Springtime bird use of agricultural landscapes examined by using autonomous recording units

. Autonomous recording units (ARUs) emerged as a novel technology for avian acoustic monitoring in the 2000s. They have since been primarily used as a substitute for human observers during the breeding season. Autonomous recording units have potential uses in springtime soundscapes, such as to study spring departure or arrival of migratory birds and territory establishment of resident breeding birds. We described springtime bird communities of agricultural landscapes based on data collected between 1 April and 15 May 2015–2019, from ARUs deployed at 32 locations across 13 counties in Iowa, USA. We compared bird communities across site types, analyzed trends in migrant arrival dates and detection rates, and examined springtime occupancy of five species that commonly use grassland cover: Common Yellowthroat ( Geothlypis trichas ), Field Sparrow ( Spizella pusilla ), Grasshopper Sparrow ( Ammodramus savannarum ), Savannah Sparrow ( Passerculus sandwichensis ), and Vesper Sparrow ( Pooecetes gramineus ). We made 4029 detections of 86 bird species. The most frequent detections were of common farmland species, including Red-winged Blackbird ( Agelaius phoeniceus ), American Robin ( Turdus migratorius ), Brown-headed Cowbird ( Molothrus ater ), Ring-necked Pheasant ( Phasianus colchicus ), and Eastern Meadowlark ( Sturnella magna ); however, detections also included 18 species of greatest conservation need. Sites with reconstructed prairie vegetation, conventionally grown corn ( Zea mays ) and soybean crops ( Glycine max ), crops with prairie strips, and crops with terraces had similar species richness. We found that, in addition to documenting species richness and departure/arrival times of migratory species, ARUs generated species-level detection probabilities similar to or higher than existing studies on breeding season occupancy. Detection probabilities of five focal species ranged from 0.36–0.89. Occupancy models revealed significant springtime land cover associations for Common Yellowthroats, Field Sparrows, Savannah Sparrows


INTRODUCTION
Autonomous recording units (ARUs) emerged as a novel technology for avian acoustic monitoring in the 2000s (Shonfield and Bayne 2017).Programmable recording schedules, affordability, and recent advances in battery life and file storage capacity of ARUs provide several advantages to researchers including broader spatial and temporal sampling, reductions in observer bias, and data quality control (Sugai et al. 2019).Autonomous recording units have been employed in studies examining settlement patterns of migratory birds (e.g., Johnson andBayne 2022, Swicegood et al. 2023), soundscape dynamics (e.g., Suer and Farina 2015, Bennocci et al. 2020, Stowell andSeur 2020), and diel patterns (e.g., Jahn et al. 2017, Pérez-Granados and Schuchmann 2020, Schroeder and McRae 2020).Although they can be deployed long-term, ARUs have been primarily used during the breeding season because birds are more vocally active while breeding and are therefore more identifiable in audio recordings (e.g., Alquezar and Machado 2015, Furnas and Callas 2015, Pérez-Granados et al. 2018).The application of ARUs for studying birds during other periods of the annual cycle, such as arrival to and departure from breeding grounds, remains underdeveloped.
There is potential for use of ARUs in springtime soundscapes, such as during spring departure or arrival of migratory birds, and territory establishment of resident and spring arriving birds.Buxton et al. (2016) used ARUs to examine shifts in arrival of migratory songbirds in Alaska and found bioacoustic indices to be useful for tracking arrival of songbirds.Swicegood et al. (2023) deployed ARUs in agricultural landscapes to investigate settlement patterns of Thick-billed Longspurs (Rhynchophanes mccownii).Other springtime studies include Sanders and Mennill (2014) and Colbert et al. (2015), which used ARUs to examine spring migratory movements and spring gobbling activity of Wild Turkey (Meleagris gallopavo), respectively.Autonomous recording units have also proven effective for studying Ruffed Grouse (Bonasa umbellus) drumming behavior using April recordings in Minnesota (Déaux et al. 2020, Grinde et al. 2021).The use of ARUs to study spring migration and territory establishment, however, is still relatively underexplored.Autonomous recording unit studies could form a "middle ground," augmenting broad-scale taxon-wide data collected through radar surveys with species-level information and individual-scale data tracked through global positioning systems with community information.With climate change, land use change, and other potential disruptors affecting temporal and spatial patterns in biological activity, networks of ARUs could potentially be deployed to track species-level shifts in migration timing or spatial concentration or dispersion.Climate change is already thought to be the primary driver of shifts in spring arrival of birds, which respond to increases in local and regional temperatures during wintering and spring migration (Bradley et al. 1999, Brown et al. 1999, Sparks et al. 2007, Swanson and Palmer 2009, Van Buskirk et al. 2009).
