Strolling through a Century: Replicating Historical Bird Surveys to Explore 100 Years of Change in an Urban Bird Community

In 1898, Herbert and Alice Walter started a 5-year survey of birds in Lincoln Park—the largest park in Chicago, Illinois—and summarized their data in an urban birding field guide, Wild Birds in City Parks. Twenty-nine years later, William Dreuth compared the relative frequency of species in the Walters’ study to that in his own 5-year Lincoln Park survey. Between 2012 and 2015, we replicated these surveys to investigate a century of bird diversity and community composition change in urban Chicago. While species richness did not change, community composition did. We found that (1) species with a greater diet breadth and (2) species that increased in statewide occupancy were more likely to increase in frequency over time. We conclude that factors at multiple scales brought temporal changes to Chicago’s bird community. Overall, this survey highlights the slow and subtle ways in which species may respond to a century of urban intensification.


Introduction
Since the start of the twentieth century, the proportion of people living in cities has increased from 10% to more than 50% (Seto et al. 2012). As a result, the amount and intensity of urban land cover have increased worldwide (Liu et al. 2018), and it has become undeniable that urbanization negatively influences biodiversity (McDonald et al. 2020). Cities are often located in biodiversity hot spots, and the associated habitat loss caused by urbanization is a major driver of species extirpations or extinctions (McDonald et al. 2008).
Yet some nonhuman species persist, and at times thrive, in cities. As such, the selective pressures of urban environments provide an excellent opportunity to investigate ecological processes.
One underexplored process in urban environments is how species composition changes over time. Certainly, longterm comparisons of natural habitat converted to urban land cover demonstrate a decrease in species richness and substantial community turnover as the historically present species are lost and replaced by generalist urban-adapted species (Aldrich and Coffin 1980;Tingley and Beissinger 2013). Yet over 100 years a forest turned suburban lot likely tells a different story than an already urbanized city over the same time frame. As a city continues to urbanize over time, selective pressures likely increase, which can potentially result in a loss of species that cannot persist at higher levels of urban intensity (Pyšek et al. 2004;Aronson et al. 2016;Fidino et al. 2020). If this is the case, we would expect species richness to decrease. Additionally, temporal changes in urban species composition could reflect large-scale temporal shifts in the distribution and abundance of species in the regional pool (Murgui 2014;Ward et al. 2018). If this is the case, species turnover within a city should reflect overall regional trends. Yet while cities have existed for millennia, explorations into their ecology have increased only in recent decades (Magle et al. 2012). As a result, long-term urban ecological data sets are exceedingly rare, making it difficult to explore how long periods of urban intensification shape community composition in a city.
To explore how species composition changes over time in a city, we replicated a historical bird survey conducted at the turn of the twentieth century in Lincoln Park, the largest park in Chicago, Illinois. The original surveyors, Herbert and Alice Walter, surveyed Lincoln Park from March to May between 1898 and 1903 and wrote a field guide for city bird-watchers (Walter and Walter 1904). Along with accurate species descriptions, the Walters' book summarized their field notes for future comparison. Unbeknownst to the Walters, their survey was then continued by William Dreuth-a stock clerk, amateur naturalist, and expert birder-between 1927 and 1932 (Clark and Nice 1950). Decades later, one of our colleagues found a copy of the Walters' book in a Chicago antique store, and subsequent investigations into the Walters' survey led us to William Dreuth's original field notes stored in the back halls of the Chicago Academy of Sciences. These discoveries inspired us to continue these Lincoln Park surveys between 2012 and 2015. Our goals were to determine (1) whether this bird community had changed over a century, (2) whether generalist species became more common over time, and (3) whether the temporal trends observed in Lincoln Park mimicked those observed throughout the state of Illinois. The extreme levels of urban intensification that Chicago experienced in the past century make this survey unique. For example, since the Walters' bird counts in 1898, the height of the tallest building in downtown Chicago more than quadrupled, the average human population density doubled, and automobiles replaced horses as a primary mode of transportation (Randall et al. 1999;Sovacool 2009). Although this longterm study is of a single location, which limits our ability to assess general trends, replicating these historical surveys highlights frequency changes of the resident and migrant birds that make up Chicago's avian community and can help clarify the impacts that 100 years of urban intensification may have on species richness and composition.

