Supplemental feeding of wild animal populations is popular across many areas of the world and has long been considered beneficial, especially to avian taxa. Over four billion dollars are spent by hobby bird feeders in the United States each year alone. However, there is mixed evidence whether wildlife feeding is beneficial, including when it is implemented as a conservation management tool, a targeted experimental design, or an avocation. Much of the current evidence suggests that providing supplemental food is advantageous to the reproductive output and general survival of focal taxa. However, many of these studies are limited in scope and duration, leaving possible negative impacts unaddressed. This is particularly true regarding passive backyard feeding, which describes the majority of supplemental feeding, including the immense effort of millions of public enthusiasts. Here we show that winter supplemental feeding prior to reproduction had no significant impact on a range of reproductive parameters in a resident, montane passerine species, the Mountain Chickadee (Poecile gambeli). This population resides in an intact natural environment with no exposure to supplemental food beyond our experimental treatments, and individual birds were tracked across six years using radio frequency identification technology. Our results add to the growing evidence that supplemental feeding alone, isolated from the effects of urban environments, may have little to no impact on the population dynamics of some avian taxa.

Subjects and study site

All data for this study were collected from December 2015–August 2021 at Sagehen Experimental Forest (Sagehen Creek Field Station, University of California, Berkeley) in the Sierra Nevada, USA. This population has been monitored year-round at two elevational sites (high: ca. 2400 m; low: ca. 1900 m separated by 3.49 km) since 2013 (Kozlovsky et al., 2018, Croston et al., 2016). Nest boxes and feeders were placed along service roads within the forest (5.87 km of non-linear road at low elevation and 4.83 km at high elevation), but nest boxes cover a larger area compared to the area with available feeders. Adult and juvenile chickadees were captured with mist nets during the fall and early spring at feeders and by hand at nest boxes during the summer (e.g. Croston et al., 2016, 2017, Tello-Ramos et al., 2018). Birds were banded with unique combinations of colored PIT-tags and plastic or aluminum leg bands. Individuals that were not first captured as nestlings were assigned an approximate age (adult or juvenile) at time of initial capture based on multiple plumage characteristics and sexed via physiological evidence if possible (e.g. brood patch or cloacal protuberance; Pyle, 1997). Individuals in their first year of life were classified as juveniles and those older were classified as adults. Nestlings were captured in nest boxes and banded with U.S. Geological Survey aluminum bands; if recaptured as adults, they were banded with an additional plastic color band and a PIT-tag. Birds are highly resident and do not disperse between sites. A total of 4,264 nestlings have been banded in nest boxes since 2013 and not a single individual was detected breeding at a different elevation than it was initially banded during the duration of the study.

Food supplementation

We provided chickadees with a constant supply of black oil sunflower seeds during the winter months (December through April) of each year. While multiple feeders were used during early autumn to capture and band chickadees, only two feeder locations per elevation were continually stocked with supplementary food throughout the winter. All other feeders were closed before winter. These permanent locations were the only feeding sites across approximately 5 km (at each elevation) and some birds regularly utilized the feeders while other more distant individuals did not visit at all. These geographic details allowed us to compare reproductive performance of birds with and without access to supplementary food during the winter prior to reproduction.

Winter feeding locations were associated with feeder arrays used for testing cognitive abilities and included eight feeders arranged equidistantly in a circular array format on a 122 x 122 cm square frame. Visitation of feeders was detected via RFID readers that record when birds marked with a PIT-tag land on a feeder’s perch (Croston et al., 2016, 2017; Tello-Ramos et al. 2018, Bridge et al., 2019). Tagged birds detected at these feeders were categorized as individuals with access to supplemental food and tagged birds that never visited the feeders were categorized as those without access to supplemental food. Birds trapped at feeders in the fall (September-November) who did not continue visiting the main feeding locations were classified as individuals without access to supplemented food (non-supplemented: 109 and supplemented: 128). Many of the adult individuals banded at nest boxes were far away from the main feeding locations and were never detected at these locations (non-supplemented: 83 and supplemented: 45). Thus, our sample of birds represents some individuals that were never near our feeders, some that may have been near our feeders but did not visit them, and others that visited the feeders. There may have been birds that remained undetected during the winter months due to pilfering seeds underneath the feeders or from the caches of those visiting feeders but we expect this number to be quite small. The coniferous forest habitats at each elevation are extremely homogenous so it is unlikely that habitat quality influenced the choice to visit feeders.