Data from ARU recordings are often analyzed though occupancy modeling (e.g., Furnas andCallas 2015, Stiffler et al. 2018).Similarly, nearly all habitat-association studies that use occupancy modeling take place during the breeding season, with little attention given to non-breeding patterns.A major reason for this is the closure assumption of occupancy modeling, i.e., that there are no changes in availability between survey periods (MacKenzie et al. 2002, Rota et al. 2009).However, Kendall et al. (2013) demonstrated that the closure assumption can be met by staggering arrival and departure times to avoid biasing occupancy estimates.Higher vocal activity near the beginning of the breeding season may also lead to higher detection rates.A study of Canadian forest birds found higher detection rates and higher species richness earlier in the breeding season (Ehnes et al. 2018).Springtime detection rates are largely unexplored in most systems, but vocal activity may be high enough for some species to develop accurate occupancy estimates.
We evaluated the utility of ARUs for studying spring bird communities as part of a larger study seeking to understand grassland bird use of agricultural landscapes of the U.S. Corn Belt.Grassland species are more susceptible to local extinctions due to increasing temperatures compared to forest birds; species occupying fragmented grasslands are at even higher risk (Jarzyna et al. 2016).Since 1970, grassland birds in North America have experienced a 53% overall decline, the steepest reduction of any bird community, primarily due to grassland habitat loss (Rosenberg et al. 2019).In North America, 80% of all grasslands have been lost through conversion to agriculture and only 1% of the original tallgrass prairie remains in most states and provinces (Knopf 1994, Samson andKnopf 1994).Corn Belt agriculture has intensified during the last several decades, creating vegetative monocultures that support fewer native wildlife species (Matson et al. 1997, Brown andSchulte 2011).Increasing agricultural efficiency has removed natural field edges and expanded artificial drainage, use of pesticides, and livestock stocking rates in pastures (Newton 1998, Brown andSchulte 2011).Many grasslandbreeding species now rely on small, semi-natural grassy features embedded within agricultural landscapes.Some features have been found to increase bird use of crop fields including grass terraces (Hultquist and Best 2001), grass waterways (Bryan and Best 1991), field borders (Conover et al. 2009, Burger et al. 2010), and riparian buffers (Berges et al. 2010).These grassy features, usually dominated by exotic cool-season grasses, are unlikely to provide high quality breeding habitat to species that once relied on extensive tracts of tallgrass prairies.Grassy areas with established non-native grasses are structurally different from prairies, with substantially higher grass height and cover.They have been associated with temporal and spatial shifts in breeding patterns of grassland birds (Andersen and Steidl 2020).
Migratory grassland birds are thus likely to be heavily affected by both climate change and habitat modification that accompanies land use change.However, little research has been devoted to their springtime habitat use when energetic resources are crucially needed by birds to support breeding effort and food availability may be low.We sought to fill this gap by using ARUs to investigate the springtime bird community across agricultural landscapes in Iowa, a central U.S. Corn Belt state.Our specific objectives were to: 1. Describe springtime bird communities of agricultural landscapes,

2.
Examine springtime detectability and occupancy, and 3. Evaluate the utility of ARUs for springtime studies of avian habitat use.
We hypothesized that species richness would increase with the amount of grass cover at study sites given the Corn Belt region was historically, prior to EuroAmerican colonization, dominated by grasslands (Conner et al. 2001).We made predictions regarding species occupancy based on published habitat associations (Brennan et al. 2020, Vickery et al. 2020, Wheelwright et al. 2020.).We also hypothesized that springtime species-level detection probabilities would be lower than those generated during breeding season studies.https://journal.afonet.org/vol95/iss1/art1/

Study area
The study area was composed of 32 sites located across 13 counties in Iowa (Appendix 1 Study sites were chosen as a part of a broader investigation of bird use of agricultural landscapes (Stephenson 2022, Giese 2023).Permission to access the land from the land manager was required for study.Sites were composed of one of four types: (1) reconstructed or restored prairie (hereafter, large patch prairies), (2) corn and soybean crops grown using conventional practices for the region and without substantial areas of conservation cover, (3) conventionally managed crops with terraces, and (4) conventionally managed crops with prairie strips.Large patch prairies ranged from 38-102 ha in size, similar to crop fields, which ranged from 17.89 ha to 93.14 ha.Crop fields with terraces included narrow berms installed to minimize soil erosion and were covered in cool-season grasses.Crop fields with prairie strips included linear non-crop areas composed of diverse, reconstructed native prairie vegetation to improve biodiversity and provide multiple ecosystem services (Schulte et al. 2017).Perennial vegetation at sites was mostly dormant during the study period and crops were planted between mid-April and early-May.

