Many songbird species have experienced significant population declines, partly because of brood parasitism by the Brown-headed Cowbird (Molothrus ater), which is positively associated with increasing landscape forest cover in the midwestern United States. However, cowbirds are also experiencing long-term population declines, which should reduce parasitism pressure and thus increase productivity of host species. We used 20 years of nest monitoring data from five sites in Missouri across a gradient of landscape forest cover to assess temporal trends in the rate and intensity of brood parasitism for Acadian Flycatchers (Empidonax virescens), Indigo Buntings (Passerina cyanea), and Northern Cardinals (Cardinalis cardinalis). We evaluated whether there were concomitant changes in fledging brood size, nest survival, a combination of the two metrics (i.e., host young produced per nest attempt), and whether such changes were more substantial with decreasing landscape forest cover. Parasitism rates and intensities declined substantially during 1991–2010. Fledging brood size and nest survival rates were positively associated with landscape forest cover, confirming the fragmentation hypothesis for Midwest forest birds. Declining parasitism rates were associated with increased fledging brood sizes, with more pronounced increases as landscape forest cover decreased. Nest survival increased insubstantially across time during laying and incubation, but not during the nestling stage. The best predictor of nest survival was parasitism status, with parasitized nests surviving at lower rates than unparasitized nests. Overall, productivity increased during 1991–2010, with more pronounced increases associated with lower levels of landscape forest cover. The negative effects of cowbirds on nest survival in addition to fledging brood size in less forested landscapes suggest that cowbirds may be a primary cause of forest fragmentation effects on songbird productivity in the Midwest. Our results underscore the dynamic nature of demographic parameters, which should be accounted for in predictive models of wildlife responses to future environmental conditions.
Citation: Cox WA, Thompson FR III, Root B, Faaborg J (2012) Declining Brown-Headed Cowbird (Molothrus ater) Populations Are Associated with Landscape-Specific Reductions in Brood Parasitism and Increases in Songbird Productivity. PLoS ONE 7(10): e47591. https://doi.org/10.1371/journal.pone.0047591
Editor: Csaba Moskát, Hungarian Academy of Sciences, Hungary
Received: May 21, 2012; Accepted: September 17, 2012; Published: October 15, 2012
This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Funding: Funding for this project was provided by the U.S.D.A. Forest Service. Funding sources for the original field studies included the U.S.D.A. Forest Service, Missouri Department of Conservation, and U.S. Fish and Wildlife Service. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Many species of Neotropical migrant songbirds have experienced significant long-term population declines , . Identifying causes of the declines is challenging because the life-cycles of migrant songbirds can involve multiple habitat types across vast spatial scales . Nevertheless, conservation biologists have made major advances in our understanding of factors that limit migrant bird populations (reviewed in ). On the breeding grounds in eastern North America, habitat loss and fragmentation have decreased songbird productivity by reducing species occurrence and pairing success in small patches  and decreasing rates of nest survival . Songbird productivity is furthered hampered by brood parasitism from Brown-headed Cowbirds (Molothrus ater; hereafter cowbird), which exhibit increased abundances that result in increased rates of brood parasitism with decreasing regional forest cover in the midwestern United States . In addition to the reduced host productivity occurring as a direct result of parasitism , nests with cowbirds may experience greater predation rates because cowbird begging attracts predators . Recent studies of video-monitored nests have also shown that cowbirds are frequent nest predators , ,  and that nest predation by cowbirds increases with decreasing regional forest cover , which further implicates cowbirds as important drivers of declines in productivity associated with forest fragmentation.
Like many other passerines, cowbirds have exhibited long-term declines in population abundances ; the North American Breeding Bird Survey (BBS) indicates a survey-wide 0.6% annual decline (95% CI: −0.9, −0.5%) in cowbird abundance between 1966 and 2010 . Such declines should lead to a reduction in parasitism rates and intensity (i.e., the mean number of cowbird eggs per parasitized nest), both of which are positively correlated with cowbird abundances , , and potentially even declines in nest predation. Few studies, however, have investigated long-term patterns of parasitism. McLaren et al.  reported non-significant declines in parasitism rates in Ontario during 1970–2000 when cowbird abundances in Ontario began to decline, but the authors cautioned that the data they used were not sampled evenly across time or space which limited the strength of their inferences. Rivers et al.  documented reduced parasitism rates and intensity on Dickcissel nests (Spiza americana) compared to an earlier study at the same location , but also noted that the observed patterns may have been a consequence of an increase in the abundance of alternative hosts at their study site.
