Research Paper

INTRODUCTION

Habitat loss is the primary threat to maintenance of global biodiversity (Tilman et al. 2017) and has driven biodiversity declines (e.g., Cook and Toft 2005, Vors et al. 2007). Due to the paucity and fragmentation of habitat for a variety of species in many regions, techniques which facilitate animal movement may be required to maintain viable breeding populations (e.g., Brown et al. 2016). Migratory species which utilize relatively ephemeral conditions, often found in marshes, have likely adapted to be flexible when choosing breeding sites among years (e.g., Beletsky and Orians 1994). A variety of environmental cues which indicate aspects of habitat quality, such as food availability (Hromada et al. 2008) or predator presence (Wesner et al. 2012), could ultimately affect where highly mobile, migratory individuals—such as birds—settle to breed each year. Thus, each spring represents an opportunity for habitat managers to a.) modify or restore conditions to influence where species attempt to breed and, depending on the management strategies chosen (Bradshaw et al. 2020), to b.) affect how successful those attempts are. However, the presence of habitat alone may not be enough to initially attract focal species given that, across numerous taxa (Buxton et al. 2020), many species often settle in the vicinity of already-present conspecifics (e.g., in birds; Valente et al. 2021). Managers may wish to actively attract focal species to an area of interest by broadcasting a conspicuous social cue that indicates high-quality habitat, especially when expensive resources were devoted to creating such habitat that has remained unused. Ideally, this would facilitate the precise attraction of focal species to desired, high-quality habitat patches rather than to similar, nearby habitat patches that could be negatively affected by planned management actions (e.g., water drawdown or mowing).

Audio playback of conspecific or heterospecific vocalizations has successfully attracted avian taxa including migratory songbirds (Mönkkönen et al. 1997, Ward and Schlossberg 2004, Anich and Ward 2017) and colonial waterbirds (Ward et al. 2011). These taxa have subsequently used the habitat where playback was broadcast from and could feasibly experience increased nesting success due to proximity to conspecifics or heterospecifics via anti-predator benefits (Sládeček et al. 2014, Šálek et al. 2022). However, conspecific attraction using audio playback has been ineffective for several bird species (e.g., Pérez‐Granados and Traba 2019) despite the overall trend of effectiveness (Valente et al. 2021). Many avian taxa, including secretive marsh birds such as rails (hereafter: marsh birds), have received little study regarding the effectiveness of conspecific or heterospecific cues for attracting individuals to habitat. Although a preliminary study found no evidence of conspecific attraction in a marsh bird (the Sora [Porzana carolina]; Goldberg et al. 2022), several characteristics of marsh birds are conducive to being attracted to habitat by conspecific cues including being migratory, vocalizing nocturnally when migration is occurring, and having clumped distributions (Schlossberg and Ward 2004). Globally, wetlands have likely declined by >85% since 1700 (Davidson 2014). Targeted attraction of marsh birds to remaining wetlands via audio playback could help facilitate discovery and use of breeding habitat managed for this taxon and could even expand geographic ranges (e.g., Anich and Ward 2017). Such efforts could be undertaken with the goal of restoring ecosystem function (Boogert et al. 2006).

Breeding habitat and geographic distribution of Sora and Virginia Rail (Rallus limicola) are similar and both species often co-occur within the same wetland unit (Johnson and Dinsmore 1986). Many regional populations, especially where substantial wetland draining has occurred (e.g., the Midwestern U.S.; Dahl and Allord 1996), are likely significantly smaller than occurred before European colonization of North America. Regional occupancy declines of Sora and Virginia Rail, e.g., in the Great Lakes region of North America (Tozer 2016), could potentially be reversed in part by attracting migratory individuals to habitat via the broadcast of conspecific calls and/or the calls of heterospecifics that use similar habitat. However, multiple factors including conditions in the nonbreeding and migratory range would influence efforts to increase populations. Given their co-occurrence, Sora and Virginia Rail are ideal species for testing the ability of confamilial audio playback (calls of species in Rallidae) to influence which breeding or stopover habitat patches are selected during spring migration by marsh birds. Species such as Yellow Rail (Coturnicops noveboracensis), King Rail (Rallus elegans), and Black Rail (Laterallus jamaicensis) could also benefit from acoustic cues when searching for habitat during spring migration in the Great Lakes Basin, where their rarity has led to various “threatened” and “endangered” statuses.

The primary goal of this study was to test the confamilial attraction hypothesis (attraction to the same taxonomic family) regarding Sora and Virginia Rail during spring migration and to determine the scale, if any, at which attraction to overnight playback occurred and the effects of habitat quality. We also incorporated audio of Yellow Rail and King Rail, rare species in our Midwestern study area, in the playback experiment. Yellow Rail generally only migrate through our study area, which is at the northern edge of the King Rail breeding range. If Sora or Virginia Rail were attracted by rail vocalizations while migrating, then we predicted that, regardless of habitat quality, more rail individuals (“marsh birds”) would be counted during a year when overnight playback was broadcast than during years when it was not. We also predicted that rails would arrive earliest, and occur at the greatest densities, in the immediate vicinity of playback stations. Finally, we predicted that if Yellow Rail, King Rail, or Black Rail were detected, then those detections would occur near to playback stations.