Data collection
For each ARU deployment (Songmeter SM3, Wildlife Acoustics, Maynard, Massachusetts, USA), we generated a random point within a farm field and placed the unit in the nearest grassy feature or otherwise unfarmed area.Autonomous recording units were deployed a minimum of 250 m apart.Each unit was mounted ~1.5 m above the ground on a steel fence post.At each site, ARUs were programmed to record for 1 hour beginning 15 minutes before sunrise and ending 45 min after sunrise each day.We analyzed data collected from 1 April-15 May, 2015-2019.We chose this period to coincide with the migratory season for grassland birds, and prior to when in-person observations through bird point counts typically begin in the study region.Data storage cards were collected every 2-3 months and stored for later analysis.Figure 1 summarizes workflow.
We analyzed each 60-min recording of the daily dawn chorus from each deployment location through the period 1 April-15 May for each year, excluding days with excessive wind, rain, or other background noise.Days with background noise deemed excessive showed a steady sound signature above 1 kHz.Of the 2088 total available recordings, 348 were deemed usable with little noise interference.Due to low availability of recordings in some years, 299 of the surveys occurred in 2016 and 2018.We used an intermittent subsampling procedure generated using R statistical All species that could not be initially identified were checked by a secondary observer.We removed 111 unknown vocalizations, mostly distant calls that could not be identified to the species level, from the analysis.Of the 87 species we detected, we classified 44 species as spring arrivers, 34 as year-round residents, and 9 as winterers, according to the Iowa Ornithologists' Union classifications (IOU 2020; Table 1).
Temperature and wind speed are known to affect bird activity and thus detectability (Robbins 1981).We obtained mean daily temperature data from the nearest regional weather monitoring station (NOAA 2022).We were unable to obtain reliable historical wind speed data but accounted for this factor's impact on observation by excluding days with excessive wind in recordings.
To examine environmental predictors of occupancy, we used aerial images provided by the National Agriculture Imagery Program (USDA NAIP 2020) to digitize land cover within 200 m of each ARU.We used field verification to resolve ambiguous land cover in aerial imagery.We calculated the land cover percentage of local environmental variables within 200 m of each ARU (Table 2).We chose 200 m based on the maximum detection distance of around 100 m for most grassland bird species, which was determined by in-person breeding bird surveys at our sites (Giese 2023).We used a five-class cover classification system: crop, grass, prairie, woody, and developed.We calculated the distance from each ARU to the nearest road, a variable that is commonly accepted as influencing the detectability of birds (Yip et al. 2017).

Statistical Analysis
All analyses were performed using package "lme4" (Bates et al. 2015) in R statistical software R 4.1.2(R Core Team 2021).We examined the relationship between site type and species richness using a generalized linear mixed model (glmm) with a Poisson error distribution and log link function.We included deployment location as a random term and accounted for variation in survey effort across deployment locations by including (log-transformed) effort (i.e., number of recordings analyzed) as an offset in the model.Differences in species richness among site types were examined using Tukey post hoc comparisons (Bretz et a. 2016).
Using occupancy models, we evaluated five species with enough variability in occurrence to allow successful model-fitting, including Common Yellowthroat (Geothlypis trichas), Field Sparrow (Spizella pusilla), Grasshopper Sparrow (Ammodramus savannarum), Savannah Sparrow (Passerculus sandwichensis), and Vesper Sparrow (Pooecetes gramineus); hereafter, we refer to them as focal species.We did not model the occurrence of common species, such as Dickcissel (Spiza americana) and Red-winged Blackbird (Agelaius phoenicius), which were present at nearly all sites.Using detection histories of the five focal species, we created single-season occupancy models using the "occu" function in R package "unmarked" (Fiske and Chandler 2011).We considered a site to be each ARU deployment location in each year (hereafter "site-year") and treated each day as a separate site visit.We assumed independence among years.We also assumed independence among sites given that minimum distances between ARUs were greater than the effective detection distances of the focal species.In preparation for occupancy analysis of grassland songbirds, we classified each species as one of the following: yearround resident, winterer, or spring arriving.All focal species were considered spring arrivers.To avoid violating the closure assumption for occupancy modeling (MacKenzie et al. 2002), we did not consider any of these species available for detection until either its first detection at a site or first detection at a nearby site if it was never detected during a specific year (Kendall et al. 2013).
