds in the magnitude or nature of reproductive responses. Prey and predator populations may well exhibit contrasting responses to precipitation among vital rates, years, and regions, masking food- and predation-mediated responses. Perhaps even more surprising, we did not find evidence of body size underlying relationships between rainfall and reproduction because body size is mechanistically linked to thermoregulatory capacity. One potential explanation is that body size was highly correlated with nest structure and often correlated with nestling development and/or parental care (SI Appendix, Figs. S5–S11). These findings suggest that there may be common suites of traits in avian reproduction that help smaller species offset the energetic demands of thermoregulation.

The lack of clear support for any of the indirect mechanisms is largely due to the absence of evidence for species-level covariates being important in mediating responses. This lack of response is particularly surprising given the implications of nest structure, developmental mode, and parental care for thermoregulation (51, 52). We expect that under certain environmental conditions, each of these proposed covariates is important in modulating the impacts of precipitation on nest success. Additionally, it is possible that analyses with more granular measures of rainfall intensity such as the magnitude of rain relative to long-term means or methods that identified thresholds, differentiating extremes of drought or rainfall, or more fine-scale partitioning of categorical groups of species-level traits, some additional differences may become apparent. Unfortunately, too few studies reported rainfall metrics with sufficient detail to adopt such approaches in this study—we urge future researchers to be as explicit as possible in their reporting to enable such refinements. It is also possible that indirect mechanisms tend to be more context-specific than the direct mechanisms. In retrospect, given the diversity of avian reproductive behaviors, many of which have evolved in the context of mitigating abiotic or biotic challenges, this lack of consistent associations may not be a surprise (53).

Birds are declining across most guilds and biomes (54), and understanding the relationships between weather, environmental characteristics, and life history traits is required if we wish to predict demographic consequences of future climates (25). Results of this study imply that species living at high elevations and latitudes will be more affected by future changes in precipitation regimes than those living at lower elevations or latitudes. Additionally, species on the far ends of the spectrum of breeding season precipitation (including very light and very heavy precipitation) or near the edge of their species-specific hygric niches (8) will be more affected by future changes in climate regimes than those living in places with more average rainfall. Species responses to climate change can be difficult to predict because species distributions change in response to climatic and nonclimatic factors (55). However, understanding responses to climate is the basis for projection models and animal conservation initiatives (56). This study provides concrete evidence for the nature of bird responses to current variation in an understudied axis of climate; it can help predict reproductive consequences of future climatic conditions while informing basic ecological knowledge regarding the mechanisms underlying endotherm responses to precipitation.

Materials and Methods

Literature Search and Inclusion Criteria.

We searched three databases for published studies relating precipitation and reproductive success. We searched the Web of Science and Scopus databases for titles and abstracts using the following search criteria: [bird* AND (nest*) NEAR (succe* OR surviv*) AND (precipit* OR rain OR rainfall)] and the Wildlife & Ecology Studies database using the following search criteria: [SU bird* AND SU nest* AND (succe* OR surviv*) AND (precipit* OR rain OR rainfall)] on August 29, 2023. We did not include “drought,” “storm,” or other search terms reflecting specific extremes of rainfall because a) we had now a priori predictions relevant to those terms, and b) we sought to avoid binary categorizations of rainfall which obscure the linear relationships we hypothesized would better reflect biological reality. Likewise, we did not separately consider “snow” as distinct from precipitation. Snow has insulative properties that can result in distinct consequences for endotherms, with snow buffering plants and animals from adverse effects of cold temperatures while rain typically exacerbates it. Furthermore, snow may have consequences for animals while it is a solid, or following melt, making it subject to phenological processes not applicable to rain (57), thus reducing our capacity to identify consequences for reproductive success across a wide range of published studies with variable methodologies across biomes. We read the abstracts for all journal articles, dissertations, conference papers, and books obtained from these searches and excluded sources that did not mention both a weather metric and reproductive success within the abstract. For the remaining sources, we screened the source and excluded those that did not include a metric of precipitation as a predictor nor a reproductive success metric as a response variable in an original analysis. We also excluded analyses that reported a quadratic relationship (only 4 studies that also met the other criteria) because a) there were too few studies to effectively analyze separately, and b) they imply potentially different constraints at each end of the rainfall gradient.

We included sources that referenced precipitation in any form (e.g., rainfall or snowfall), but excluded sources that assessed only relationships between humidity and reproductive success as those would involve a different set of mechanisms than those we considered. We also excluded sources that assessed the consequences of river or lake levels on avian reproduction, as these depend on factors that may occur outside the geographic scope of the original study (e.g., snowmelt from distant mountains, site hydrology). We excluded sources that used year as a proxy for precipitation, as multiple environmental variables vary annually. We excluded sources that pooled estimates across sites or species and those that did not describe the nature (i.e., positive or negative) of the relationship between precipitation and reproductive success (e.g., included precipitation in analytical model but did not describe the nature of the relationship or provide an estimated effect size).