METHODS

Study area

This study occurred within impounded coastal wetlands during 2021 and 2022 in central Michigan, USA, at two adjacent, publicly-owned properties—Shiawassee River State Game Area (SGA) and Shiawassee National Wildlife Refuge (NWR)—which together encompassed ~ 80 km². These properties occurred near the center of Saginaw County, which encompasses approximately 2100 km² and includes several marsh complexes. Dominant landcover at both properties consisted of a mix of perennial herbaceous emergent vegetation, emergent woody vegetation, upland forests, and moist-soil units (which were managed to contain less perennial, and more annual, wetland plants) within 5 km of the Shiawassee River. Both Soras and Virginia Rails had been commonly documented (>5 times) via eBird (Basic Dataset 2023) throughout both properties during the five years before our study occurred, primarily within herbaceous emergent vegetation. There had been two previous King Rail reports (in 2019 and 1996) and one Yellow Rail report (in 2001) by eBird contributors, suggesting that habitat may occur in our study area for these less-common species. There were no eBird reports of Black Rails on either property, though this uncommon species could feasibly occur in the study area.

Pre-experiment monitoring

We assumed that all marsh units could serve as stopover or breeding habitat for our focal species. All such units contained vegetation, though in some cases only at margins, to provide cover and water to generate food, which together could facilitate occupancy and so constitute habitat (Hall et al. 1997). However, we assumed that the relative quality of habitat (which we define as the degree to which a patch could support rails) within these units could be inferred based on detection or non-detection of our focal species which are often not detected, especially at low population densities, even when present (Conway and Gibbs 2005). We assumed that higher-quality habitat (hereafter HQH) would support more individuals in the absence of any experimental manipulation, thus increasing the likelihood of us detecting individuals that were present. Similarly, we assumed that lower-quality habitat (LQH) would support substantially less individuals and so generally result in non-detection in the absence of experimental manipulation.

In 2021, we categorized units within our study area as HQH and LQH based on detection and non-detection, respectively, of Soras and/or Virginia Rails (only focal species detected). We determined within-unit presence of focal species in 2021 after a single detection and non-detection after ≥3 surveys (methodology to follow) without a detection. We assumed that subsequent presence of focal species at LQH units in 2022 would be due primarily to experimental manipulation—i.e., overnight playback of rail calls—that occurred in 2022 but not in 2021 given the relative constancy of environmental conditions in the diked units that we studied. By including LQH as a focus of our study we attempted to isolate effects of overnight playback on migratory rails. By including HQH in our experiment, we sought to demonstrate results that wetland managers may experience and to determine how habitat quality, in addition to our primary factor of interest—overnight playback of rail calls—affected site selection by our focal species. For the purposes of our exploratory study, we assumed that less-common rails (King Rails, Yellow Rails, and Black Rails) would be more likely to occur in the units we considered HQH than in the units we considered LQH, given the general need of these species for dense, herbaceous emergent vegetation (Eddleman et al. 1988). Although sites could have been settled due to factors other than habitat quality, we used a priori predictions of occurrence or absence to strengthen our subsequent habitat categorizations.

An experienced observer (D.B.) conducted call-broadcast surveys (hereafter: “surveys”) weekly between 8 April and 14 June in 2021, which included the migratory and breeding periods for focal species, to categorize wetland units as LQH or HQH. Water depth, a useful predictor of rail occupancy (Malone et al. 2021), was measured immediately before each survey. Surveys occurred between 30 minutes before sunrise and 3.5 hours after sunrise and were based on the Standardized North American Marsh Bird Monitoring (SNAMBM) protocol guidelines (Conway 2011). Individuals of all bird species detected within a 125-m radius of survey points were noted. Audio tracks included 1 minute of vocalizations with call bouts separated by ~ 5 seconds of silence for each focal species, except for Black Rails (no Black Rail calls were broadcast due to rarity of this species and logistical constraints), followed by 1 minute of silence such that species were presented in order of ascending “intrusiveness” (i.e., loudness; Conway 2011).

These pre-experimental surveys occurred both within units where, based on communication with property staff and eBird records, we expected to detect Soras and/or Virginia Rails (predicted HQH units) at least once and within units where we did not expect to detect these species (predicted LQH units). Predicted HQH units (n = 7) were dominated by perennial, herbaceous emergent vegetation such as Typha. Predicted LQH units (n = 6) were experiencing moist-soil management regimes meant primarily to benefit waterfowl during the nonbreeding season (via the seeds produced by annual plants), though contained wetland vegetation—especially at the edges of units—that could feasibly be used as cover by marsh birds. Such moist-soil units are known to be used by marsh birds during migration (Wilson et al. 2018).

We randomly selected survey points with the constraint that they occurred ≤50 m from the base of a dike within a wetland unit not dominated by woody, floating, or submergent vegetation. Surveys generally occurred at each point every other week and were completed in the same order each day. Using this approach, we categorized five units as LQH and seven units as HQH (Appendix 1 for additional details).

Playback experiment

We selected seven pairs of points (14 points total), in seven separate units, that were used for the experiment. These units each included an experimental point—where overnight playback was broadcast from—and a control point separated by 250 m (Fig. 1). This distance is greater than that used by some secretive marsh bird studies, such as Webb et al. (2022; 150 to 200 m), and represented the finest scale at which we tested the precision of overnight playback as an attractant. The units in which these pairs occurred were chosen, in part, to maximize the number of surveys that could be completed in a morning. Four point-pairs occurred within LQH and three point-pairs occurred within HQH. A fourth point-pair where overnight playback occurred in HQH was excluded from analysis (Appendix 1). We assumed that within-unit pairing of experimental and control points would minimize environmental variance between them and improve comparison. The centers of each point-pair were ≥800 m from all other point-pair centers (Fig. 1) and occurred at pre-experimental survey locations from 2021 except in cases where conditions, like water level, required within-unit shifting of point-pairs. Logistical constraints required that the experimental point within each pair be closest to an access point. We clustered point-pairs by habitat category such that all LQH point-pairs were at least 2.5 km from the nearest HQH point-pair (Fig. 1). We intended for this distance to allow for a comparison of focal species response between the area where audio was broadcast overnight from LQH (“LQH playback zone”) and the area where audio was broadcast overnight from HQH (“HQH playback zone”). We established three unpaired survey points within HQH (each in a separate unit; Fig. 1), where no playback occurred, in units adjacent to those containing LQH point-pairs. These HQH control points were ≥800 m from overnight playback and >2.5 km from the nearest playback from within HQH. Thus, we were able to investigate a habitat-driven question and a scale-driven question via use of these unpaired HQH points. The habitat-driven question was: can overnight playback within LQH induce rails to preferentially use LQH adjacent to HQH? The scale-driven question was: will our unpaired HQH control points (~800 m from overnight playback) attract less rails than our paired HQH control points (250 m from overnight playback) or our experimental HQH points (0 m from overnight playback). In the HQH playback zone, there were no LQH units adjacent to HQH where unpaired points could be placed.