Study Area
Lincoln Park is mostly linear and is located along the western shore of Lake Michigan, roughly 4.5 km north of downtown Chicago ( fig. 1). Since the original bird survey, the park has increased in size from about 125 ha in 1898 to 481 ha currently ( fig. 1), with most new land added to the north of Lincoln Park's original boundary (Clark and Nice 1950; fig. 1). Primarily built for recreation, Lincoln Park has ponds, sports fields, nature areas, and expanses of turfgrass peppered with mature trees and shrubs. These qualities have changed little since 1898. The Lincoln Park Zoo, which lies in the middle of the original park boundaries, was also present across all three survey periods.
Throughout Chicago, temperatures increase from near freezing at the start of the avian migratory season to roughly 157C by May (NOAA 2019). In March, sunrise begins near 7:30 a.m. and, by May, advances to about 5:00 a.m. Average precipitation between March and May is 24.28 cm (NOAA 2019). Snow is common in March, but sometimes it snows in April (NOAA 2019).

Replicating the Historical Bird Surveys
There was little information on the routes that the Walters and Dreuth walked to count birds. Thus, we assumed that the Walters (1898-1903) did not follow a standard methodology, surveyed the entire park, and counted species by sight and sound because their birding guide included apt descriptions of birds and their songs (Walter and Walter 1904). Dreuth, who surveyed between 1927 and 1933, included the count's date and time as well as the end points of the path traveled in his field notes. Neither historical survey described the path traveled in the park, the distance from the path at which birds were identified, or the speed at which a surveyor walked.
At a minimum, the historical surveys included the number of days per year that the counts were conducted. The Walters averaged 75.66 (min p 66, max p 87) counts per year. Because the Walters included a figure of observed species richness per day between May 7 and 20-a time that they deemed the height of migration-in their birding guide, we assumed that they counted birds each weekday in March and April but every day in May (Walter and Walter 1904 We replicated these surveys using this information but followed a standard methodology to increase repeatability. We delineated a 2.45-km line transect from the northernmost point to the southernmost point of the original park boundaries ( fig. 1). In the past century, many walking paths have not changed. Therefore, we used walking paths whenever possible. To count birds, one trained observer (M. Fidino or K. Limbrick) walked the transect at about 2 km h 21 and started roughly 1 h past local sunrise during clear weather. Species were identified by sight and sound within 50 m of the transect. A count's start point was switched daily and started in the north or the south of the park ( fig. 1). We conducted counts each weekday of March and April and then daily in May.
Some species were not included in the historical surveys. House sparrows (Passer domesticus) were-and still areabundant in Lincoln Park, but the Walters and Dreuth did not count them. Additionally, even though the Walters provided a list of 31 water-and shorebirds encountered in Lincoln Park during their survey, they excluded gannets and grebes, cranes and rails, ducks and geese, loons, terns and gulls, and pelicans from their counts "due to a lack of sufficient data" (Walter and Walter 1904, p. 42). While we recorded all species encountered in our counts, we were able to compare only with what the historical surveyors recorded.

