Natural history and community science records confirm rapid geographic shifts in the distribution of Bachman’s Sparrow ( Peucaea aestivalis ) since 1850

. North American grassland birds colonized emerging habitat created by expanding agriculture in a pattern of eastward expansions from the mid-1800s to mid-1900s. These birds have been declining, since at least the mid-1900s, largely as result of anthropogenic landscape change. Only one bird that now breeds predominantly in southeastern pine savannas is thought to have experienced a concurrent range expansion into this region: Peucaea aestivalis (Bachman’s Sparrow). However, our understanding of the P. aestivalis expansion, and subsequent retraction to the southeastern United States, is largely based on a contemporaneous review of only a subset of historical records from beyond its modern, northern limit. We suggest an alternative explanation for these historical records is that P. aestivalis historically occurred more broadly than was recognized in contemporaneous literature. To evaluate these hypotheses, we reviewed field observations from literature, natural history collections, and eBird to show how P. aestivalis presence throughout eastern North America has shifted since the mid-1800s. To confirm that these findings were not the result of detection bias, we repeated our analysis on a common sparrow species ( Spizella pusilla ) with a largely overlapping breeding range, but no history of expansion and retraction. We confirm that P. aestivalis expanded its range, but add that prior to that expansion, its historical distribution was broader than commonly acknowledged today. As a result, we identify the northwestern historical limit of P. aestivalis , the Ouachita and Ozark highlands, as a potential source region for an eastward expansion that is consistent with those of other North American grassland birds of the era. We discuss the potential evolutionary and conservation implications of this range expansion on P. aestivalis given our more nuanced understanding of it. Anthropogenic landscape change initially provided additional habitat for P. aestivalis but has ultimately resulted in a reduction of the P. aestivalis distribution


INTRODUCTION
Birds adapted to disturbance-mediated habitats are particularly threatened by human activities (Brawn et al. 2001). One of the clearest examples documents the decline of North American grasslands birds since the mid-1900s (Peterjohn and Sauer 1999). This decline is a product of agricultural intensification over the course of the 1900s coupled with the primary effects of habitat loss due to agricultural expansion and the suppression of historical fire regimes (Askins et al. 2007). In the hundred or so years leading up to the documentation of these declines, a pattern of eastward expansion of grassland birds occurred from the central prairies to the northern portions of the central and eastern plains and plateaus (Askins 2000). Species exhibiting this pattern included Eremophila alpestris (Horned Lark; Thomas 1951, Hurley andFranks 1976), Spiza americana (Dickcissel; Gross 1956, Hurley andFranks 1976), Sturnella neglecta (Western Meadowlark;Lanyon 1956), and Chondestes grammacus (Lark Sparrow;Lanyon 1956, McNair 1982. These expansions were also attributed to anthropogenic landscape change-primarily expanding agriculture and the decimation of eastern forests (Askins 2000). Despite the recent declines, these species still occur in portions of the new areas they colonized. The propensity of these now declining birds to colonize novel habitat created by anthropogenic disturbance in recent history is encouraging amid continued anthropogenic landscape change.
Another potentially informative case of range expansion related to anthropogenic landscape change occurred for Peucaea aestivalis (Bachman's Sparrow) [1] in the late 1800s and early 1900s. This characteristic grassland bird of the southern pine savannas (Askins et al. 2007) is now uncommon across its distribution; however, a century ago, P. aestivalis was recorded beyond the limits of its modern distribution (hereafter extralimital), as far north as Ontario (Saunders 1917, Snyder 1929) and upstate New York (Meade 1941). The long-standing explanation for these records assumes the following: (1) the historical P. aestivalis distribution prior to these extralimital records was consistent with its modern distribution, and (2) this distribution expanded, generally northward, and subsequently retracted back to the Southeast (Harlan 1989, Dunning et al. 2020). However, this explanation is complicated by extensive habitat loss and the unresolved taxonomy of the extralimital birds, and it relies heavily on a contemporaneous review (Brooks 1938) of only a portion of extralimital records for P. aestivalis.