The group to which our focal species belong migrate at night (e.g., Harrity and Conway 2020). Therefore, we broadcasted audio of rail breeding calls overnight from ~1.5 hours after sunset to 1.5 hours before sunrise (start and stop times reprogrammed every ~2 weeks). Calls were downloaded from xeno-canto. Playback stations consisted of a FOXPRO dual-speaker game caller (Model: NX4) connected to an AMICO digital timer (Model: CN101A) and a marine battery within a plastic box that was placed on a platform 1 m above the ground (Appendix 1: Fig. A1). The audio track that we broadcast overnight consisted of a 1-hour recording of calls from our focal species, except for Black Rail, that was continuously looped during broadcast hours (Appendix 1, for details about calls used).

We broadcast audio from experimental points within point-pairs between 25 March and 14 June, 2022. Due to logistical constraints (e.g., flooding), playback stations in the LQH playback zone began broadcasting audio the night of 25 March, whereas within the HQH playback zone one station began broadcasting audio the night of 29 March and three others began broadcasting audio the night of 2 April. However, these later dates were before we anticipated focal species would arrive to the study area. Autonomous recording unit (ARU) data suggest that this presumption was accurate.

We used ARUs to establish when focal species arrived at each point and to determine how long these species remained. ARUs were programmed to record from 2 hours before sunrise to 3 hours after sunrise at all survey points in the LQH playback zone. We did not have enough ARUs to deploy in the HQH playback zone. The ARUs were primarily (10 of 12) AudioMoths (Open Acoustic Devices; model: 1.0.0; sample rate: 32 kHz). We deployed two SM4 recorders (Wildlife Acoustics; sample rate: 24 kHz) in separate units, both where no playback occurred, within the LQH playback zone. All recordings produced were in the WAV format and occurred between 18 March and 14 June. On all days between 25 March and 6 June the majority of ARUs were recording, though technical issues caused random, sporadic gaps in the recording effort. We assumed that focal species would be detectable in audio recordings from approximately the edges of our survey plots (125 m radius) and not beyond them (Turgeon et al. 2017).

We conducted surveys at each point to estimate rail abundance. These surveys occurred every other week, with weekly surveys alternating between zones (Fig. 1), and used the same methodology described above in the “Pre-Experiment Monitoring” section between 28 March and 14 June, 2022. This resulted in each point being surveyed two times per survey period (survey period details to follow). To minimize habituation, the audio used in these surveys was different from the audio that was broadcast overnight (but, again, breeding rail calls were broadcast). We completed surveys ≥1 hour after game callers had automatically turned off. Before every survey we measured water depth at the survey point and quantified the percent coverage of herbaceous vegetation within a 5-m radius plot surrounding each survey point. We also recorded the most dominant plant species in each plot. After completing each survey at experimental points, the observer checked the on-site playback station to ensure that it was functioning properly and adjusted automatic timers, to account for changing sunrise, for the playback stations and AudioMoths.

Analyses

Our analytical approach sought to accomplish related objectives (Table 1). Objective 1 was to determine if rail abundance and occurrence (definitions below) were greater in the year when overnight playback was broadcast compared to years when it was not and how this related to habitat quality. Objective 2 was to determine if proximity to overnight playback affected rail arrival and abundance during the year when overnight playback was broadcast and how this related to habitat quality.

We compared experimental and control treatments based on in-person surveys at three spatiotemporal scales: a.) between-year (2021, 2022), within-unit (objective 1); b.) within-year (2022), within-unit (objective 2); and c.) within-year (2022), between-unit (objective 2; Table 1). Treatments occurred at the unit scale (for between-unit comparison) and at the point scale (for within-unit comparison), which both included HQH and LQH (Fig. 1). Experimental treatments were defined by broadcast of overnight playback (in 2022) and control treatments were defined by the absence of overnight playback (in 2021 and 2022).

We split the survey period into the following temporal segments for analysis: initial migration (28 March to 20 April), peak migration (21 April to 18 May), and breeding (19 May to 14 June). We defined these periods based on available life history knowledge for Soras and Virginia Rails (e.g., Fournier et al. 2015, Hengst 2021), though acknowledge that they are approximate. We pooled counts of individual rails across all focal species during each survey, thus reducing zero-inflation and number of statistical comparisons, to provide an index of rail abundance at survey points (hereafter, rail abundance).