Statistical Analysis
With the vegan package (ver. 2.5-6) in R (ver. 4.0.3; Oksanen et al. 2019; R Core Team 2020), we estimated a and b diversity for each survey period to quantify (1) whether species richness decreased over time and (2) how similar communities were among survey periods. For the a diversity analysis, we created a community matrix for the three survey periods (1898-1903, 1922-1927, and 2012-2015) with species along the columns and survey years along the rows. If a species was detected in a given survey year, the associated cell would equal 1 in that matrix; otherwise, it was 0.
To account for potentially missed species in a survey period and correct for differences in sampling effort, we estimated a diversity from the communities' matrices with rarefaction in vegan, which provides a point estimate of species richness as well as standard errors of the point estimate (O'Hara 2005). To estimate b diversity, we compared proportional similarity in bird communities as 1 minus the Jaccard dis-similarity using the proportion of days a species was observed each year during a given survey period, where a value of 1 would indicate that communities are identical (Legendre and Legendre 2012). To quantify how much variability in community composition the survey period (i.e., 1898-1903, 1922-1927, and 2012-2015) explained, we applied a permutational multivariate ANOVA (PERMANOVA) to the same community matrices that we used to calculate b diversity, treating survey period as a categorical variable (Anderson 2017). Following this, we graphically represented the community each year in multivariate space using nonmetric multidimensional scaling, using the Bray-Curtis index to measure distances among the bird communities for each survey year.
To determine whether, over time, generalist species became more frequent or whether species trajectories reflected statewide trends, we used a binomial generalized linear mixed model (GLMM). Our response variable was the proportion of days a species was observed per year weighted by the number of counts conducted per year. We used this response variable because abundances were not available in the first survey period. We quantified each species' generalist ability by deriving their diet and foraging breadth from a global bird functional trait database (Wilman et al. 2014). For diet breadth we counted the number of different diet categories that each species was known to forage on, which included invertebrates, mammals and birds, reptiles, fish, vertebrates, carrion, fruit, nectar, seeds, or other plant material, such as seedlings, weeds, or lichen (n p 10, mean p 1:6, min p 1, max p 6). For foraging breadth, we used the same functional trait database and counted the number of different foraging strategies that a species was known to use, which could include foraging below water, on water, on the ground, in the understory, in the middle to high levels of trees, in the tree canopy, or in the air (n p 7, mean p 2:4, min p 1, max p 5). We subtracted 1 from the foraging and diet breadth of each species so that the model intercept represented a species with a foraging and diet breadth of 1. To determine whether changes in species' relative frequency followed statewide trends, we used statewide estimates of the absolute change in occupancy of breeding birds throughout Illinois between 1906(Ward et al. 2018. Statewide frequency data were scaled so that a 1-unit difference represented a 10% change in statewide occupancy. Only 55% of the birds in our analysis were represented in the Ward et al. (2018) data set. Birds not represented, such as nonbreeding birds migrating through our study area, were given a value of 0.
Our GLMM also included survey periods (1898-1903, 1922-1927, and 2012-2015) as categorical variables. We treated the first survey period as the reference category. Given the much longer 85-year span between our own sampling and the middle survey relative to the much shorter 19-year time span between the first and second survey periods, we included only interactions between the 2012-2015 period and the three continuous variables (i.e., diet breadth, foraging breadth, and change in statewide occupancy) to quantify whether the response to these variables changed in the last survey period relative to the other two periods. For the random effect structure of the model, we allowed all three survey periods to vary by species. This model was fitted in version 2.19.1 of Stan (Stan Development Team 2018) using the stan_glmer function in rstanarm (Goodrich et al. 2018). Following a 1,000-step burn-in, the posterior was sampled a total of 24,000 times across six chains. The intercept of the model was given a vague Cauchy(0, 5) prior, while slope terms were given a Cauchy(0, 2:5) prior (Gelman et al. 2008). Some data had to be censored from our GLMM analysis. For example, a species was removed if it was observed in only one of the three survey periods. Likewise, waterbirds, shorebirds, and house sparrows were excluded from this analysis because of their omission from the first survey pe-riod. Data for this analysis have been archived in the Dryad Digital Repository (https://doi.org/10.5061/dryad.w3r2280rb; Fidino et al. 2021).

Bird Richness and Similarity
In total, 145 species in 34 families were observed across all survey periods. Birds known to breed in northeastern Illinois represented~63%, 57%, and 60% of the community for each survey period, ordered from the oldest survey to the most recent. Parulid warblers represented the greatest portion of the species pool (n p 33 species), followed by New World sparrows in Passerellidae (n p 19 species) and blackbirds in Icteridae (n p 10 species). After we accounted for potentially missed species, fewer species were observed between 1898 and 1903 (mean p 118:15, SE p 4:87), while a similar number of species was observed between 1927 and 1932 (mean p 135:31, SE p 6:89) and between 2012 and 2015 (mean p 131:22, SE p 5:15). Community composition, as estimated by 1 minus the Jaccard dissimilarity, was most similar between 1898-1903 and 1927-1932 (0.59), which were closer together in time, and most dissimilar between 1898-1903 and 2012-2015 (0.45). Community composition was also dissimilar between 1927-1932 and 2012-2015 (0.48). With the PERMANOVA, the survey period explained 60.06% of the variation in this bird community (P ! :001; fig. 2). The difference in the 10 most common species per survey period also helps illustrate the shift in community composition (table 1). Only the American robin (Turdus migratorius) and common grackle (Quiscalus quiscula), for example, remained in the 10 most common species across survey periods (table 1).