Generally, P. aestivalis breeding habitat depends on frequent firereturn intervals: the primary vegetation is grasses and forbs that are densest below one meter and cover a patchwork of litter and bare ground amid an otherwise open environment with scattered shrubs, trees, or other perching structures (Dunning and Watts 1990, Haggerty 2000, Jones et al. 2013. Peucaea aestivalis primarily occurs in southern pine (Pinus spp.) savannas today, and most recent research on P. aestivalis has been conducted in the remaining longleaf (P. palustris) pinelands found throughout the Southeast (Dunning et al. 2020). Peucaea aestivalis is also regularly found in utility right-of-ways maintained by anthropogenic disturbance (Champlin et al. 2015), clearcuts associated with dense pine plantations (Haggerty 1988(Haggerty , 2000, treeless dry prairie habitat in south-central Florida (Korosy et al. 2013), and a remnant fragment of oak (Quercus spp.) savanna maintained by frequent fire on Fort Campbell Military Base in central Kentucky and Tennessee (Barrioz et al. 2013, Hockman 2013). However, the species has not been detected recently in oakwood and pineland savannas undergoing restoration in Tennessee and Kentucky that lie outside of Fort Campbell (Barrioz et al. 2013). Similarly, it was once thought P. aestivalis may have been abundant in the limestone glades of the Ozarks (Hardin et al. 1982), but it has only been detected in small numbers (Roach et al. 2019) or not at all (Comer et al. 2011) in recent studies of remnant oak and pineland savannas (including some undergoing ecological restoration in the Ozark and Ouachita highlands).
Peucaea aestivalis has experienced severe habitat loss and fragmentation as the distributions of longleaf pine and midwestern transitional oak savannas have been reduced to 3% (Frost 1993) and 1% (Nuzzo 1986) of their historical distributions, respectively (Fig. 1). Peucaea aestivalis habitat loss at the northern extent of its recognized distribution can also be attributed to the similarly drastic reduction of mixed-shortleaf (P. echinata) pine forests (Moser et al. 2007). Longleaf pine forests were successively deforested from Virginia south through the Atlantic Coastal Plains and Piedmont from 1840 to the late 1860s and then west across the Gulf Coastal Plains as logging, settlement, and agricultural expansion depleted the forests, leaving few oldgrowth longleaf pinelands by the early 1900s (Frost 1993). Similarly, rapid and intensive shortleaf pine logging in the Ouachita and Ozark highlands between 1880 and 1920 facilitated hardwood encroachment into the oak-pine savannas of the region (Cunningham 2007). More generally, the reduction of transitional oak-pine savanna throughout the Midwest tended to occur with settlement and the accompanying fire-suppression and landscape conversion to agriculture (Nuzzo 1986).
The persistence of P. aestivalis and other birds that inhabit these characteristically open landscapes is dependent on restoring the historically frequent fire-return intervals that maintain suitable ground cover conditions (Engstrom 1993, Wilson et al. 1995, Barrioz et al. 2013. The application of prescribed fire has been most successful throughout the historical range of longleaf pine, particularly in Gulf Coast states (Melvin 2018). In contrast, there has been a reduction in prescribed fire application over the last decade in the region encompassing the northern limits of shortleaf pine and the southern extent of the historical distribution of oakpine savanna in Missouri, Kentucky, and Tennessee (Melvin 2018). Still, P. aestivalis density remains high (~0.5 territorial males hectare -1 ; Tucker et al. 2004), or increases (Wilson et al. 1995), in areas with frequent fire-return intervals (≤ 3 years). That longleaf pine savannas have benefitted most from this prescribed fire activity is perhaps why research on P. aestivalis has become so focused on its association with this particular habitat.
In his review of P. aestivalis occurrence in southeastern Ohio, West Virginia, and southwestern Pennsylvania, Brooks (1938) states that P. aestivalis began its expansion into his focal region ( Fig. 1) in the late 1890s, inhabiting early successional habitat like fallow agricultural fields and pastureland, hillside meadows, and grassy open oak groves. Brooks (1938) found the sparrow to be a locally common breeder in the region until reaching its peak abundance between 1915 and the early 1920s. Its abundance then declined until it was considered rare by the 1930s. Brooks (1938) attributed the appearance and subsequent disappearance of P.  (BirdLife International and NatureServe 2015), the natural range of longleaf pine (Pinus palustris; Prasad and Iverson 2003) in dark green, the natural range of shortleaf pine (Pinus echinata; Prasad and Iverson 2003) in light green, and the historical distribution of transitional midwestern oak savannas, recreated based on Nuzzo (1986), in ochre. The black rectangle encompasses the focal region reviewed in Brooks (1938).
aestivalis in this region to the abandonment of agricultural lands in the region and subsequent ecological succession of this open land to shrubby habitats unfavorable to P. aestivalis. This explanation has been restated, first by Weston (1968), to explain historical P. aestivalis records from the focal region of Brooks (1938) in addition to other extralimital regions that were not thoroughly reviewed at the time. We refer to this inductive generalization of the changes Brooks described for his focal region as the expansion-retraction hypothesis.