The decline of cowbirds across time could have substantial positive implications for the conservation of host species. We used a long-term nest monitoring data set to evaluate changes in parasitism rate, parasitism intensity, fledging brood size (i.e., the number of host young per successful nest), nest survival rates, and an overall productivity metric that combined fledging brood size and nest survival (i.e., number of host young per nest attempt) for three species of forest songbirds across a gradient of forest cover in Missouri, USA, landscapes that have remained largely unchanged during the past 20 years. We predicted that a statewide reduction in cowbird abundances (1.3% annual decline during 1966–2010 [95% CI: −1.9, −0.6%]; Fig. 1;) would result in reduced frequency and intensity of brood parasitism and a concomitant increase in fledging brood size. We further predicted that rates of nest survival would increase across time during the incubation and nestling stages because of reduced nest predation by cowbirds. By contrast, we predicted that nest survival should remain largely unaffected during the laying stage as cowbird abundance changed because a cowbird should lay an egg rather than depredate a nest during that stage and the species we studied rarely abandon an active nest in response to brood parasitism [21, 22, W.A. Cox unpubl. data]. Finally, we predicted that increases in fledging brood size and nest survival should result in increased overall productivity, with more substantial increases occurring in less forested landscapes where cowbirds are most abundant and parasitism rates are highest.
We compiled nest monitoring data from multiple studies that occurred on public lands in Missouri, USA, during 1991–2010 (Fig. 2). The Riparian Ecosystem Assessment and Management (REAM) project occurred at three northern Missouri riparian floodplain sites during 1994–2002. The Missouri Ozark Forest Ecosystem Project (MOFEP) is an ongoing study located in the Ozark mountains in southern Missouri where data were collected during 1991–2010 except 1996, in mature secondary successional oak-hickory forest, along roadsides, in even- and uneven-aged timber harvest plots, and in wildlife food plots. Data from the University of Missouri Baskett Research and Education Area (Baskett) were from four separate studies that occurred in old fields and mid- and late-successional oak-hickory forests during 1992–2002 and 2007–2010. All studies were permitted by the Missouri Department of Conservation and the U.S. Fish and Wildlife Service as required by law. We limited our analysis to three songbird species that were well represented at all study sites. The Indigo Bunting (Passerina cyanea, hereafter bunting) builds nests in shrubs within old fields, along forest edges, and in dense forest understory vegetation. The Acadian Flycatcher (Empidonax virescens, hereafter flycatcher) is a forest-interior species that typically nests in the branches of sub-canopy trees. The Northern Cardinal (Cardinalis cardinalis, hereafter cardinal) is a habitat generalist that nests in shrubs and trees at a variety of heights in old fields and wooded habitats. Researchers used systematic search or behavioral cues to find nests  and usually monitored them every 1–4 d until the nest fledged or failed. We considered a nest to be parasitized if ≥1 cowbird egg or nestling was in the nest at any point while being monitored. We considered parasitism intensity to be the maximum number of cowbird eggs and/or nestlings observed in a nest on any monitoring visit. To minimize the potential for bias in survival estimates, we right-censored nests with unknown or questionable fates (i.e., >4 d between the penultimate and final nest check and no evidence of fledglings or signs of predation noted) as well as those that failed because of researcher activities .
Locations of REAM (triangles), Baskett (square), and MOFEP (circle) sites in a study of temporal trends of brood parasitism and nest survival in Missouri, USA, 1991–2009. Shading indicates percent forest cover in a 10-km radius, with darker shading indicating more forest (range: <1–97%). Numbers next to each site indicate number of nests used in nest survival analysis for Acadian Flycatchers, Indigo Buntings, and Northern Cardinals, respectively, and the mean percent forest cover for nests at each site.
We obtained geographic coordinates recorded from handheld GPS units (n = 1,186) or from nest locations marked on gridded maps in a GIS (n = 959). One study at Baskett did not have exact nest locations mapped but did record subplot locations (e.g., specific old fields or forest sections); we defined the center of the subplot as the location for these nests (n = 718). The remaining nests were assigned locations based on the center of a breeding territory (n = 45), or an estimated location based on written descriptions (n = 4).
We used landscape forest cover as a metric of habitat fragmentation (sensu ). We assessed temporal changes in landscape forest cover to determine whether it was appropriate to use a single National Land Cover Database (NLCD) for all of our nests and to determine whether the hypothesized changes in parasitism rates and intensity could be due to broad-scale habitat change (i.e., increased forest cover). First, we downloaded U.S.D.A. Forest Service Forest Inventory Analysis (FIA) data (http://www.fia.fs.fed.us/) for all the counties that fell within a 10-km radius of our nests and assessed forest cover change from 1989 (data were not available for 1990–1991) to 2010. Second, we downloaded two NLCD land cover change databases (http://www.mrlc.gov/; 1992 versus 2001 and 2001 versus 2006) that correct for compatibility issues between releases and identify pixels that have changed between releases. We used the pixels within the 10-km buffers surrounding nests that changed to or from forest to calculate the percent change in forest cover.
We concluded it was appropriate to use the 2001 database to calculate landscape forest cover for all nests to avoid compatibility issues between different NLCD releases (see results). We used ArcMap 9.3  to reclassify land cover as forest (composed of deciduous, evergreen, and mixed forests, shrub/scrub [a category which included transitional forests], and woody wetlands) or non-forest (all other land types). We used the focal statistics tool to calculate percent forest surrounding each pixel and chose a 10-km radius because it best explains variation in nest predation  and is a strong predictor of parasitism risk and cowbird abundance , ,  for forest songbirds in the United States.