Species Trends over Time
Of the  On their own, a species' diet breadth (b p 0:18, 95% CI p 20.07 to 0.45) or foraging breadth (b p 0:11, 95% CI p 20.24 to 0.46) was not associated with changes in a species' frequency. However, species with a greater diet breadth did, on average, increase in frequency over time, as evidenced by the positive interaction term between diet breadth and the third survey period in our GLMM (b p 0:35, 95% CI p 0.12 to 0.58). With other parameters held constant at zero, species that consumed one diet category were observed on an average of 3.50% of survey days (95% CI p 1.66% to 7.33%) between 2012 and 2015. Conversely, species that consumed three and five different categories were observed on an average of, respectively, 9.50% (95% CI p 5.17% to 16.83%) and 23.34% (95% CI p 9.97% to 44.58%) of survey days between 2012 and 2015. Unlike diet breadth, there was insufficient evidence of an association between foraging breadth and changes in species frequency through time (b p 20:26, 95% CI p 20.57 to 0.04).
Species that increased in statewide occupancy were, on average, more likely to increase in frequency over time in Lincoln Park, as evidenced by the positive interaction term between the change in statewide trends and the third survey period (b p 0:40, 95% CI p 0.21 to 0.57). There were some species, however, that increased statewide but became less common in Lincoln Park and vice versa ( fig. 2). For example, American crows (Corvus brachyrhynchos) decreased in statewide occupancy over time, but their absolute frequency change in Lincoln Park between the historic surveys and our survey was 77.09% (95% CI p 72.11% to 81.32%; fig. 2). Conversely, brown-headed cowbirds (Molothrus ater) increased in statewide occupancy, but their absolute frequency change in Lincoln Park between the historic surveys and our survey was 228.30% (95% CI p 222.08% to 234.10%; fig. 2). Overall, of the 121 species analyzed in our GLMM, 58 of them exhibited changes in their observation frequency over time at the a p 0:95 level ( fig. 3).