Alternatively, P. aestivalis may simply have occurred more broadly in the past. During the expansion period described by Brooks (1938), P. aestivalis was also recorded in several other regions where it no longer occurs, including southeastern Missouri (Woodruff 1907), southern Illinois, southwestern Indiana (Cooke 1914), Virginia (Daniel 1899a, Cooke 1914, Murray 1933, and portions of Texas (Lloyd 1887, Ragsdale 1892, Tennessee (Gainer 1921), and Kentucky (Blincoe 1921). Prior to the expansion period, another species, P. illinoensis (Oakwoods Sparrow; Ridgway 1879), was described from several specimens collected in southern Illinois, but was quickly subsumed into P. aestivalis (Brewster 1885). Thus, it is possible that prior to its expansion into Brooks's focal region in the 1890s, P. aestivalis occurred more widely than its modern recognized distribution and experienced a prolonged contraction to its modern distribution. We refer to this possibility as the broad-historicaldistribution hypothesis. The reason this hypothesis may not have been considered previously may stem from the close association P. aestivalis now has with southeastern longleaf pine savannas.
Whether the extralimital records are the result of an expansion or a broad-historical-distribution is unclear, in part because of the unresolved taxonomy of the extralimital birds. The first iteration of P. aestivalis taxonomic revision began two decades prior to the proposed start of the expansion. The first edition of the American Ornithologists' Union (AOU) Check-list of North American Birds listed two P. aestivalis subspecies: P. a. aestivalis (Pinewoods Sparrow; Lichtenstein 1823) from Florida and southern Georgia and P. a. bachmani (Bachman's Sparrow; Audubon 1834) ranging from South Carolina and Alabama through eastern Texas and north to southern Illinois and Indiana (AOU 1886). However, P. a. illinoensis (Oakwoods Sparrow; Ridgway 1879), which ranged from central Texas to southern Illinois in oak savannas and fallow agricultural fields, had been described and identified as a synonym of P. a. bachmani (Brewster 1885), but was not listed as such in the first four editions of the AOU check-list (AOU 1886(AOU , 1895(AOU , 1910(AOU , 1931. Peucaea aestivalis illinoensis was first included as a third subspecies in the fifth edition of the AOU checklist, the last to treat subspecies (AOU 1957). The inclusion seems to have been the result of the reevaluation of a series of Tennessee specimens (1882-1883) identified as hybrids of P. a. bachmani and P. a. illinoensis (Wetmore 1939). Notably, the description of P. a. illinoensis places P. aestivalis extralimital to its modern distribution prior to the proposed start of the expansion. Finally, P. aestivalis taxonomy has not considered potential ecological differences among the subspecies suggested by the habitat descriptions included in their original classifications: The northeastern limit of P. a. illinoensis was described as southern Illinois (Ridgway 1879) through central Indiana (AOU 1957), though P. aestivalis was thought to have colonized the extralimital region through southern Ohio (Brooks 1938, 1952, Weston 1968, occupying similar habitat to that described for P. a. illinoensis and distinct from its southern pinewoods habitat. The case of the P. aestivalis expansion has not been holistically reviewed since Brooks (1938) despite the possibility for new insight on the effects anthropogenic landscape change has had on grassland birds in the modern era and the wealth of easily accessible data stored in natural history collections (Meineke et al. 2019, Nelson andEllis 2019) and eBird (Sullivan et al. 2009). Here we examine recent historical shifts in the geographic distribution of P. aestivalis through extensive review of occurrence records from historical field observations, natural history collection, and eBird. By identifying the geographic coordinates of all available historical records and summarizing them in maps, we aim to evaluate whether historical records support the expansion-retraction hypothesis or the broadhistorical-distribution hypothesis.

Record collection and parsing
We compiled natural history records by querying two web-based biodiversity databases, Vertnet (http://vertnet.org/) and Global Biodiversity Information Facility (GBIF; https://www.gbif.org/), for Peucaea aestivalis and Aimophila aestivalis to account for a recent genus re-classification (DaCosta et al. 2009). We downloaded all Vertnet records and all GBIF occurrence records categorized as preserved specimens (GBIF 2021a, b). This resulted in four sets of natural history collection records that we parsed to remove duplicate records using a combination of R 4.1.0 (R Core Team 2021) code and manual editing in spreadsheets. We identified duplicate records as records with identical entries in the fields: institutioncode, catalognumber, and basisofrecord. We filtered the remaining unique records in R 4.1.0 to those including collection year, month, and locality specified to at least county. We also removed fossil records and those associated with specimens collected since 2002 when eBird was established (Sullivan et al. 2009) and dramatically increased available observations. Finally, to focus on records from the breeding season, we removed all records from October through February. All natural history collections from which we compiled records from Vertnet for the final dataset are listed in Appendix 1.