We did not incorporate individual visitation rates into the study as our data are inherently incomplete due to the timing of when an individual was captured in each season (early or late fall or winter) and due to limited time in which the feeders recorded visits by the tagged birds (feeders were active all winter, but we only collected visitation data during the specific interval of cognitive testing), making this comparison problematic. Visitation numbers also reflect both consumption and caching in this species which cannot be easily disentangled. Our categorization tests the idea of whether the presence and access (birds that visited the feeders even a few times were considered to have full access to the feeders while birds that never visited the feeders were considered to have no access to supplementary food) of human-provided food during winter has any effect on reproduction in the following season. We also conducted a more conservative analysis that only included individuals never detected at feeders and individuals that completed at least 20 trials during cognitive testing (a trial results in a bird getting food during a visit to the feeder array, Croston et al. 2016, 2107; Tello-Ramos et al. 2018). All birds that completed at least 20 trials were visiting the feeders regularly for several weeks leading up to cognitive testing and most exceeded 20 trials during their participation (Trials: range 20–4,444, mean: 603.35, median 337), so this analysis amplified potential differences between pairs that regularly visit the feeders and those that and did not access supplemental food.

Breeding

We monitored chickadees for breeding behavior starting in mid-April each year. Chickadees began nesting approximately one to two months after food supplementation was completely removed. Mountain chickadees are secondary cavity-nesting birds that readily use human-made nest boxes (McCallum et al., 2020). Nest boxes were monitored for the onset of nest building, first egg date, onset of incubation and hatch date on a weekly to biweekly basis (Kozlovsky et al., 2018, Pitera et al., 2021, Sonnenberg et al., 2020). After the onset of incubation was detected, nests were targeted for monitoring based on expected incubation duration to limit disturbance and allow for precise detection of hatch dates. Thus, first egg dates are approximations with an error rate of ±1 day but all hatch dates are exact. After nestlings hatched, any unbanded parents were captured and banded between eight and twelve days post-hatch. Nestlings were counted, banded, and weighed using an electronic scale at 16 days post-hatch (Kozlovsky et al., 2018, Pitera et al., 2021, Sonnenberg et al., 2020). We assigned parental breeding pairs to four categories depending on their visitation to feeders in the winter prior to each breeding attempt: pairs in which both individuals visited feeders, pairs in which only one individual visited feeders (male or female), and pairs in which neither individual visited feeders. Any nest with a parent that was not banded in the previous winter was excluded.

Statistical methods

We tested whether access to supplemental food in winter influenced breeding performance of mountain chickadees using five breeding variables: first egg date, clutch size, brood size (number of living nestlings on day 16 post-hatch), mean nestling mass, and the coefficient of variation of nestling mass. We only included data from initial nests, leaving out renests within a season, as these attempts are very infrequent and often unsuccessful in this system. We modeled each breeding variable separately for our high- and low-elevation sites due to previously reported major differences in breeding performance between elevations in this population which were not the main focus of this particular study (Kozlovsky et al., 2018, Pitera et al., 2021). Analyzing elevations separately allowed us to test whether supplementary food had an effect on reproduction within each elevation. This resulted in 10 models, five for each elevation. Each model contained the breeding variable as the response variable, parent access to supplemental food as a fixed effect (four levels: both parents accessed feeders, male only, female only, and neither parent accessed feeders), breeding year as a fixed effect (six levels: 2016 – 2021) and parent age as a fixed effect (four levels: both parents adults, adult male and juvenile female, juvenile male, and adult female and both parents juveniles). Including year and parent age as fixed effects allowed us to control for environmentally driven differences in breeding across years (Kozlovsky et al., 2018), and effects of parent breeding experience (Pitera et al., 2021) previously reported for this system. We included male ID and female ID as separate random effects in each model to control for individual variation and repeated nesting attempts across years. We did not include an interaction between year and parent access to food due to small sample sizes when the data were divided this way.

We used linear mixed models to model the effects of winter feeding on first egg date, mean nestling mass, and the coefficient of variation of nestling mass. To model the effects of winter feeding on clutch size and brood size, we used generalized linear mixed models with a Generalized Poisson distribution and log link in the R package ‘glmmTMB’ due to our data being underdispersed (Brooks et al., 2017, Joe & Zhu, 2005). When modeling clutch size, we included first egg date of the nest as a fixed effect, as previous research in this system has indicated that earlier nests have larger clutches (Kozlovsky et al., 2018). When modeling mean nestling mass and coefficient of variation of nestling mass, we included brood size as a fixed effect in the model, as brood size is known to influence nestling mass in some species (Dijkstra et al., 1990). Coefficient of variation of nestling mass was log-transformed prior to modeling to improve residual fit. We tested the importance of each fixed effect in all models using type II Wald chi-square tests. When the ‘parent access to food’ fixed effect was significant (p < 0.05), we calculated the estimated marginal mean for each level using the ‘emmeans’ R package (Lenth, 2021) and calculated the significance of differences among the categories using a Tukey HSD test and Tukey correction for multiple comparisons through the ‘pairs’ function in R. We tested the residual fit of all models using the R package ‘DHARMa’ (Hartig, 2018). We calculated marginal R2 values for each model following Nakagawa et al. (2017) implemented in the R package ‘performance’ (Lüdecke et al 2021). All statistics and graphics production were performed in R version 4.1.1 (R Core Team, 2021, Wickham et al., 2019).

The program R is the only required software.