Our overall approach was to evaluate sets of candidate models explaining variation in nest parasitism rates, parasitism intensity, fledging brood size, and nest survival within an information-theoretic approach . We then combined model-based predictions of nest survival and fledging brood size using a bootstrapping approach to estimate the effects of forest cover and year on the number of young produced per nest attempt (hereafter productivity).
We modeled parasitism rates using logistic regression with a binary response variable (parasitized versus unparasitized) with the GLIMMIX procedure in SAS . We constructed five candidate models (Table 1) representing hypotheses explaining variation in parasitism rates and ranked models by calculating Akaike's Information Criterion (AIC) values and the difference between the top model and other candidate models (ΔAIC). We then used the ΔAIC values to calculate weights (wi) and evaluate the relative support of each model in the candidate set. All models included study site as a random effect to acknowledge the potential for correlated fates within sites and to account for site-specific variation in parasitism rates. All models except the null also included host species and nest site habitat type, both of which can influence parasitism rates , . We assigned nest site habitat types by collapsing the original researchers' designations into three categories (forest, clearcut, non-forest). If the original researcher did not designate a habitat type, we used the nest location within a GIS to assign the habitat type. Non-forested habitat included fields, old fields, forest/field edges, and roadsides. The MOFEP experiment also includes group selection and thinning harvest regimes (see  for details), but because treatments retained substantial tree canopy and do not influence cowbird abundance , we assigned all nests in those stands to the forest habitat type. We included landscape forest cover and year as fixed effects of interest in candidate models. We also included an interaction term (forest cover × year) in one model to assess the hypothesis that the decline in parasitism rates would be more substantial in fragmented landscapes compared to highly forested landscapes. We did not consider an interaction between species and year because all three species exhibit pronounced increases in parasitism rates with decreasing forest cover in Missouri (W.A. Cox, unpubl. data) likely because of increased cowbird abundances across the same gradient. We assessed the same model set for parasitism intensity using the number of cowbird eggs or young as a normally distributed response variable, an approach that is most robust to deviations from an assumed distribution when analyzing egg or nestling count data .
Because interpretation of model-averaged coefficients is problematic when some parameters occur as both additive and interactive terms in various models , and because we needed parasitism rates to produce estimates of nest survival and fledging brood size (see below), we generated model-averaged predictions rather than coefficients from the sets of candidate models. We held terms in our regression models at specified values (e.g., an equal proportion of all three species; see  for an overview of empirical versus model-based estimation) to produce predictions that were not biased by our sample, which was uneven across some covariates.
We used nests for which exact counts of host young were known to model fledging brood size as a function of forest cover and year. We used the same approach and model set as described for parasitism rates and intensity (Table 1), but because parasitized nests typically fledge substantially fewer young than unparasitized nests , all models except the null also included a term for parasitism status. We covaried values for parasitism rate, forest cover, and year according to the predictions produced from our parasitism rate analysis to produce predictions of fledging brood size.
We used the logistic exposure method  within an information-theoretic framework  to model nest survival using the NLMIXED procedure in SAS . All models (Table 2) included a random-effect term for study site and all but the null model included fixed-effect terms for the nuisance parameters (songbird species, nest stage, and habitat type) as each influences rates of survival in our study system , . We included a term for brood parasitism status in all but a true null model and a model with just the nuisance parameters because parasitized nests may be under greater risk of predation . We included a model with landscape forest cover and year to evaluate whether landscape-specific rates of predation (i.e., low nest survival in fragmented habitats) changed across time, and a term with a year × forest cover interaction to assess our hypothesis that overall predation rates would decline in more fragmented landscapes where cowbirds are more abundant but remain unchanged in highly forested landscapes where cowbirds are less abundant. We included a model with a stage × year interaction to assess the hypothesis that if temporal trends in nest survival were caused by cowbirds, they would occur during incubation or nestling stages because cowbirds should lay eggs rather than depredate nests during the laying stage (the species we studied infrequently abandon nests because of brood parasitism [21, 22, W.A. Cox unpubl. data]). Finally, we included a global model with all constitutive terms from the candidate models.
To assess temporal changes in productivity, we combined fledging brood size and nest survival estimates. To do this, we used a parametric bootstrapping approach that allowed us to incorporate the error associated with the fledging brood size and nest survival estimates . First, we generated model-averaged predictions for fledging brood size and model-averaged predictions for nest survival across years. Then for each year, we randomly selected a value for fledging brood size and a value for nest survival probability from a normal distribution of possible values that was constrained by each term's standard error. We then calculated productivity as fledging brood size × nest survival probability. We repeated this 10,000 times and treated the mean value as a point estimate of productivity and the 2.5% and 97.5% values as the confidence limits. All analyses were performed using SAS . Mean values are presented with standard errors unless otherwise indicated.