We added additional records that passed the above filters from three natural history collections and literature. We searched for natural history collections in the midwestern United States with holdings that are not published on Vertnet and GBIF in an attempt to identify as many P. aestivalis specimens collected in the extralimital region as possible. We obtained P. aestivalis records from two additional collections: Zoology Collections at Cincinnati Museum Center, Cincinnati, Ohio and the Aves Collection of Joseph Moore Museum at Earlham College, Richmond, Indiana. We contacted the Natural History Museum, Tring, United Kingdom to identify any extralimital records in their expansive collection. Finally, we queried Searchable Ornithological Research Archive (SORA; https://sora.unm.edu/) for Peucaea aestivalis, Aimophila aestivalis, Bachman's Sparrow, Oakwoods Sparrow, and Pinewoods Sparrow to identify additional extralimital records.
We obtained P. aestivalis occurrence records from eBird (eBird 2018) and manually examined historical records prior to the establishment of eBird to remove any that matched records in our natural history collection dataset. We removed observations that were duplicates or occurred between October and February using auk 0.3.0 package (Strimas-Mackey et al. 2018) in R 4.1.0 (R Core Team 2021). Detailed documentation of all filtering and the associated R code is available as are the data used for this study.

Assigning geographic coordinates and visualizing records on maps
We examined the records in our filtered natural history dataset to identify and remove obvious misidentifications, e.g., a specimen from Oaxaca, Mexico. To assign geographic coordinates to each natural history record, we parsed the associated locality into independent clauses. Then, we identified the most specific locality clause together with the corresponding spatial feature and locality type following georeferencing best practices Wieczorek 2020, Zermoglio et al. 2020). We identified the geographic coordinates and coordinate uncertainty radius of the most specific locality clause for each record with the point-radius method (Wieczorek et al. 2004) using GEOLocate (Rios and Bart 2010). During georeferencing we removed 20 additional records with dubious or vague localities. For the eBird data, we used the geographic coordinates provided with each entry.
The natural history and eBird records were grouped into four time periods that correspond to the expansion and retraction described in Brooks (1938) and Weston (1968). To visualize changes in P. aestivalis occurrence in the recent past, we created four heatmaps Although it has become commonplace to generate species distribution models (SDMs) from presence-only, natural history collection data (Elith and Leathwick 2009), SDMs are not wellsuited to this study because they identify environmental variables associated with species occurrence data and project the likelihood of environmental suitability across space and/or time (Elith and Leathwick 2009). The relevant predictor variables needed to build meaningful SDMs are not available for the earlier time periods studied here. Instead, the heatmaps produced here depict the evidence for sparrow presence in a given region and time period. The maps were created using the point density tool in ArcGIS Pro 2.8 (ESRI 2021) with default settings and the mean density of records/hectare per period. Results are presented using a color scheme that is based on the standard deviation of each raster cell from the mean density of records for each period.

Accounting for bias
Our ability to interpret mapped P. aestivalis records is challenging because opportunistic occurrence data are not randomly sampled (Boakes et al. 2010, Daru et al. 2018. In an attempt to distinguish between a lack of P. aestivalis records versus putative absence in a given region and period, we also applied our methods to a similar but more common sparrow. We chose Spizella pusilla (Field Sparrow) because it is common across its distribution, is another visually nondescript sparrow, and occupies structurally similar early successional habitats as P. aestivalis. Moreover, P. aestivalis and S. pusilla occurred together during the expansion period within the limit of the modern distribution of P. aestivalis in northern Mississippi (Allison 1899) and in the extralimital range (Eifrig 1915), and more recently in Ozark limestone glades (Hardin et al. 1982). Finally, S. pusilla was first described in 1810 (Wilson), just a decade before P. aestivalis was described in 1823 (Lichtenstein) and should have been similarly familiar to naturalists and ornithologists of the time.
We compiled similar occurrence records for S. pusilla and created heatmaps for the focal time periods described for P. aestivalis with the following exceptions: (1) we queried only Vertnet for S. pusilla natural history records, because of the overall consistency we found between P. aestivalis records available on Vertnet and GBIF; (2) to process the larger number of S. pusilla records, we used geographic coordinates provided by VertNet and performed batch processing of remaining S. pusilla localities using GEOLocate. We output only the geographic coordinates, because we did not make use of the P. aestivalis uncertainty radius. When GEOLocate batch processing did not identify coordinates for a record, we assigned geographic coordinates and an uncertainty radius using the methods applied to P. aestivalis.

RESULTS
The historical P. aestivalis records we mapped (n = 3042) included 1600 from 53 natural history collections, 56 from historical literature, and 1386 from historical data included in eBird. The historical S. pusilla records we mapped (n = 54,991) came primarily from eBird (n = 49,774) but also included 4827 from Vertnet. Modern maps were based on 11,232 P. aestivalis and 574,596 S. pusilla observations. At least an order of magnitude more data was available for S. pusilla for all periods except the pre-expansion and expansion period. The expansion period had the sparsest data for both species (Table 1).