We completed all inferential analyses in R (4.1.3) and used linear mixed models via the lme4 package (Bates et al. 2015) with a random intercept to account for repeated measures at each location. Due to our small sample size, we used an alpha value of 0.1 to minimize the prevalence of false negatives. Post-hoc comparisons were conducted using the emmeans package (Lenth 2022). All initial models included only one predictor variable to avoid overfitting. Given our small sample size, and the consequent difficulty in some cases of satisfying distributional assumptions, parameter estimates should be interpreted with caution though are likely unbiased (Schielzeth et al. 2020). For within-unit comparisons, we calculated the difference between experimental and control entities (points or units) for each spatially-linked pair of observations regarding all response and predictor variables. We refer to this value for each variable as “relative variable name” (e.g., “relative rail abundance”). Thus, a significant finding for an intercept-only model indicated a main effect, with a positive value indicating that, overall, experimental entities were larger than their respective controls whereas a negative value indicated the opposite. For between-unit comparisons, both survey points were used to calculate a unit mean for each variable in units where an experimental and control point existed and otherwise one point was used (Fig. 1). Each variable was generally measured two times at each point per period. For all analyses, we calculated means across repeated measurements within the scale of inquiry (point or unit) for each variable within each period. Therefore, depending on the analysis, points or units in each period had a single value for each variable and zero-inflation was reduced.

We sought to determine if site characteristics (water depth and/or herbaceous cover) varied as predicted between habitat categories and to confirm that site characteristics did not vary between experimental and control treatments in a fashion that could have influenced results. We also investigated the effects of habitat category (HQH or LQH) and site characteristics, experimental audio playback, and period (initial migration, peak migration, or breeding) on rail abundance. Refer to Appendix 1 (Table A1) for a comprehensive overview of models.

We used eBird data (Basic Dataset 2023) to further evaluate objective 1 (Table 1). Specifically, we compared rail abundance and occurrence during the year when overnight playback occurred (2022) compared to other years at the scale of our study area—i.e., the adjacent properties constituting the wetland complex where treatments occurred—and in Saginaw County, wherein the wetland complex was located. We predicted that overnight playback would be associated with a greater increase in rail abundance and occurrence at the scale of our study area compared to the broader county-level scale due to the locality of the treatment. This also allowed us to confirm that results during our experimental year were not driven by broader-scale factors. We summarized eBird observations submitted via “checklists”—which list all individual birds identified during a session by an observer or group—between 2013 and 2023. This allowed us to compare the control year (2021), the experimental year (2022), and the post-experimental year (2023) to the preceding eight-year period at two different scales. We compiled all observations between 1 April and 31 July, when we assumed focal species were most likely to be present, and confirmed that no observations occurred during minutes when overnight playback had been broadcast from our stations. We calculated two standardized values for each year which indicated temporal change in marsh bird relative abundance (number of marsh bird individuals reported per checklist) and occurrence (proportion of checklists with at least one marsh bird reported).

We used BirdNET (Kahl et al. 2021) to analyze every other ARU recording until at least one of our five focal species was confirmed via manual review, thus further addressing objective 2 (Table 1). We then analyzed each day prior to first detection at each point to increase the precision of our estimate regarding when focal species arrived. Similarly, we analyzed recordings at all points during the week before and after a King Rail was detected in-person to gain information about when this individual arrived and where it may have moved.

RESULTS

We detected only Soras (n = 10, control year; n = 61, experimental year) and Virginia Rails (n = 22, control year; n = 66, experimental year) during formal, in-person surveys. Therefore, only these species were used to calculate rail abundance (i.e., Black Rail, Yellow Rail, and King Rail did not contribute). Only outputs of models that indicated evidence of an effect are reported in the main text to improve readability. Output for all models used in this study, including those that did not indicate an effect, are provided as tables in Appendix 1 (A2, A3, and A4).

Objective 1: Were more rails counted during the experimental year?

We found that water depth tended to be deeper in the experimental year than in the control year (mean difference = 5.49 ± 2.21 cm [SE]; t = 2.48; df = 6; P = 0.05). However, relative water depth did not predict relative rail abundance. As predicted, we detected greater rail abundance during the experimental year than in the same units during the control year (mean difference = 0.86 ± 0.38 [SE]; t = 2.25; df = 6; P = 0.07), a 218% increase, and this difference varied by period (F[2, 12] = 4.5; P = 0.04). Post-hoc comparison indicated (t = -2.90, P = 0.03) that mean relative rail abundance was 2.18 ± 0.75 [SE] higher during peak migration compared to initial migration. Rail abundance was particularly increased in the peak migration and breeding periods in HQH during the experimental year compared to the control year (Fig. 2). We found that habitat category and period interacted (F[2, 10] = 2.91; P = 0.1) such that relative rail abundance was greater in HQH than in LQH during peak migration (mean difference = 2.88 ± 0.38; t = 2.88; df = 14.6; P = 0.1) but not the other periods. However, we were unable to statistically test if habitat category alone affected relative rail abundance due to a singular fit.

Analysis of eBird data indicated that our experimental year (2022) had exceptionally high rail abundance and occurrence in our study area (greater than the 95 percent confidence interval of the previous eight years), as did our post-experimental year (2023) to a lesser degree, though our control year (2021) did not (Fig. 3). During our experimental year within our study area, there were 0.93 marsh birds per eBird checklist (n = 558) whereas the 95% CI for the mean of the pre-study period was 0.25 to 0.74. At this scale, there were more marsh birds per eBird checklist during our experimental year than any previous year analyzed except for 2015, when there were 1.28 marsh birds per checklist (n = 286) following a notable increase in water levels in nearby Lake Michigan–Huron (refer to Discussion). The proportion of checklists with a marsh bird detected (0.26) within our study area was higher during the experimental year compared to the upper bound of the 95% CI for the mean of the pre-study period (0.11 to 0.16). This proportion was higher than any previous year, including 2015 (0.17). The post-experimental year had the second highest occurrence rate (0.18; n = 866) in our study area. At the county scale, neither the control, experimental, nor post-experimental years were exceptional compared to the eight years before field work began regarding abundance or occurrence (Fig. 3).