We included sources that reported metrics related to nest success (e.g., daily nest survival, hatching success, fledging success), but excluded sources that included a combined metric of nest or brood success with postfledge survival, and those studies that reported only counts of clutch size or brood size. If studies included separate analyses for postfledge survival and other methods of reproductive success (e.g., egg survival), we included only metrics that did not include postfledge survival because nestlings and fledglings incur different risks and have different anatomical and behavioral strategies to cope with these risks. For example, fledglings often have more developed feather structure, are no longer affected or protected by their nesting structure or substrate, and have more mobility to escape predators. Therefore, our a priori predictions for life history and environmental drivers of postfledge survival would differ from those experienced at the nest. We excluded sources that reported only reproductive phenology, vital rates of young birds following fledging, and vital rates of adults. If sources referenced reproductive success estimates from another study, we obtained the original study to include in our analysis. We made a list of all citations within our included sources and repeated the review process on the referenced articles to obtain as large a sample of studies as possible.

Predictions and Data Collection.

We tabulated biogeographic and species-specific covariates designed to distinguish among the alternative mechanistic hypotheses regarding the ways that rainfall could influence avian reproduction (Fig. 1; see “Hypotheses and Predictions” section of SI Appendix for more detailed descriptions of the relationships hypothesized below and SI Appendix, Table S1 for all species, the covariate values, and references for those data used in this study). The period over which precipitation occurred in relation to the nesting period (i.e., precipitation timing) allowed us to distinguish whether responses resulted from direct or indirect mechanisms. Thermoregulatory-mediated nest failure should manifest during or immediately following precipitation events because physiological responses that lead to hypothermia or nest abandonment occur on the order of hours. Similarly, destruction due to flooding would lead to nest destruction in the immediate aftermath of heavy rain. Conversely, we expected responses mediated by indirect mechanisms such as vegetation structure, predator abundance, or prey availability to be more strongly correlated with longer-term metrics of precipitation (summarized over seasonal or annual periods, or occurring prior to the nesting season) to be more strongly associated with nest success due to the increased time required for plant, predator, or prey populations to change in ways that would affect birds.

To assess relationships between precipitation timing and responses to precipitation, we recorded the precipitation variables used in the original study. We coded studies as belonging to one of three temporal scales relevant to how and when they would affect nest success. “Direct” precipitation metrics were those measured at any point throughout the nesting period, such as daily rainfall throughout the entire nesting period or daily rainfall during incubation. “Breeding season” metrics were those measured at the temporal scale of the breeding season (e.g., breeding season total, or precipitation prior to egg laying). “Lagged” variables were those that occurred prior to the breeding season (e.g., winter, annual, or previous breeding season’s precipitation).

The different mechanisms potentially underlying the effects of precipitation on avian reproductive success predict varying biogeographic and/or species-specific correlates (Fig. 1). We coded elevation, latitude, and biome from information provided in the original source articles. We recorded the mean elevation of the study site (meters above sea level) and absolute latitude (decimal degrees) as reported by the source authors. We used absolute latitude because we predicted responses to precipitation would vary with distance from the equator and not on a positive/negative latitudinal axis. If the elevation or latitude were not reported in the source, we estimated the mean latitude and elevation by locating the study site on Google Earth (earth.google.com/web/) based on the site description. We coded the biome in which the study took place based on site descriptions in the source articles, coding it as one of four types: 1) “aquatic” for coastal and wetland habitats such as marshes, beaches, and floodplains; 2) “open” for grasslands, agricultural areas, tundra, and desert habitats, all characterized by having little tree cover; 3) “wooded” for sites in forest and in areas having extensive shrub cover; 4) “urban” for sites in cities and zoos, typically also characterized by high tree cover and extensive human-built structures providing potential shelter.

We tabulated all species-specific covariate data from individual species accounts in the Birds of the World online (58). We categorized prey types into the following categories and noted the presence or absence of each in the diet of each species: vegetation, fruit, invertebrates, or vertebrates (i.e., birds, reptiles, mammals, fish, or carrion). We grouped seeds and leaves under one category (“vegetation”) because the species that primarily eat leaves or other vegetative matter (i.e., geese) also eat seeds; therefore, these were a natural grouping. We then coded each species as a “specialist” if they typically consume one prey type, or a “generalist” if they take two or more types of prey. We coded nest shape as belonging to one of three categories based on descriptions and photos within the study or Birds of the World (58): 1) “minimal” for relatively flat nests or nests with little to no nesting substrate (i.e., scrape and platform nests); 2) “structured” for nests with vertical nesting substrate (i.e., cup, bowl, and adherent nests) but open to the top; 3) “protected” for nests with substrate atop the majority of the nest (i.e., cavity, burrow, domed, and pendulous nests). We coded nest height as either 1) on the ground, or 2) more than 1 m off the ground. For species that nest both on and above the ground, we coded it according to what authors of Birds of the World species accounts considered the most common height class. We coded a species as being “altricial” if they exhibited any type of altricial or semialtricial nestling development, or “precocial” if they exhibit precocial or semiprecocial nestling development based on Birds of the World species accounts. To characterize the amount of parental care, we coded species as one of the following: 1) “uniparental” for species in which one parent primarily raises the brood; 2) “biparental” if two individuals share roughly equally in raising the brood; or 3) “cooperative” if three or more individuals raise the brood. If species typically exhibited both biparental and cooperative care, we coded them as “Multiparental.” Parental care was thus treated as a binary variable in analyses (uniparental vs. multiparental). We acknowledge that broad-brush categorizations of species-level traits which are inherently continuous in nature do not capture the richness of avian phenotypes; such simplification is an unfortunate cost of the generality gained by conducting studies of multiple taxa at global scales. We calculated the mean body size (grams) for adult males and females and average clutch size, as reported by Birds of the World species accounts. We Z-transformed all continuous variables (i.e., elevation, latitude, average body size, and average clutch size). A complete list of the species included in this study, the coding of each covariate, and references for both the main precipitation–reproduction relationship and covariate data appears in SI Appendix, Table S1.