Maps of pre-expansion P. aestivalis records are consistent with both the expansion-retraction and broader-historical-distribution hypotheses. P. aestivalis records occurred in a portion of the extralimital region from southeastern Missouri through southwestern Indiana ( Fig. 2A) prior to the expansion period and provide support for the broader-historical-distribution hypothesis. . The color of each raster cell reflects its deviation from the mean density of records/ha for each period: red cells have the highest density and blue cells have the lowest density of records within a given time period and species. The dashed line in (E) represents our approximation of the region encompassed by P. aestivalis during its expansion for comparison to the modern time period. For each map, the mean density of records/ha, its standard deviation, the maximum density of records/ha, and the raster cell size are reported in Table 1.
However, S. pusilla occurrence records are more broadly distributed throughout the Midwest (Fig. 2F), indicating that ornithologists were collecting specimens and natural history observers were documenting other bird species in the region prior to 1890. The presence of S. pusilla in the extralimital region during the pre-expansion period suggests an absence of P. aestivalis here and supports the expansion-retraction hypothesis. Peucaea aestivalis records also occurred farther west in Texas than the limit of the modern distribution. Disjunct records also confirm the presence of P. aestivalis throughout the Atlantic and Gulf coastal plains. Spizella pusilla records are similarly sparse compared to the expansion period and suggest this disjunct pattern does not reflect the absence of P. aestivalis throughout the Gulf Coastal Plain (Fig. 2G).
Expansion-period records provide support for the expansionretraction hypothesis (Fig. 2B). During expansion, P. aestivalis records are more widely distributed throughout the Midwest than prior to 1890 (Figs. 2A, 2B). The records indicate P. aestivalis began to occur in more northern portions of Illinois and Indiana, east to southwestern Pennsylvania, the western extremes of Maryland, and Virginia. The greatest density of extralimital records during expansion occurred in southern Ohio and northern West Virginia. The absence of P. aestivalis records in central and northwestern Illinois and northeastern Indiana coupled with concurrent S. pusilla records in the same regions (Fig. 2G) suggest that P. aestivalis presence in the northwestern portion of the extralimital region was disjunct. When compared to the pre-expansion period ( Fig. 2A), an increase in the density of P. aestivalis records occurs outside the limits of its modern distribution in the inland Atlantic coastal plains in northern Alabama and western North Carolina. There is also a dearth of P. aestivalis records in central and northeastern Texas where concurrent S. pusilla records occur (Fig. 2G). Maps of retractionperiod records suggest P. aestivalis continued to occupy portions of the extralimital region into the 1960s, specifically in the central portion of the extralimital region (southeastern Indiana through northern West Virginia) as indicated by the greatest density of extralimital records (Fig. 2C).
In comparison with the earlier time periods, post-retraction P. aestivalis records occurred most consistently within the modern distribution ( Figs. 1 and 2D). Still, several records indicated that P. aestivalis occurred locally in a disjunct portion of its extralimital distribution in southern Ohio and West Virginia until 1974 and 1989, respectively, and slightly north of its accepted distribution in Indiana until 1979. The absence of P. aestivalis in areas where it previously occurred in the extralimital region is supported by concurrent S. pusilla records in the Midwest and Northeast (Fig. 2I).
The modern map shows restriction of P. aestivalis to the Southeast (Fig. 2E). In comparison to the post-retraction period, the P. aestivalis distribution has become further reduced during the 21st century. Its northern limits are fragments in northeastern North Carolina and southern Virginia, the central Kentucky-Tennessee border, and extreme south-central Missouri. Moreover, the modern map demonstrates a more fragmented contemporary distribution of P. aestivalis compared to the oft-referenced range map (Fig. 1). Its fragmented geographic distribution is especially apparent in the western Gulf coastal plains. Throughout all time periods, the greatest densities of records consistently occurred in the coastal regions of central South Carolina, south through Georgia and into central Florida (Fig. 2).

DISCUSSION
Our review of records of P. aestivalis since the mid-1800s supports aspects of both the expansion-retraction and broad-historicaldistribution hypotheses. Literature review and maps largely support the timing and geographical extent of the P. aestivalis expansion into the region described by Brooks (1938), though additional data suggest the bird's distribution was broader during the pre-expansion period than commonly acknowledged today. This holistic review of historical records points to the western extent of the historical distribution as a potential source region for the colonization of the extralimital region, in addition to the previously stated southern origin.