Rail abundance, based on our surveys, was greatest during peak migration in both years and for all unit-level treatments that resulted in a rail detection (Fig. 4). During peak migration, rail abundance was especially elevated where overnight playback was broadcast from HQH and nearly all LQH detections occurred in this period (all LQH detections were associated with overnight playback; Fig. 4). Sora and Virginia Rail were more abundant during the experimental year than during the control year based on total number of individuals detected and number of individuals detected per survey (Fig. 5). Although the number of detections were particularly increased during peak migration for both species during the experimental year, only Virginia Rails appeared to be markedly more abundant (~0.4 more individuals/survey) during the breeding period of the experimental year (Fig. 5).

On 25 May of the experimental year, we detected a King Rail in a LQH unit where overnight playback was being broadcast (LQH unit furthest to east; Fig. 1) while preparing to survey elsewhere. Refer to Appendix 1 for additional detail about this observation.

Objective 2: Did proximity to playback affect rail abundance/arrival?

We did not find evidence of differences between experimental and control points within units regarding site characteristics (as predicted) or regarding relative rail abundance (contrary to prediction; Fig. 2). Neither relative herbaceous cover, habitat category, nor period predicted relative rail abundance. We were unable to determine if relative water depth influenced relative rail abundance due to a singular fit.

We confirmed that, as predicted, HQH units during our experimental year had deeper mean water depths than LQH units (30.11 ± 4.12 cm vs. 5.37 ± 5.08 [SE] cm; t = 3.77; df = 8; P = 0.01), though herbaceous vegetative cover, by percentage, did not vary between HQH and LQH units. Cocklebur (Xanthium sp.) and foxtail (Setaria sp.) were each the most dominant plant species at two LQH units, whereas, in HQH, cattail (Typha sp.) was most dominant in five units and a graminoid in another unit. As predicted, there was greater mean rail abundance at experimental HQH units (2.57 ± 0.68 [SE]) compared to experimental LQH units (0.35 ± 0.59 [SE]; t = 2.48; df = 5; P = 0.06). However, there was not greater rail abundance at experimental LQH units than at nearby control HQH units (contrary to prediction). Upon moving beyond comparison of HQH and LQH, we found no differences between experimental and control HQH units regarding water depth (as predicted), herbaceous vegetation coverage (as predicted), or mean rail abundance (contrary to prediction; but refer to discussion for potential explanation). Finally, we found no evidence that experimental and control units, without respect to habitat category, differed regarding mean rail abundance (contrary to prediction).

Our comparison of rail abundance within HQH at three proximities (0 m, 250 m, and 800 m) to overnight playback revealed no significant difference in rail abundance. However, when counts were averaged across all periods, mean rail abundance within HQH was greatest at 0 m (2.9 rails per survey) from overnight playback, less at 250 m from playback (2.2), and least at 800 m from playback (1.1). This trend was in accord with our prediction that increasing distance from overnight playback within HQH would be associated with reduced rail abundance, though no such trend was apparent in LQH at the within-unit scale (0 m from playback: 0.25 rails per survey; 250 m from playback: 0.55 rails per survey).

We detected at least one focal species at every point in 2022 (n = 11) where ARUs were deployed (all in the LQH playback zone), including in four LQH units where no focal individuals were detected during surveys in 2021. Among our focal species, only Soras, Virginia Rails, and a King Rail were detected via ARU. Our first focal rail detections (Sora and Virginia Rail) occurred at experimental and control points within HQH on 5 April (Table 2) in both playback zones. Both experimental LQH points that we had ARU data for indicated that Sora and Virginia Rail arrived within three days of the first overall detections, though at three of the four control LQH points the first detection via ARU did not occur until ≥2 weeks later (Table 2). The first in-person detections occurred after the first ARU detections at every applicable point (mean days later = 16.2; Table 2). ARU recordings only detected a King Rail in the eastern-most LQH unit of the LQH playback zone at the experimental point (Fig. 1). This detection occurred the morning (25 May) that a King Rail was detected by the observer who was arriving for the day’s surveys that occurred elsewhere.

DISCUSSION

Habitat selection by migratory birds can be impacted by social and non-social environmental cues during a multi-step process that begins at a broad spatial scale and continues through progressively finer scales (Johnson 1980). Our results provide insights about how Sora and Virginia Rail responded to overnight playback during migration and about the precision with which this tool could be used to attract these species.

We found greater rail abundance during the experimental year relative to the same units during the control year, which supports the confamilial attraction hypothesis as it pertains to Soras and Virginial Rails. The between-year difference within our study area was apparently driven by an increased rail abundance in HQH during the experimental year (Fig. 2), which in addition to our within-year results underscores the importance of habitat quality. The effect was particularly apparent during the peak migration and breeding periods for both species (Fig. 2), though by the breeding period most Soras had apparently left the area and so Virginia Rails drove the effect. Differences in rail abundance at a location between any two years could be due to a variety of factors, especially given the dynamic nature of wetlands (Keddy 2010), and so comparing our experimental year to additional years was necessary. Compared to the eight years before field work began, eBird data within our study area suggest that our experimental year, but not our control year, was exceptional regarding number of rail individuals reported per checklist and proportion of checklists with a rail reported within our study area (Fig. 3). These data indicate that the post-experimental year also had increased rail abundance and occurrence, though to a lesser degree than the experimental year, indicating the possibility of a weak carry-over effect between years. Given that the increase indicated by eBird data was much more pronounced at the study area scale than the county scale (Fig. 3), it appears that our results were not driven by broader-scale migration patterns. Although in 2015 there were more rails per checklist than during our experimental year, this may have been due to increasing Lake Michigan–Huron water levels, which were above the long-term average for a full year for the first time in over 15 years (NOAA 2024). This uncertainty, however, demonstrates that additional study is required to confirm the trends that we identified. Nonetheless, taken together, our in-person surveys and eBird data suggest that overnight playback contributed to increased rail abundance and occurrence in our study area.