Forest cover in the 10 counties surrounding our nests increased 7% (640,628 ha to 682,725 ha) between 1989 and 2010 according to FIA data. In contrast, NLCD data from a 10-km radius surrounding each nest suggested that forest cover declined by 1.8% from 1992 to 2001, and by 0.4% from 2001 to 2006. Of the nests monitored wherein contents were reliably observed, 423 of 1,524 (28%) bunting nests, 76 of 1,021 (7%) flycatcher nests, and 79 of 367 (22%) cardinal nests were parasitized. All three species were well represented across the gradient of forest cover (Fig. 2; bunting mean: 58±1%, range: 23–95%; flycatcher mean: 70±1%, range: 23–95%; cardinal mean: 44±1%, range: 23–95%). Forest cover strongly influenced parasitism rates with a mean predicted parasitism rate of 33% (95% CI: 28–37%) for a population of nests balanced across species at 23% forest cover (the 5th percentile of observed forest cover values) versus 3% (95% CI: 2–4%) for a population of nests at 94% forest cover (the 95th percentile). A parameter for year was in the two best-supported parasitism rate models, which combined for 100% of the overall weight of evidence (Table 1). Parasitism rates differed between species and declined across time (Fig. 3a). A parameter for year was also in the top two parasitism intensity models, which combined for 98% of the overall weight of evidence (Table 1). The mean number of cowbird eggs per parasitized nest differed among species (buntings: 1.43±0.03; flycatchers: 1.09±0.04; cardinals: 1.23±0.06) and declined across time (Fig. 3b). There was a substantial effect of brood parasitism on fledging brood size; fledging brood sizes were greater for unparasitized versus parasitized nests for buntings (2.79±0.12 versus 1.47±0.14 fledglings), flycatchers (2.50±0.13 versus 1.17±0.15 fledglings), and cardinals (2.82±0.15 versus 1.50±0.16 fledglings). The best-supported model for fledging brood size did not include a covariate for year, but there was considerable model-selection uncertainty (Table 1), and model-based predictions of fledging brood size increased across time (Fig. 4) because the predictions incorporated the negative association between parasitism rates and year. The effect size of the increase in fledging brood size from 1991 to 2010 was greater for nests as forest cover declined, with a 22% increase in brood size for nests at 23% forest cover (1.83±0.09 in 1991 versus 2.35±0.19 in 2010) compared to a 3% increase for nests at 94% forest cover (2.83±0.13 versus 2.93±0.31; Fig. 4).
Model-based predictions of Brown-headed Cowbird parasitism rates (A) and intensity (B) for three songbird species in Missouri, USA, 1991–2010. Rates are for a population at the median value for forest cover (53%) and at observed frequencies of nests in three habitat types (forest, clearcut, non-forest). Rates in (B) are for a balanced population of the three songbird species. Intensity values <1 are not possible (i.e., there cannot be <1 cowbird egg or nestling in a parasitized nest), so predictions of parasitism intensity should be interpreted with regard to temporal changes rather than absolute values. Error bars in (A) and dashed lines in (B) represent 95% confidence intervals, with some bars offset for visual clarity.
Model-based predictions of fledging brood size at 5%, median, and 95% observed values of forest cover (23%, 53%, and 94%, respectively) across a balanced population of three songbird species in Missouri, USA, 1991–2010. Estimates are for a population at observed frequencies of nests in three habitat types (forest, clearcut, non-forest), with parasitism rates varying across years and forest cover levels on the basis of the parasitism rate model-selection analysis. Error bars represent 95% confidence intervals and are offset for visual clarity.
The total effective sample size  for our nest survival analysis was 33,698. There was considerable model-selection uncertainty, with the best-supported model having 24% of the overall weight of evidence (Table 2). There was limited support for our prediction that temporal variation in nest survival was nest stage-specific. A model with a stage × year interaction term was the second-best supported in the candidate set, but overall models with this term only had 41% of the cumulative AIC weight. Model-averaged estimates of nest survival for each stage were contrary to our predications; there was an insubstantial increase in nest survival across time during laying and incubation stages but not during the nestling stage (Fig. 5a). There was also limited support for our prediction that temporal trends in nest survival would be landscape-specific. The forest cover × year interaction term did not appear in any of the top three models (Table 2), and although there was a small increase in overall nest survival across time, the difference between landscapes was relatively constant (Fig 5b). Instead, the best predictor of nest survival was parasitism status, as evidenced by the substantial improvement in model likelihood between the model with species, stage, and habitat type variables (AIC = 10,292.25) and the same model that also included parasitism status (AIC = 10,231.41; Table 2). This effect was not due solely to nest abandonment or the total loss of host young to cowbirds, as a post hoc analysis that included only successful nests and those that failed because of nest predation indicated that the period survival rate (i.e., cumulative survival probability across all three nest stages) for parasitized nests (0.14, 95% CI: 0.11–0.18) was substantially lower than for unparasitized nests (0.26, 95% CI: 0.24–0.29).