Our results coupled with contemporaneous literature largely support the description of the expansion and retraction by Brooks (1938). In southeastern Ohio, West Virginia, and southwestern Pennsylvania, P. aestivalis was first observed after 1890 ( Fig. 2A,  2B; Dawson 1903, Gano 1904, Brooks 1912, Roads 1936). Subsequently, P. aestivalis remained a locally common breeder in the region until the mid-1920s (Gano 1904, Brooks 1914, Dickey 1917, Brooks 1938. In contrast to the precipitous decline in P. aestivalis numbers by the mid-1930s reported by Brooks (1938), our results show a similar distribution in this region during retraction as during expansion (Fig. 2B, 2C). Additional contemporaneous literature confirms the timing of the decline in West Virginia and southwestern Pennsylvania (Bibbee 1934, Brooks 1934, 1938, but not in Ohio (Hicks 1935). Moreover, within the focal region identified by Brooks (1938), most records are from Ohio across all time periods and the highest density of records from the retraction period occurred in southern Ohio. This suggests that southern Ohio may have been the stronghold of the extralimital population as it was extirpated from its northeastern limit.
Despite our overall agreement with the expansion-retraction hypothesis offered by Brooks (1938), our results confirm that the P. aestivalis range expansion and subsequent retraction was more complex than Brooks (1938) described. Our review highlights evidence for the broad-historical-distribution hypothesis, but not that P. aestivalis occurred throughout the entire extralimital region prior to its expansion. Prior to the late 1800s, P. aestivalis occurred farther west in Texas than currently accepted and as far north as southern Illinois and Indiana (Figs. 1 and 2A). A majority of these records likely influenced Ridgway's description of P. a. illinoensis (Ridgway 1879) and an additional record places P. aestivalis as far north as Bloomington, Indiana, more than a decade earlier than the records considered by Brooks (1938, Cooke 1914. However, even during the expansion period it is unlikely that populations ever established beyond northeastern Illinois and southwestern Pennsylvania. There are records during the retraction period from Michigan (Campbell 1944), Ontario (Saunders 1917, Snyder 1929, New York (Meade 1941), and New Jersey, but none are breeding records (Fig. 2C). There are breeding records from Maryland, but records from the other states represent casual occurrences (AOU 1957). Notably, no P. aestivalis have been detected in the extralimital region in at least the last two decades (Fig. 2E, 2J), despite the increased search effort confirmed by eBird activity.
Other birds Lanius ludovicianus (Loggerhead Shrike), Chondestes grammacus (Lark Sparrow), and Thryomanes bewickii (Bewick's Wren) that seem to have experienced expansions most consistent in geography and timing to that of P. aestivalis are not southern pinewood specialists. The lack of corresponding distributional shifts for pine specialists such as Dryobates borealis (Redcockaded Woodpecker) and Sitta pusilla (Brown-headed Nuthatch) is unsurprising because they are cavity nesters and dependent on the presence of pine trees. Another earlysuccessional, ground-nesting bird that breeds predominantly in the southern pinelands is Colinus virginianus (Northern Bobwhite; Askins et al. 2007), but the species inhabits a broader suite of habitats than P. aestivalis and occurs across a wider distribution including the extralimital region. And though the southern pinelands are the primary breeding habitat of L. l. ludovicianus, another migratory subspecies from the eastern extent of the central prairies, L. l. migrans, is thought to have expanded eastward (Cade and Woods 1997). Both L. ludovicianus and T. bewickii appear to have extended their breeding ranges throughout northeastern North America in response to agricultural landscape change concurrent with the P. aestivalis expansion (Graber et al. 1973, Wilcove 1990). These expansions in part composed the region in which P. aestivalis became abundant during its expansion, including the Corn Belt Plains, Interior River Valleys and Hills, the northern portion of the Interior Plateau, and the Allegheny Plateau. Over the same time period in which P. aestivalis was extirpated from this region, L. ludovicianus declined precipitously (Cade and Woods 1997) and T. bewickii was extirpated (Wilcove 1990) despite being generalists that use open landscapes and are more tolerant of brushy and scrubby habitat than P. aestivalis (Kennedy andWhite 2020, Yosef 2020). Another generalist bird of open habitats, C. grammacus, experienced an eastward expansion from the Great Plains (Brooks 1952, Martin andParrish 2020) most consistent with the timing and geographic limits of P. aestivalis. Often the appearance (Brooks 1914), abundance (Wilson 1922), and decline of C. grammacus is mentioned in the contemporaneous literature describing P. aestivalis expansion (Brooks 1934). Notably in the midwestern region of concurrent expansion, P. aestivalis and T. bewickii were extirpated, while L. ludovicianus and C. grammacus precipitously declined but still occur locally.