Soras and Virginia Rails were commonly detected during both years of our study and generally appeared to respond similarly to overnight playback (Fig. 2). Our data indicate that Soras primarily migrate through the study area, as Hengst (2021) found regarding this species in northwestern Ohio, whereas Virginia Rails are more likely to stay to breed especially if overnight playback occurs. It is feasible that overnight playback could be useful for attracting focal species to both stopover habitat (e.g., Soras in our study area)—if habitat associations are considered (Webb et al. 2022)—and to breeding habitat (e.g., Virginia Rails in our study area). In addition to migratory behavior, future investigators should consider life history traits of Soras and Virginia Rails, such as diet (Horak 1970), when identifying species-specific factors that could influence effectiveness of overnight playback as a conservation tool.

Surveys indicated that rails were somewhat common in LQH units where overnight playback occurred during peak migration (Fig. 4; detected at 5 of 8 LQH points in the LQH playback zone), but rail detections were nearly absent from surveys there during the breeding period (Fig. 4; no detections at seven of eight points). ARU data indicated presence of focal species at three of these seven LQH points during the breeding period, which could indicate that overnight playback can cause sustained use of LQH (as in Betts et al. 2008) by rails. Regardless, the contrast between breeding period survey detections in HQH (9 of 9 points) and in LQH (1 of 8) may indicate that LQH, if used, tends to be visited by migrants that do not stay to breed.

Overnight playback at our sites likely contributed to the attraction of at least one King Rail, a marsh bird of conservation concern, to our study area. Based on our observations, and analysis of ARU data, we assume that a King Rail arrived to the LQH playback zone during the night of 25 May. By 1 June, that bird may have begun settling into a home range in the HQH unit immediately south of where it was initially attracted to. Anecdotal reports suggest that a breeding attempt may have occurred at this location, though only one individual King Rail was noted in eBird reports and we found no direct breeding evidence. The last eBird observation of a King Rail in our study area was on 4 July and all observations (N=17) were concentrated in the same area. Over 80,000 people visit Shiawassee NWR annually with the majority (beginning 1 June most years) driving vehicles on the wildlife drive that passes through the center of our LQH playback zone. Many of them are skilled bird observers and would accurately report encounters with a King Rail. However, prior to our experimental year, there had only been two King Rail observations reported on eBird in our study area. Given their rarity in the region, Black and Yellow Rails require further study regarding attraction to overnight playback.

We found little evidence that our overnight playback approach can attract rails specifically to a point or even a wetland unit. Rather, our results are consistent with attraction of rails broadly to the wetland complex where overnight playback occurred. Rails may have settled around playback stations in a radiating pattern that led to high densities even in areas where no playback occurred. This may have also happened in an investigation of rail response to conspecific cues that did not find an effect of overnight playback (Goldberg et al. 2022). Thus, managers may not be able to specifically attract rails to habitat patches without also attracting them to nearby habitat patches. Using fewer playback stations within a wetland complex may increase ability to attract rails to specific habitat patches.

Our results suggest that habitat quality is important for rail habitat selection. Rail abundance tended to be less at LQH sites where overnight playback occurred compared to nearby HQH sites where it did not occur (Fig. 4). Our finding that, among experimental units, rail abundance was greater within HQH than LQH further illustrates the importance of habitat quality when rails are selecting habitat. Cornell and Donovan (2010) also demonstrated the importance of habitat quality, even in the presence of social cues, when a songbird selected habitat. However, other studies have found that social cues can lead to preferential selection of LQH by birds (e.g., Betts et al. 2008).

Our ARU data, though limited by equipment failure, suggest that overnight playback at LQH may have initially attracted rails at approximately the same time as HQH without overnight playback (Table 2). LQH without overnight playback, however, may have experienced delays in rail arrival by ≥2 weeks (Table 2). Soras and Virginia Rails first arrived to our study area on 5 April in 2022, which is comparable to previous estimates of arrival by these species in northwestern Ohio based on first captures (24 March to 17 April; Fournier et al. 2015). Arrival dates of focal species could be considered by managers deciding when to manipulate water levels or other habitat variables and ARUs clearly help to identify these dates.

The logistical challenges of this study (e.g., moving heavy batteries through marsh) limited sample size, statistical power, and the breadth of inference possible. We encourage further experimental efforts following similar methods to test overnight playback as a conservation tool. Such investigations could refine inference from confamilial to conspecific attraction and so advance understanding of this phenomenon in birds (Valente et al. 2021) and in general (Buxton et al. 2020). Ideally, audio playback would be broadcast as little as possible to reduce potential negative effects, such as repelling other species (Fletcher Jr. 2007) and/or masking their vocalizations. Quantifying potential negative effects of attraction to a location, such as increased depredation (Ward et al. 2011), disease rates, or lower reproductive output in broadcast areas, could inform conservation action by managers and provide useful, but thus far lacking, information. Until such knowledge is generated, ideally via long-term monitoring, conservation practitioners should be aware of the potential for unknown negative outcomes. Further, high-quality habitat should be carefully identified before attempting to attract focal species with social cues to avoid inadvertently creating an ecological trap (Battin 2004).

Our study indicates that overnight playback can attract migratory rails, including Soras, Virginia Rails, and potentially King Rails, to a wetland complex. Wetland managers may thus be able to attract these species via broadcasted cues during years when, for example, appropriate water depths will be maintained throughout a large proportion of a wetland complex. Efforts to restore wetland biodiversity and ecosystem function could benefit from using overnight playback of rail calls as a conservation tool.