Model-based predictions of daily songbird nest survival rate by nest stage (A) and for 5%, median, and 95% observed values of forest cover (23%, 53%, and 94%, respectively; B) in Missouri, USA, 1991–2010. Predictions in (A) are for a balanced population of three songbird species at the median (53%) value of forest cover and at observed frequencies of nests in three habitat types (forest, clearcut, non-forest). Predictions in (B) are for a balanced population of three songbird species and three habitat types, with unbalanced nest stages (lay: 11%, incubation: 49%, nestling: 40%) that reflect the average time spent in each stage across species. Parasitism rates were varied across years and forest cover levels on the basis of the parasitism rate model-selection analysis. Error bars represent 95% confidence intervals and are offset for visual clarity.
The influence of landscape forest cover on fledging brood size and nest survival led to a substantial increase in overall productivity as forest cover increased (Fig. 6). Similar to the fledging brood size results upon which estimates were partly based, the effect size of the increase in productivity from 1991 to 2010 was larger as forest cover declined, with a 30% increase in productivity for nests at 23% forest cover (0.38, 95% CI: 0.21–0.48 in 1991 versus 0.55, 95% CI: 0.35–0.76 in 2010) compared to a 10% increase for nests at 94% forest cover (0.84, 95% CI: 0.67–1.02 versus 0.93, 95% CI: 0.68–1.20; Fig. 6).
Average productivity per nest attempt across time at 5%, median, and 95% observed values of forest cover (23%, 53%, and 94%, respectively) for a balanced population of three songbird species in a study of brood parasitism and nest predation in Missouri, USA, 1991–2010. Error bars represent 95% confidence intervals and are offset for visual clarity.
Several factors may have contributed to cowbird declines in Missouri during 1966–2010. Cowbird abundances increase with proximity to grazing livestock , and cattle production in Missouri in 2009 declined >40% from its peak in 1975 . Further, although landscape forest cover remained largely unchanged at the 10-km scale surrounding the nests we monitored, increased forest cover throughout Missouri (FIA data suggest an 11% increase in forested acreage between 1989 and 2010) may also reduce regional cowbird densities by reducing habitat used for foraging and/or increasing the distance between spatially distinct foraging and breeding habitats . Other factors that limit bird populations such as broad climatic patterns  may also affect cowbird abundances. Regardless of the mechanisms driving cowbird declines, our data from 20 years of monitoring nests at five Missouri sites suggest productivity of three songbird species increased concurrent with these declines. Our results also provide further evidence of the negative effect of forest fragmentation on songbird productivity, though these effects may be changing over time.
Concordant with documented declines in cowbird abundance, the rate and intensity of parasitism declined substantially during 1991–2010 for the three species we studied. Parasitized nests of most passerine bird species exhibit reduced host productivity , an effect that is more pronounced in nests with >1 cowbird nestling . Declines in the rate and intensity of parasitism between 1991 and 2010 at our study sites resulted in increased predicted fledging brood sizes as forest cover declined, where cowbirds are most abundant and parasitism rates are highest. Population dynamics of songbirds are most sensitive to changes in adult and post-fledging survival, but fledging brood size can be an important determinant in population stability .
Because parasitism rates increase and nest survival rates decrease with increasing forest fragmentation in the Midwest , and because cowbirds are established nest predators that depredate nests more frequently in fragmented landscapes , we predicted that nest survival rates would increase as cowbird abundances decreased across time, more so in fragmented landscapes where cowbirds are abundant. Further, we predicted that changes in nest survival would be stage-specific because cowbirds should have little incentive to depredate nests when hosts are laying eggs. However, nest survival only increased modestly over time, and there was considerable uncertainty surrounding our predictions (Fig. 5b). Further, nest survival increased in a manner contrary to both of our predictions. This is perhaps unsurprising given the diverse suite of predators known to contribute to overall rates of predation in our study system ,  that also have exhibited long-term population changes. For example, Blue Jays (Cyanocitta cristata) are frequent predators during incubation  and have declined 1.1% annually in Missouri during 1991–2010 (95% CI: −2.0, −0.3%; ), which may contribute to the temporal patterns of predation we observed during the laying and incubation stages. By contrast, populations of Broad-winged Hawks (Buteo platypterus) and Barred Owls (Strix asio), frequent predators that depredate nests almost exclusively during the nestling stage , may have increased substantially in Missouri during 1991–2010 (Broad-winged Hawk: 4.3% [95% CI: −0.5, 9.2%], Barred Owl: 5.8% [95% CI: 3.3, 8.9%]; ). Correlations between broad-scale predator population trends and local demographic metrics should be interpreted with caution, but they are concordant with previous studies relating predator abundance and avian reproductive performance .