Prior to this study, modern discussion of the P. aestivalis expansion included little to no discussion of the source region for the expansion (Dunning et al. 2020). Moreover, previous reviews of the expansion suggested it occurred from the southeastern U. S. (Brooks 1938, 1952, Weston 1968). Brooks first hypothesized that the colonization of his focal region began at the Allegheny Plateau of southeastern Ohio from a longstanding source population in Kentucky (Brooks 1938(Brooks , 1952. Weston (1968) further posited that a rapid increase in the southeastern P. aestivalis population following the wholesale deforestation of the southern pine forests ultimately spurred the expansion north, first into Tennessee and then Kentucky before reaching the focal region of Brooks (1938). Our results and historical literature provide some support for this hypothesis as P. aestivalis was present in eastern Tennessee and southcentral Kentucky prior to 1890 ( Fig.  2A) and throughout the early 1900s (Blincoe 1921, Gainer 1921, Wilson 1922. Weston (1968) acknowledges the stark differences between the fallow fields of the Allegheny region and southern pinelands but does not address the habitat differences between the extralimital birds and their southern counterparts; however, P. aestivalis was recorded at the edges of active (Blincoe 1921) and fallow agricultural fields (Strong 1918, Gainer 1921 in Kentucky and Tennessee, and, most relevant to Weston's hypothesis, in fallow fields burned regularly for grazing livestock in Georgia (Smith 1901).
Another plausible source for extralimital regions is the Ozark Highlands of southeastern Missouri and southern Illinois. There is ample support for this northwestern limit of the historical distribution of P. aestivalis as the or at least a source for the expansion from the combination of historical records ( Fig. 2A) and the context of similar and concurrent eastward expansions of several other birds discussed above. This potential source region is not mutually exclusive from Brooks's colonization hypothesis for his focal region. It is possible that individual P. aestivalis found refuge in available habitat in Kentucky and Tennessee (Strong 1918, Blincoe 1921, Gainer 1921 following the deforestation of the southern Missouri highlands (Moser et al. 2007), and also that the population was already expanding into these adjacent regions as habitat became available there. Peucaea aestivalis was common in remaining pine and oak mixed forests, but it was considered rare in clearcuts in southeastern Missouri at the time its first breeding record occurred in the state (Woodruff 1907). Considering the Ozark Highlands as a source region for the expansion is plausible in explaining colonization of the extralimital region, including the northern portions of Illinois and Indiana that have not been adequately addressed. Eastward expansion of P. a. illinoensis from the eastern limit of midwestern transitional oak savannas ( Fig. 1) is more consistent with the broader pattern of grassland specialists expanding eastward during the mid-1800s through the mid-1900s (Askins 2000).
Range expansion by P. aestivalis suggests an ability to colonize emerging suitable habitat; however, P. aestivalis only occurs locally throughout most of its modern distribution and is absent from available and seemingly suitable, early-successional habitat (Dunning et al. 2020). Recent research assessing sparrow occupancy points to the importance of landscape scale factors such as the amount of habitat available at a regional scale as well as the presence of suitable habitat (Taillie et al. 2015). This is consistent with studies highlighting the importance that large blocks of suitable habitat may have in maintaining grassland birds (Herkert 1994a, b, Vickery et al. 1994. Peucaea aestivalis did occur locally in portions of the extralimital region into the 1980s (Fig. 2D), but it is unclear whether the extralimital regions that were colonized contained productive habitats. Furthermore, we know that suitable conditions at the scale of individual territories are best maintained by frequent, ≤ 3-year, fires (Tucker et al. 2004) and this presents additional challenges today (Stephens et al. 2019). Today, P. aestivalis is most abundant in portions of the southeastern U.S. that fall within the former range of longleaf pine (Figs. 1, 2E), a region that leads the nation in its application of prescribed fire (Melvin 2018). In comparison, P. aestivalis is uncommon in the Ouachita and Ozark highlands (Comer et al. 2011, Roach et al. 2019) and at the historical northern limit of shortleaf pines in Kentucky and Tennessee ( Fig. 2E; Barrioz et al. 2013), with the exception of Fort Campbell (Hockman 2013). This may be a result of less consistent prescribed fire in the historically open shortleaf pine woodlands and mixed pine-oak savannas of the region (Guldin 2007, Melvin 2018. Although the extralimital population may have occupied sink habitats that were incapable of supporting stable populations, the alternative explanation is that P. a. illinoensis was more suited to colonizing the emerging suitable habitat, given that it likely exhibited the farthest seasonal migrations of the P. aestivalis subspecies based on its northernmost breeding range. If so, it is possible P. a. illinoensis has been extirpated because of the combined loss of habitat and hybridization with other subspecies during the expansion and retraction. Recent study of microsatellite and mitochondrial DNA throughout the southern extent of the P. aestivalis distribution did not support current subspecies designations but instead reflected panmixia (Cerame et al. 2014). Additional research to resolve the taxonomy of P. aestivalis subspecies and especially that of extralimital individuals is necessary.