LITERATURE CITED

Anich, N. M., and M. P. Ward. 2017. Using audio playback to expand the geographic breeding range of an endangered species. Diversity and Distributions 23(12):1499-1508. https://doi.org/10.1111/ddi.12635

Bates, D., M. Mächler, B. Bolker, and S. Walker. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67(1):1-48. https://doi.org/10.18637/jss.v067.i01

Battin, J. 2004. When good animals love bad habitats: ecological traps and the conservation of animal populations. Conservation Biology 18(6):1482-1491. https://doi.org/10.1111/j.1523-1739.2004.00417.x

Beletsky, L. D., and G. H. Orians. 1994. Site fidelity and territorial movements of males in a rapidly declining population of Yellow-Headed Blackbirds. Behavioral Ecology and Sociobiology 34:257-265. https://doi.org/10.1007/BF00183476

Betts, M. G., A. S. Hadley, N. Rodenhouse, and J. J. Nocera. 2008. Social information trumps vegetation structure in breeding-site selection by a migrant songbird. Proceedings of the Royal Society B: Biological Sciences 275(1648):2257-2263. https://doi.org/10.1098/rspb.2008.0217

Boogert, N. J., D. M. Paterson, and K. N. Laland. 2006. The implications of niche construction and ecosystem engineering for conservation biology. BioScience 56(7):570-578. https://doi.org/10.1641/0006-3568(2006)56[570:TIONCA]2.0.CO;2

Bradshaw, T. M., A. G. Blake-Bradshaw, A. M. V. Fournier, J. D. Lancaster, J. O’Connell, C. N. Jacques, M. W. Eichholz, and H. M. Hagy. 2020. Marsh bird occupancy of wetlands managed for waterfowl in the Midwestern USA. PLOS ONE 15:e0228980. https://doi.org/10.1371/journal.pone.0228980

Brown, N. L., M. M. Peacock, and M. E. Ritchie. 2016. Genetic variation and population structure in a threatened species, the Utah Prairie Dog Cynomys parvidens: the use of genetic data to inform conservation actions. Ecology and Evolution 6(2):426-446. https://doi.org/10.1002/ece3.1874

Buxton, V. L., J. K. Enos, J. H. Sperry, and M. P. Ward. 2020. A review of conspecific attraction for habitat selection across taxa. Ecology and Evolution 10(23):12690-12699. https://doi.org/10.1002/ece3.6922

Conway, C. J. 2011. Standardized North American marsh bird monitoring protocol. Waterbirds 34(3):319-346. https://doi.org/10.1675/063.034.0307

Conway C. J., and J. P. Gibbs. 2005. Effectiveness of call-broadcast surveys for monitoring marsh birds. The Auk 122(1):26-35. https://doi.org/10.1093/auk/122.1.26

Cook, L. F., and C. A. Toft. 2005. Dynamics of extinction: population decline in the colonially nesting Tricolored Blackbird Agelaius tricolor. Bird Conservation International 15(1):73-88. https://doi.org/10.1017/S0959270905000067

Cornell, K. L., and T. M. Donovan. 2010. Scale-dependent mechanisms of habitat selection for a migratory Passerine: An Experimental Approach. The Auk 127(4):899-908. https://doi.org/10.1525/auk.2010.09171

Dahl, T. E., and G. J. Allord. 1996. Technical aspects of wetlands: history of wetlands in the conterminous United States, United States Geological Survey water supply paper 2425. United States Geological Survey.

Davidson, N. C. 2014. How much wetland has the world lost? Long-term and recent trends in global wetland area. Marine and Freshwater Research 65(10):934-941. https://doi.org/10.1071/MF14173

eBird Basic Dataset. 2023. Version: EBD_relAug-2023. Cornell Lab of Ornithology, Ithaca, New York, USA.

Eddleman, W. R., F. L. Knopf, B. Meanley, F. A. Reid, and R. Zembal. 1988. Conservation of North American Rallids. The Wilson Bulletin 100:458-475.

Fletcher Jr, R. J. 2007. Species interactions and population density mediate the use of social cues for habitat selection. Journal of Animal Ecology 76(3):598-606. https://doi.org/10.1111/j.1365-2656.2007.01230.x

Fournier, A. M. V., M. C. Shieldcastle, T. Kashmer, and K. A. Mylecraine. 2015. Comparison of arrival dates of rail migration in the Southwest Lake Erie Marshes, Ohio, USA. Waterbirds 38(3):312-314. https://doi.org/10.1675/063.038.0313

Goldberg, D. L., A. M. Tauber, M. P. Ward, and A. P. Capparella. 2022. Soras over Illinois: nocturnal settlement patterns in Porzana carolina are site-specific but show no evidence of visual or acoustic cues. Waterbirds 45(1):82-90. https://doi.org/10.1675/063.045.0110

Hall, L. S., P. R. Krausman, and M. L. Morrison. 1997. The habitat concept and a plea for standard terminology. Wildlife Society Bulletin 25:173-182.

Harrity, E. J., and C. J. Conway. 2020. Satellite transmitters reveal previously unknown migratory behavior and wintering locations of Yuma Ridgway’s Rails. Journal of Field Ornithology 91(3):300-312. https://doi.org/10.1111/jofo.12344

Hengst, N. M. 2021. Movements and habitat relationships of Virginia Rails and Soras within impounded coastal wetlands of Northwest Ohio. The Ohio State University, Columbus, OH, USA.

Horak, G. J. 1970. A comparative study of the foods of the Sora and Virginia Rail. The Wilson Bulletin 82:206 -213.