Lower nest survival and reduced fledging brood sizes associated with low landscape forest cover led to a substantial negative correlation between forest cover and our combined productivity metric (i.e., host young produced per nest attempt). With predicted productivity values ≥50% lower at 23% forest cover compared to 94% forest cover regardless of year (Fig. 6), our data serve as a stark reminder of the negative effects of forest fragmentation on breeding birds in the midwestern United States as described by Robinson et al. . We suggest that cowbirds may be an under-acknowledged mechanism behind reduced nest survival in fragmented landscapes. The presence of a cowbird nestling can result in zero host young fledging from on otherwise successful nest because of egg removal or destruction (e.g., ), host ejection , or host mortality due to competition . In addition, louder and more frequent begging by cowbird young  and increased parental activity at nests with cowbird young  may result in higher predation risk  and contribute to lower rates of nest survival for parasitized nests as seen here and elsewhere , . The negative effects of parasitism on fledging brood size and nest survival (and thus on our productivity measure) combined with the fact that cowbirds depredate nests more frequently in less forested landscapes  suggest that cowbirds may be a primary cause of the forest fragmentation effect on songbird productivity in the Midwest.
Despite the substantial decline in predicted parasitism rates during 1991–2010, the concomitant increase in productivity was comparatively modest because the strongly negative impact of parasitism on the productivity of a single nest is muted across a population of nests wherein most are not parasitized and many parasitized nests are depredated. Nevertheless, lower landscape forest cover was associated with a greater increase in predicted productivity, which provides support for our hypothesis that temporal trends in productivity should be landscape-specific. It also suggests that some habitat patches that were formerly population sinks (sensu ) may now produce enough young to be considered sources, which exemplifies the potential value of incorporating temporally dynamic source-sink models into the management of migratory songbirds.
We stress that the patterns we observed in Missouri almost assuredly do not apply throughout the extensive range of the Brown-headed Cowbird. The BBS data suggest that temporal trends in cowbird abundances are not uniform; 19 U.S. states have seen substantial cowbird declines between 1966–2010, but 12 states have had significant increases in cowbird abundances during the same timespan . Furthermore, current abundance trends may not persist into the future, as it is possible that current cowbird declines are part of a long-term population oscillation. Only seven states had significant declines during 2000–2010, and even though cowbird abundance has declined in Canada throughout the entire BBS survey period (1.5% annual decline [95% CI: −2.2, −1.0%]), they actually increased in abundance during 2000–2010 (2.4% annual increase [95% CI: 0.9, 4.1%]; ). Finally, Missouri is near the historical center of the cowbird breeding range ,  where abundances are typically highest , and it may be difficult to detect biologically meaningful changes in the effect of cowbirds on productivity in locations where cowbird abundances are substantially lower.
Predictive models designed to assess the effects of climate change on wildlife distributions and abundances into the next century are increasingly common in the literature. Such models will be most useful when they incorporate important biotic interactions ,  such as those between brood parasites and their host species. Given the inherent complexity of virtually all ecosystems, any such model is likely to be surrounded by substantial uncertainty. Nevertheless, our study underscores the fact that critical demographic parameters are not static in time but can instead exhibit long-term temporal trends. Demographers rely upon empirical data collection to parameterize their models, but our data demonstrate that values derived from older studies may not be reflective of current or future conditions. Studies such as this that monitor populations across extended time periods are essential if we are to accurately predict future trends in the abundance, occurrence, or productivity of wildlife.
We thank all of the primary investigators and graduate students who contributed data for this manuscript; they include D. Burhans, D. Dearborn, A. George, T. Fessler, R. Hirsch-Jacobson, D. Morris, and P. Porneluzi. We also thank the countless field technicians for their help over the past 20 years. We thank members of the Avian Ecology lab at the University of Missouri for providing comments on previous versions of this manuscript.
Conceived and designed the experiments: WAC FRT. Performed the experiments: WAC FRT. Analyzed the data: WAC FRT. Wrote the paper: WAC FRT BR JF.
- 1. Robbins CS, Sauer JR, Greenberg RS, Droege S (1989) Population declines of North American birds that migrate to the Neotropics. Proc Natl Acad Sci USA 86: 7658–7662.
- 2. Sauer JR, Link WA (2011) Analysis of the North American Breeding Bird Survey using hierarchical models. Auk 128: 87–98.
- 3. Sherry TW, Holmes RT (1995) Summer versus winter limitaton of populations: what are the issues and what is the evidence? In: Martin TE, Finch DM, editors. Ecology and management of Neotropical migratory birds. New York: Oxford University Press. 85–120.
- 4. Faaborg J, Holmes RT, Anders AD, Bildstein KL, Dugger KM, et al. (2010) Conserving migratory land birds in the New World: do we know enough? Ecol Appl 20: 398–418.
- 5. Van Horn MA, Gentry RM, Faaborg J (1995) Patterns of pairing success of the Ovenbird (Seiurus aurocapillus) within Missouri forest fragments. Auk 112: 98–106.
- 6. Robinson SK, Thompson III FR, Donovan TM, Whitehead DR, Faaborg J (1995) Regional forest fragmentation and the nesting success of migratory birds. Science 267: 1987–1990.
- 7. Chace JF, Farmer C, Winfree R, Curson DR, Jensen WE, et al. (2005) Cowbird (Molothrus spp.) ecology: a review of factors influencing distribution and abundance of cowbirds across spatial scales. Ornithol Monogr 57: 45–70.