We tested for correlations among spatial variables before constructing models.We used temperature and distance-to-road to model detection of each focal species.Other variables commonly used to model detection, such as observer and time of day, did not vary in our study.After determining the best predictor of detection for each species, we then used spatial covariates to construct occupancy models.Among spatial variables, both prairie cover and developed cover were zero-inflated due to their absence at many ARU deployment sites.To account for this, we converted both variables to a binomial factor based on their presence or absence within 200 m of the ARU.We created interaction models for each species based on known habitat associations and life history traits, predicting that Common We used model averaging and multimodel inference with ΔAIC c < 2.0 across all candidate models to estimate the effects of covariates on occupancy using package "AICcmodavg" (Mazerolle 2020) in R statistical software 4.1.2(R Core Team 2021).After constructing model sets, we tested for collinearity among covariates using variance inflation factor (VIF).We reported parameter estimates derived from AIC model selection with 85% confidence intervals as recommended for limited sets of a priori models (Arnold 2010).We reported 95% confidence intervals for all other parameter estimates.

RESULTS
We Adjusted for survey effort, the relationship between site type and species richness was significant (Table 3), but a Tukey post hoc comparison among site types revealed no statistically significant pairwise differences at the p = 0.05 level.Among comparisons with the largest differences in species richness, conventional crops and crops with terraces was insignificant (β = -0.131,95% CI: -0.634-0.373,p = 0.98); and crops with prairie strips and crops with terraces was insignificant (β = -0.431,95% CI: -0.951-0.083,p = 0.61).Most birds were found in multiple site types, but Darkeyed Junco (Junco hyemalis) and White-crowned Sparrow (Zonotrichia leucophrys) were only detected at a control site with The date of first detection of our five focal species varied (Table 4), and the number of occupied sites increased steadily throughout the study period (Fig. 2).All occupancy models met goodness-of-fit criteria and were unadjusted.Naive detection probabilities for our five focal species ranged from 0.36-0.89(Table 3).The covariates for the top detection probability models for each focal species were temperature for Common Yellowthroat, distance to road for Field Sparrow and Vesper Sparrow, and a constant (i.e., null) for Grasshopper Sparrow and Savannah Sparrow.

DISCUSSION
We examined the utility of ARUs to study springtime avian habitat use within agricultural landscapes.We found that, in addition to documenting species richness of vocalizing species springtime avian communities, ARUs generated species-level detection probabilities similar to or higher than studies on breeding season occupancy of grassland birds (Sidie-Slettedahl We detected 87 total species across our study sites, including 44 species as spring arrivers, 34 as year-round residents, and 9 as winterers, corresponding to 28% of species that regularly occur in the state throughout the year (IOU 2022).Among dates of first detections, we documented an Eastern Kingbird (Tyrannus tyrannus) on 14 April 2016, which is among the three earliest detections of the species in the state (IOU 2020), and a Brown Thrasher (Toxostoma rufum) on 9 April 2016, which is the earliest detection of the species in the state (IOU 2020).These detections occurred during an exceptionally warm period with statewide temperatures 3-5 degrees Celsius above average (NWS 2022).
Our hypothesis of increasing species richness with increasing grassland cover was not statistically supported.Although species richness is a coarse metric of community response, the lack of difference suggests either many bird species did not perceive a difference during the spring season among the site types comprising our study or we were not able to capture differences in their responses based on our methods.Most species of greatest conservation need with more than one detection were documented across all site types with the exception of Greater Yellowlegs and Northern Bobwhite, which were not found in large patch prairie (Table 1).Greater Yellowlegs are migratory during our study period, but Northern Bobwhite breed in Iowa and prefer mosaics of small patches of vegetation including grasslands and early successional vegetation (Brennan et al. 2020).During nonbreeding seasons, Janke and Gates (2013) found that Northern Bobwhites selected early successional woody cover over grassland cover.Our large patch prairie sites contained little woody cover and were surrounded primarily by row crops.
Springtime occupancy varied among the five focal species we studied and are consistent with previous studies on breeding habitat preferences (Fig. 3).Our hypotheses regarding Common Yellowthroat and Field Sparrow occupancy were supported, but  (Guzy and Ritchison 2020), but no study has quantified their springtime habitat preferences.