Discussion
In one century, Lincoln Park's breeding and migratory bird community has profoundly changed. The rock pigeon (Columba livia) and American crow were rare 100 years ago but are now common (table 1). Blue jays were historically common year-round but are now seen only during the migratory season (Walter and Walter 1904). Our analysis of this century of data provided two key results that indicate what factors may be associated with temporal changes in this urban bird community. First, we found that species with a more generalist diet were more likely to increase in frequency over time. Second, species that increased in statewide occupancy were also more likely to increase in Lincoln Park. When considered together, these results likely demonstrate that factors at both local and regional scales brought changes to Chicago's bird community. As urbanization negatively influences local primary productivity (Imhoff et al. 2004) and generalist species are predicted to be favored in areas of low primary productivity (Evans et al. 2005), urban environments are often composed of generalist species (Callaghan et al. 2019(Callaghan et al. , 2020. Our analysis agrees with this observed pattern, although we did observe that the strength of this relationship increased over time. We found that species with a greater diet breadth (i.e., more generalist) became more common in Lincoln Park over the past century, although we cannot causally attribute this result to increased levels of urban intensity. Such a result may indicate that the local selective pressures of the Chicago landscape have changed to favor species able to take advantage of the dynamic, heteroge-neous resources a heavily urbanized city provides, and therefore maladapted, nongeneralist species are filtered out at the local scale (Aronson et al. 2016).
Because generalists increased in frequency, we had also expected species richness to decrease, which we did not observe. Instead, we observed substantial community turnover and large shifts in species frequencies yet comparable species richness through time. As other long-term comparisons of urban bird communities have also documented dramatic changes in community composition and species frequency over similar time frames, these patterns are perhaps general rules for urban environments (Major and Parsons 2010;Shultz et al. 2012;Strohbach et al. 2014;Fidino and Magle 2017). One apparent disagreement, however, is how a diversity may change through time, as other long-term studies have shown an increase, a decrease, or no net change in species richness (Fidino and Magle 2017 and references therein). What causes this disagreement may be partly due to the adaptive potential of species in the regional pool. If species are lost, others may take their place, and so the extent to which species can adapt to and colonize urban environments could result in variable differences in a diversity over time (Evans et al. 2009(Evans et al. , 2010. In our study, this apparent stability in species richness may be due to the time of year birds were sampled combined with Lincoln Park's size and location along Lake Michigan, which make it an ideal stopover site for migrating birds. In Lincoln Park, migrant birds represented nearly one-third of the observed bird community during this study, and given the park's size and location, it is therefore likely to see a high diversity of birds during the migratory season. Nevertheless, while richness may not have changed over time, we do conclude that, as in other urban environments, the Lincoln Park bird community has become largely dominated by a small number of common urban-adapted species (table 1; McKinney 2006) but otherwise is peppered with an impressive array of species migrating through.
We also observed that, on average, species that increased in statewide occupancy increased in frequency in Lincoln Park. We attribute this to several regional changes throughout northern Illinois over the past century. For example, the landscape in northern Illinois, where Chicago resides, has become more forested and urban between 1898 and 2015, which benefited some birds (Walk et al. 2010). Northern cardinals and red-bellied woodpeckers were rare in northern Illinois but expanded their range northward for increased forest cover (Walk et al. 2010). Both species are now common in Lincoln Park. The black-capped chickadee-a common cavity nester of Illinois's deciduous or mixed forestswas historically rare in Lincoln Park but became the seventh most frequent species (table 1). While we attribute most of the chickadee's success to Illinois's increased forest cover, its increased presence may also be because large older parks like Lincoln Park offer a high density of natural cavities that chickadees nest in (LaMontagne et al. 2015;Bovyn et al. 2019). Other urban-tolerant species, like the chimney swift (Chaetura pelagica), European starling (Sturnus vulgaris), American robin, and common grackle, have become more ubiquitous statewide (Ward et al. 2018). In Lincoln Park, these species either increased in frequency or remained common through time (table 1; fig. 1). As such, many differences in the Lincoln Park migratory community reflect statewide population trends over the past century.
While Illinois is fortunate to have a long-term statewide bird survey (Walk et al. 2010), Chicago was added to this bird census only in the 2000s. In fact, few cities have historical ecological records, possibly because they have long been regarded as inadequate habitats for wildlife (Fidino and Magle 2017). This is unfortunate, not only because large urban green spaces offer significant resources for migratory birds (Brawn and Stotz 2001) but also because many species persist in cities. In the face of dramatic urban intensification, we did not see a decrease in species richness. In fact, species richness may have slightly increased over time. We did see, however, a dramatic change in community composition over a century, whereby the community generally became dominated by species that had increased statewide, had a greater diet breadth, or both. Replicating historical surveys-like these-can provide insights into long-term community changes that would be impossible for a single naturalist to carry out. We hope that our efforts in standardizing and compiling these data across surveys will be useful for others and may encourage a future naturalist to replicate the survey again-but perhaps in less than 75 years.

Acknowledgments
We thank A. Sacerdote-Velat for finding a copy of Wild Birds in City Parks, D. Roberts for help locating and copying William Dreuth's original field notes, and the Chicago Academy of Sciences for storing said field notes for decades. We also thank E. D. Brodie III, M. P. Ward, and two anonymous reviewers, whose comments greatly improved the quality of the manuscript. We thank the Prentice Foundation, the Grainger Foundation, and the Davee Foundation for funding. Finally, we thank whoever is willing to take up the mantle and re-create these surveys again in the future. One piece of advice: wear a hat in April and May. Red-winged blackbirds are quite territorial during this time of year.