- 8. Lorenzana JC, Sealy SG (1999) A meta-analysis of the impact of parasitism by the Brown-headed Cowbird on its hosts. In: Morrison ML, Hall LS, Robinson SK, Rothstein SI, Caldwell Hahn D et al.., editors. Research and management of the Brown-headed Cowbird in western landscapes. Studies in avian biology vol 18. Camarillo: Cooper Ornithological Society. 241–253.
- 9. Dearborn DC (1999) Brown-headed Cowbird nestling vocalizations and risk of nest predation. Auk 116: 448–457.
- 10. Benson TJ, Brown JD, Bednarz JC (2010) Identifying predators clarifies predictors of nest success in a temperate passerine. J Anim Ecol 79: 225–234.
- 11. Rodewald AD, Kearns LJ (2011) Shifts in dominant nest predators along a rural-to-urban landscape gradient. Condor 113: 899–906.
- 12. Reidy JL, Thompson III FR (2012) Predator identity can explain nest predation patterns. In: Ribic CA, Thompson III FR, Pietz PJ, editors. Video surveillance of nesting birds. Studies in avian biology vol 43. Camarillo: Cooper Ornithological Society. 135–148.
- 13. Cox WA, Thompson III FR, Faaborg J (2012) Landscape forest cover and edge effects on songbird nest predation vary by nest predator. Landsc Ecol 27: 659–669.
- 14. Peterjohn BG, Sauer JR, Schwarz S (2000) Temporal and geographic patterns in population trends of Brown-headed Cowbirds. In: Smith JNM, Cook TL, Rothstein SI, Robinson SK, Sealy SG, editors. Ecology and management of cowbirds and their hosts. Austin: Universtiy of Texas Press. 21–34.
- 15. Sauer JR, Hines JE, Fallon JE, Pardieck KL, Ziolkowski Jr DJ, et al.. (2011) The North American breeding bird survey, results and analysis 1966–2010. 12.07.2011 ed. Laurel: USGS Patuxent Wildlife Research Center.
- 16. Hoover JP, Brittingham MC (1993) Regional variation in cowbird parasitism of wood thrushes. Wilson Bull 105: 228–238.
- 17. Thompson III FR, Robinson SK, Donovan TM, Faaborg JR, Whitehead DR, et al.. (2000) Biogeographic, landscape, and local factors affecting cowbird abundance and host parasitism levels. In: Smith JNM, Cook TL, Rothstein SI, Robinson SK, Sealy SG, editors. Ecology and management of cowbirds and their hosts. Austin: University of Texas Press. 271–279.
- 18. McLaren CM, Underwood TJ, Sealy SG (2006) Conflicting temporal changes in the frequency and intensity of cowbird parasitism on four common hosts in Ontario over 130 years. Condor 108: 238–242.
- 19. Rivers JW, Jensen WE, Kosciuch KL, Rothstein SI (2010) Community-level patterns of host use by the brown-headed cowbird (Molothrus ater), a generalist brood parasite. Auk 127: 263–273.
- 20. Zimmerman JL (1983) Cowbird parasitism of Dickcissels in different habitats and at different nest densities. Wilson Bull 95: 7–22.
- 21. Halkin SL, Linville SU (1999) Northern Cardinal (Cardinalis cardinalis). In: Poole A, editor. The Birds of North America Online. Ithaca: Cornell Lab of Ornithology.
- 22. Payne RB (2006) Indigo Bunting (Passerina cyanea). In: Poole A, editor. The Birds of North America Online. Ithaca: Cornell Lab of Ornithology.
- 23. Martin TE, Geupel GR (1993) Nest-monitoring plots: methods for locating nests and monitoring success. J Field Ornithol 64: 507–519.
- 24. Manolis JC, Andersen DE, Cuthbert FJ (2000) Uncertain nest fates in songbird studies and variation in Mayfield estimation. Auk 117: 615–626.
- 25. ESRI Inc (2008) ArcGIS 9.3. Redlands: Environmental Systems Research Institute.
- 26. Lloyd P, Martin TE, Redmond RL, Langner U, Hart MM (2005) Linking demographic effects of habitat fragmentation across landscapes to continental source-sink dynamics. Ecol Appl 15: 1504–1514.
- 27. Donovan TM, Thompson III FR, Faaborg JR (2000) Cowbird distribution at different scales of fragmentation: trade-offs between breeding and feeding opportunities. In: Smith JNM, Cook TL, Rothstein SI, Robinson SK, Sealy SG, editors. Ecology and management of cowbirds and their hosts. Austin: University of Texas Press. 255–264.
- 28. Burnham KP, Anderson DR (2002) Model selection and multimodal inference: a practical information-theoretic approach. New York: Springer-Verlag.
- 29. SAS Intitute Inc (2008) SAS/STAT 9.2 User's Guide. 9.2 ed. Cary, North Carolina: SAS Institute.
- 30. Shifley SR, Kabrick JM, editors (2000) General Technical Report NC-227: Proceedings of the second Missouri Ozark forest ecosystem project symposium: post-treatment results of the landscape experiment. St. Paul: U.S.D.A. Forest Service North Central Research Station.