The distribution of P. aestivalis, like most species, has likely shifted over evolutionary time as Pleistocene glacial cycles transformed forest communities and shifted their limits throughout eastern North America (Delcourt 2002). The best estimate for the divergence of the P. aestivalis lineage is sometime in the late Pliocene or early Pleistocene (Barker et al. 2015), which is consistent with when pine-dominated, though likely closed, forests established in Florida (Stambaugh et al. 2017). Still, it remains unclear whether during the glacial maximum longleaf pine persisted in Florida or was restricted southwest of its current distribution to present day southern Texas and northeastern Mexico (Stambaugh et al. 2017), where the closest relatives of P. aestivalis occur today. Any climate-related distributional shifts are distinct from the recent expansion and retraction we describe, likely occurring over periods of at least hundreds and often thousands of years. Leading up to and throughout the Holocene, consistent fire use by Native Americans in combination with natural disturbances such as wildfire from lightning strikes resulted in the open forests (Carroll et al. 2002) that dominated the landscape of eastern North America as late as the early 1800s (Hanberry et al. 2020), though southeastern coastal plains did not become pine-dominated until more recently in the mid-Holocene (Stambaugh et al. 2017). The open forests of eastern North America historically averaged fire-return intervals between 2 and 25 years depending on their composition of pine and oaks and other environmental factors but have transitioned to closed forests as a result of fire exclusion since European settlement (Hanberry et al. 2020). Moreover, Native American agriculture and land abandonment in the last several thousand years prior to European settlement and interior population collapse in the 1500s following introduction of European diseases (Carroll et al. 2002) could have prompted P. aestivalis ranges shifts similar in geography and time to this most recent expansion and retraction that we describe.
Although we cannot confirm the absence of P. aestivalis from the extralimital region prior to expansion, we believe that expansion and subsequent retraction are better supported based on the many S. pusilla records we found. The presence of S. pusilla records throughout what would become the P. aestivalis extralimital distribution prior to the P. aestivalis expansion ensures both observation and specimen collection were occurring in the region at the time. We have assumed that had unfamiliar or range extending encounters occurred, a specimen would have been collected, or at a minimum, an account written such as Ridgway's account of his first encounter with P. aestivalis in southern Illinois (Ridgway 1879), Brooks's accounts of the increasing abundance of P. aestivalis from northwestern to central West Virginia (Brooks 1912), and the first specimens taken in Texas (Ragsdale 1892), Virginia (Palmer 1897, Daniel 1899b, southwestern Pennsylvania (CM P32563), Ontario (Saunders 1917), and near Chicago (Eifrig 1915).
In confirming a recent expansion and retraction in P. aestivalis, we highlight it as an interesting case for evaluating the influence of range expansion, and thus gene flow, on the evolutionary trajectory of a grassland bird imperiled by large-scale, anthropogenic landscape change. Peucaea aestivalis is an especially interesting case among other grassland birds that have also experienced recent range expansions driven by anthropogenic landscape change: the extent of the P. aestivalis decline since the expansion is greater and is coupled with persistence near the colonized region in a single, disjunct area at the border of Kentucky and Tennessee (Barrioz et al. 2013, Hockman 2013. Uncertainties surrounding the P. aestivalis range expansion and subsequent retraction could be addressed by continuing to make use of the preserved specimens within natural history collections (Schmitt et al. 2019). Analysis of historical DNA from museum specimens (Billerman and Walsh 2019) could help: (1) identify the source regions for specimens collected in the extralimital region; (2) formally test Weston's (1968) hypothesis that a southern population increase spurred the expansion; and (3) evaluate the validity of the three described subspecies based on estimates of genetic differentiation among specimens from type localities. In addition, stable isotope analysis of feathers of extralimital specimens (e.g., Hobson et al. 2010) could address point 1 and morphological tests of similarity among type specimens could address point 3.

CONCLUSION
Revisiting the P. aestivalis range expansion has yielded a more nuanced understanding of its extent and rate of change. The investigation also provides new insights on the potential source region of the colonization. Collectively, we hope these findings will motivate additional research with better informed hypotheses in addition to highlighting the importance of mixed pine-oak woodlands to the sparrow's future conservation. Similar case studies on other grassland birds of eastern North America are possible using the vast information archived in eBird and natural history collections and could aid in conservation and in understanding the long-term effects of anthropogenic landscape change on the distributions of birds adapted to disturbancemediated habitats.
__________ [1] The common name for this species is Bachman's Sparrow, but this name is currently under review by the North American Classification Committee (NACC) and American Ornithological Society (AOS), which are evaluating the use of eponymous common names. For additional information please visit https:// americanornithology.org/english-bird-names/ and https:// birdnamesforbirds.wordpress.com/historical-profiles/profiles-a-z/ bachman-john/. For this reason, we use the scientific name of this, and for consistency, all bird species throughout this paper.