To create a response variable for each analysis within each study, we recorded the nature of the relationship between precipitation and reproductive success (i.e., positive or negative) for the given study. We recorded the effect size (i.e., parameter estimate, model-averaged parameter estimate, or median of the posterior distribution) of the relationship between precipitation and the nest-related metric, along with the associated SE or CI, if reported. We recorded sample size, response variable (e.g., daily nest success, nest survival), and metric of precipitation (e.g., annual rainfall). In addition to analyzing all reproductive success responses together, we split our dataset into subsets based on the temporal extent of response variables: 1) daily metrics of reproductive success (e.g., daily nest survival rates; hereafter: “daily survival rate”), 2) metrics of reproductive success throughout the nesting period (e.g., hatching success, nestling survival, overall nest success; hereafter: “nest success”), and 3) metrics of productivity (e.g., number of offspring fledged; hereafter: “productivity”). Although sample sizes were limited when we divided studies in this way, we analyzed these three groups separately because we expected precipitation might have different effects over distinct time periods reflecting different mechanistic links (see SI Appendix, Tables S3 and S4 for results of these analyses). We also fit models including all metrics together (hereafter: “reproductive success”) to fit models with the maximum power to identify factors mediating rainfall-driven effects on reproductive success. The dataset and code are publicly available (59).

Vote-Counting Analyses.

Given that many studies did not report effect sizes, we first conducted vote-counting analyses to maximize sample size within our synthesis. We included all studies for which we could obtain information about the precipitation variable tested and nature of the response (i.e., positive, negative, or no response). We assigned a “1” to all studies that reported a positive relationship between precipitation and reproductive success, a “0” to all studies that reported no relationship, and a “−1” to all studies that reported a negative relationship between precipitation and reproductive success. We fit multiple linear regression models (MLR); one for the entire dataset and one for each response variable group (i.e., daily survival rate, nest success, and productivity). We fit MLRs using the package stats in R v. 4.3.1 (60). Each model included all predictors (i.e., precipitation timing, latitude, elevation, nestling development, body size, clutch size, parental care, nest shape, nest height, biome, and diet breadth), unless variables were highly correlated (r > 0.60), in which case we excluded the predictor that correlated with the most variables within the subsetted data. Correlation matrices for variables included in each MLR are available in SI Appendix, Figs. S5–S8.

Phylogenetic Meta-Analyses.

For a subset of studies, we fit phylogenetic meta-analyses to account for the magnitude of the response, the sample size of the original dataset, and the phylogenetic nonindependence of species (hereafter “meta-analysis dataset”). We included any studies in the meta-analysis dataset for which we could obtain an effect size (i.e., parameter estimate), associated SE, and sample size. We excluded studies that included an effect size for nest failures instead of nest success, because these effect sizes indicate the opposite relationship as the other studies, therefore making them incomparable.

We downloaded 1,000 phylogenetic trees for the species represented in our meta-analysis dataset from vertlife.org (61). We calculated Blomberg’s k (62) to assess the strength of the phylogenetic signal across the phylogenetic trees. We calculated variance for each study based on the effect size, effect size SE, and sample size; studies with smaller variation among replicates or larger sample sizes carried more weight in the model. We created a funnel plot to assess publication bias in studies included in the phylogenetic meta-analyses. Funnel plots may appear asymmetrical due to publication bias, in which positive and/or negative results are more likely to be published than nonsignificant results (63). To avoid potential phylogenetic bias due to the nonindependence of effect sizes for closely related species in our meta-analysis dataset (64), we fit sets of multivariate mixed-effects models using the software metaphor (65) to the entire meta-analysis dataset as well as separate models of daily survival rate and nest success as the response variables. There were too few studies to analyze productivity in a phylogenetic context. Each set of models included 1,000 models (one model fit to each of the 1,000 phylogenetic trees). Each model included all predictors (i.e., precipitation timing, elevation, latitude, biome, diet breadth, nest shape, nest height, nestling development, parental care, body size, clutch size), unless variables were highly correlated (r > 0.60), in which case we excluded the predictor that correlated with the most variables. Correlation matrices for variables included in each MLR are available in SI Appendix, Figs. S9–S11.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Original data have been deposited in Zenodo (https://doi.org/10.5281/zenodo.18625703) (59).