- 31. Annand EM, Thompson III FR (1997) Forest bird response to regeneration practices in central hardwood forests. J Wildl Manage 61: 159–171.
- 32. McDonald TL, White GC (2010) A comparison of regression models for small counts. J Wildl Manage 74: 514–521.
- 33. Shaffer TL, Thompson III FR (2007) Making meaningful estimates of nest survival with model-based methods. In: Jones SL, Geupel GR, editors. Beyond Mayfield: measurements of nest-survival data. Studies in avian biology vol 34. Camarillo: Cooper Ornithological Society. 84–95.
- 34. Shaffer TL (2004) A unified approach to analyzing nest success. Auk 121: 526–540.
- 35. Cox WA, Thompson III FR, Faaborg J (2012) Species and temporal factors affect predator-specific rates of nest predation for forest songbirds in the Midwest. Auk 129: 147–155.
- 36. Thompson III FR, Burhans DE (2003) Predation of songbird nests differs by predator and between field and forest habitats. J Wildl Manage 67: 408–416.
- 37. Efron B, Tibshirani RJ (1993) An introduction to the bootstrap. New York: Chapman and Hall.
- 38. Rotella JJ, Dinsmore SJ, Shaffer TL (2004) Modeling nest-survival data: a comparison of recently developed methods that can be implemented in MARK and SAS. Anim Biodiv Cons 27: 187–205.
- 39. Goguen CB, Mathews NE (2000) Local gradients of cowbird abundance and parasitism relative to livestock grazing in a western landscape. Conserv Biol 14: 1862–1869.
- 40. U.S. Department of Agriculture (2012) National Agricultural Statistics Service website. Available: http://www.nass.usda.gov/Data_and_Statistics/Quick_Stats_1.0/index.asp. Accessed 5 Mar 2012.
- 41. Thompson III FR (1994) Temporal and spatial patterns of breeding Brown-headed Cowbirds in the midwestern United States. Auk 111: 979–990.
- 42. Anders AD, Post E (2006) Distribution-wide effects of climate on population densities of a declining migratory landbird. J Anim Ecol 75: 221–227.
- 43. Trine CL (2000) Effects of multiple parasitism on cowbird and Wood Thrush nesting success. In: Smith JNM, Cook TL, Rothstein SI, Robinson SK, Sealy SG, editors. Ecology and management of cowbirds and their hosts. Austin: University of Texas Press. 135–144.
- 44. Donovan TM, Thompson FR (2001) Modeling the ecological trap hypothesis: a habitat and demographic analysis for migrant songbirds. Ecol Appl 11: 871–882.
- 45. Schmidt KA, Ostfeld RS (2008) Numerical and behavioral effects within a pulse-driven system: consequences for shared prey. Ecology 89: 635–646.
- 46. Smith JNM (1981) Cowbird parasitism, host fitness, and age of the host female in an island Song Sparrow population. Condor 83: 167–181.
- 47. Dearborn DC (1996) Video documentation of a brown-headed cowbird nestling ejecting an indigo bunting nestling from the nest. Condor 98: 645–649.
- 48. Marvil RE, Cruz A (1989) Impact of Brown-headed Cowbird parasitism on the reproductive success of the Solitary Vireo. Auk 106: 476–480.
- 49. Dearborn DC, Anders AD, Thompson FR, Faaborg J (1998) Effects of Cowbird parasitism on parental provisioning and nestling food acquisition and growth. Condor 100: 326–334.
- 50. Martin TE, Scott J, Menge C (2000) Nest predation increases with parental activity: separating nest site and parental activity effects. P Roy Soc B-Biol Sci 267: 2287–2293.
- 51. De Mársico MC, Reboreda JC (2010) Brood parasitism increases mortality of Bay-winged cowbird nests. Condor 112: 407–417.
- 52. Stumpf KJ, Theimer TC, McLeod MA, Koronkiewicz TJ (2012) Distance from riparian edge reduces brood parasitism of southwestern willow flycatchers, whereas parasitism increases nest predation risk. J Wildl Manage 76: 269–277.
- 53. Pulliam HR (1988) Sources, sinks, and population regulation. Am Nat 132: 652–661.
- 54. Friedmann H (1929) The cowbirds: A study in the biology of social parasitism. Springfield: C.C. Thomas.
- 55. Mayfield HF (1965) The Brown-headed Cowbird with old and new hosts. Living Bird 4: 13–28.
- 56. Van Der Putten WH, Macel M, Visser ME (2010) Predicting species distribution and abundance responses to climate change: why it is essential to include biotic interactions across trophic levels. Philos Trans R Soc Lond B Biol Sci 365: 2025–2034.
- 57. Wenger SJ, Isaak DJ, Luce CH, Neville HM, Fausch KD, et al. (2011) Flow regime, temperature, and biotic interactions drive differential declines of trout species under climate change. Proc Natl Acad Sci U S A 108: 14175–14180.
- 58. Brambor T, Clark WR, Golder M (2006) Understanding interaction models: improving emprical analyses. Polit Anal 14: 63–82.