Hromada, M., M. Antczak, T. J. Valone, and P. Tryjanowski. 2008. Settling decisions and heterospecific social information use in shrikes. PLOS ONE 3(12):e3930. https://doi.org/10.1371/journal.pone.0003930

Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61(1):65-71. https://doi.org/10.2307/1937156

Johnson, R. R., and J. J. Dinsmore. 1986. Habitat use by breeding Virginia Rails and Soras. The Journal of Wildlife Management 50(3):387-392. https://doi.org/10.2307/3801092

Kahl, S., C. M. Wood, M. Eibl, and H. Klinck. 2021. BirdNET: a deep learning solution for avian diversity monitoring. Ecological Informatics 61:101236. https://doi.org/10.1016/j.ecoinf.2021.101236

Keddy, P. A. 2010. Wetland ecology: principles and conservation. Cambridge University Press, Cambridge, UK. https://doi.org/10.1017/CBO9780511778179

Lenth, R. V. 2022. Emmeans: estimated marginal means, aka least-squares means.

Malone, K. M., E. B. Webb, D. Mengel, L. J. Kearns, S. W. Matteson, and A. E. McKellar. 2021. Secretive marsh bird habitat associations in the Mississippi Flyway: a meta-analysis. Avian Conservation and Ecology 16(2):12. https://doi.org/10.5751/ACE-01924-160212

Mönkkönen, M., P. Helle, G. J. Niemi, and K. Montgomery. 1997. Heterospecific attraction affects community structure and migrant abundances in northern breeding bird communities. Canadian Journal of Zoology 75(12):2077-2083. https://doi.org/10.1139/z97-842

NOAA. 2024. The Great Lakes water level dashboard. Great Lakes Environmental Research Laboratory. https://www.glerl.noaa.gov/data/wlevels/dashboard/

Pérez-Granados, C., and J. Traba. 2019. Testing the conspecific attraction hypothesis with Dupont’s Larks, a resident species of songbird in Central Spain. Journal of Field Ornithology 90(3):277-285. https://doi.org/10.1111/jofo.12306

Šálek, M., M. Sládeček, V. Kubelka, J. Mlíkovský, D. Storch, and P. Šmilauer. 2022. Beyond habitat: effects of conspecific and heterospecific aggregation on the spatial structure of a wetland nesting bird community. Journal of Avian Biology 2022(2):e02928. https://doi.org/10.1111/jav.02928

Schielzeth, H., N. J. Dingemanse, S. Nakagawa, D. F. Westneat, H. Allegue, C. Teplitsky, D. Réale, N. A. Dochtermann, L. Z. Garamszegi, and Y. G. Araya‐Ajoy. 2020. Robustness of linear mixed-effects models to violations of distributional assumptions. Methods in Ecology and Evolution 11(9):1141-1152. https://doi.org/10.1111/2041-210X.13434

Schlossberg, S. R., and M. P. Ward. 2004. Using conspecific attraction to conserve endangered birds. Endangered Species Update 21:132-138.

Sládeček, M., V. Kubelka, J. Mlíkovský, and M. Šálek. 2014. Coping with nest predation risk in a species-rich bird community inhabiting a Siberian wetland. Folia zoologica 63(4):256-268. https://doi.org/10.25225/fozo.v63.i4.a5.2014

Tilman, D., M. Clark, D. R. Williams, K. Kimmel, S. Polasky, and C. Packer. 2017. Future threats to biodiversity and pathways to their prevention. Nature 546:73-81. https://doi.org/10.1038/nature22900

Tozer, D. C. 2016. Marsh bird occupancy dynamics, trends, and conservation in the southern Great Lakes Basin: 1996 to 2013. Journal of Great Lakes Research 42(1):136-145. https://doi.org/10.1016/j.jglr.2015.10.015

Turgeon, P. J., S. L. Van Wilgenburg, and K. L. Drake. 2017. Microphone variability and degradation: implications for monitoring programs employing autonomous recording units. Avian Conservation and Ecology 12(1):9. https://doi.org/10.5751/ACE-00958-120109

Valente, J. J., C. L. LeGrande-Rolls, J. W. Rivers, A. M. Tucker, R. A. Fischer, and M. G. Betts. 2021. Conspecific attraction for conservation and management of terrestrial breeding birds: current knowledge and future research directions. Ornithological Applications 123(2):duab007. https://doi.org/10.1093/ornithapp/duab007

Vors, L. S., J. A. Schaefer, B. A. Pond, A. R. Rodgers, and B. R. Patterson. 2007. Woodland Caribou extirpation and anthropogenic landscape disturbance in Ontario. The Journal of Wildlife Management 71(4):1249-1256. https://doi.org/10.2193/2006-263

Ward, M. P., and S. Schlossberg. 2004. Conspecific attraction and the conservation of territorial songbirds. Conservation Biology 18(2):519-525. https://doi.org/10.1111/j.1523-1739.2004.00494.x

Ward, M. P., B. Semel, C. Jablonski, C. Deutsch, V. Giammaria, S. B. Miller, and B. M. Mcguire. 2011. Consequences of using conspecific attraction in avian conservation: a case study of endangered colonial waterbirds. Waterbirds: The International Journal of Waterbird Biology 34(4):476-480. https://doi.org/10.1675/063.034.0410

Webb, E. B., E. B. Hill, K. M. Malone, and D. C. Mengel. 2022. Secretive marsh bird habitat relationships at mid-continent spring migration stopover sites. The Journal of Wildlife Management 86(4):e22205. https://doi.org/10.1002/jwmg.22205

Wilson E. D., W. E. Jensen, and R. Schultheis. 2018. Habitat use by secretive marsh birds during spring and fall migration in moist-soil wetlands in Kansas. The Wilson Journal of Ornithology 130(2):385-396. https://doi.org/10.1676/16-215.1

Wesner, J. S., E. J. Billman, and M. C. Belk. 2012. Multiple predators indirectly alter community assembly across ecological boundaries. Ecology 93(7):1674-1682. https://doi.org/10.1890/11-2061.1