We predicted that woody cover would be the best predictor of Field Sparrow occupancy.In this analysis, we found woody cover to be the best predictor and have a positive relationship with Field Sparrow occupancy.Developed cover was also an important predictor (Table 6).Field Sparrows prefer fields with a wealth of tree or shrub perches (Carey et al. 2020).During the winter and spring, Field Sparrows used abandoned agricultural fields and forest edges (Allaire and Fisher 1975).The species is more often found in less disturbed edge habitats in the eastern and southeastern U.S. (Marcus et al. 2000, Smith et al. 2005).
Our hypothesis that Grasshopper Sparrow occupancy would show a positive relationship with percent grass cover was not supported due to uninformative models.Grasshopper Sparrows prefer grass-dominated fields and avoid crops (Vickery 2020).
During the winter and spring, the species regularly occupies weedy fields in the southeastern U.S. and co-occurs with Savannah Sparrows and Song Sparrows (Melospiza melodia; Dunning and Pulliam 1989).
We predicted that grass cover would be the best predictor of Savannah Sparrow occupancy.Conversely, we found woody cover to be the best predictor, having a negative relationship with Savannah Sparrow occupancy.Crop cover was also an important predictor (Table 6).Savannah Sparrows prefer open country including grassy meadows, cultivated fields, and lightly grazed pastures and avoid areas with extensive woody cover (Wheelwright and Rising 2020).During the winter and spring, Savannah Sparrows are found in open fields, coastal marshes, and near surface water (Wheelwright and Rising 2020).The pattern we observed is thus unexpected and may be due to the highly anthropogenic character of our study landscapes, in which there is little natural vegetation (Table 1).Crop cover dominates all but the large patch prairie site types, and grassland land cover is dominated by a few species that tend to form monocultures or near-monocultures, such as the non-native cool-season grass smooth brome (Bromus inermis) or reed canary grass (Phalaris arundinacea).
We predicted that crop cover would be the best predictor of Vesper Sparrow occupancy, but we found woody cover to be the best predictor, having a negative relationship (Table 6).Vesper Sparrows breed in dry, open areas with limited woody cover (Jones and Cornely 2020).During winter, Vesper Sparrows are found in grasslands, weedy fields, and savannahs (Howell et al. 1995).In spring, they use pastures and weedy areas near fields and roadsides during migration (Jones and Cornerly 2020).Surprisingly, Vesper Sparrow occupancy was lower in conventional crops than Field Sparrow, Grasshopper Sparrow, and Savannah Sparrow occupancy (Table 5).This may be an artifact of our site selection because sites with conventional crops often contained grass waterways.
Due to the precipitous decline of grassland birds, information on their habitat requirements across the entire annual cycle is vital, especially for migratory species which spend periods of time in geographically distinct areas.In addition to springtime habitat use, ARUs have proven useful for tracking arrival times of migratory birds (Johnson andBayne 2022, Swicegood et al. 2023), which have important consequences for reproduction (Currie et al. 2000, Smith andMoore 2005).Further investigation of nonbreeding habitat use of migratory species will provide insight into potential cross-seasonal effects on populations and allow for effective monitoring (Norris and Marra 2007).
Our work was part of a larger project that also employed traditional methods such as bird point counts (Giese 2023) and nest searches (Stephenson 2022) to study bird use of agricultural landscapes.This is the first study of non-breeding bird use of prairie strips, a conservation practice with multiple ecological benefits (Schulte et al. 2017).Combined with bird point counts conducted at the same locations during the breeding season (Giese 2023), our springtime study of bird use of agricultural landscapes provides information on multi-season habitat associations of grassland birds.
Given the limitations of ARUs, we employed several methods to ensure the robustness of our analyses.We did not start survey periods for occupancy analysis until after a site was known to be occupied by a species because we could only be certain of the dates when birds began vocalizing, and not the true arrival dates of migrant birds.Because acoustic recordings are commonly obstructed by ambient noise, particularly wind (Digby et al. 2013), we removed recordings with excess wind and analyzed selected recordings from days with ideal survey conditions.Being able to collect data over a large number of days, but remove surveys conducted under poor observational conditions was an advantage of long-term deployment of ARUs.However, ambient noise still may have reduced our ability to detect some individuals.
Comparisons between acoustic data and those collected by community scientists would reveal how accurate ARUs are for found seasonal timing measured by ARUs to be consistent to that quantified by eBird across multiple taxonomic levels.In our study, such comparisons would be difficult given the low human population densities in rural Iowa.Thus, ARUs could provide an especially important tool for tracking changes in migration phenology due to climate and land use change in areas with low community science coverage.
We identified three sources of potential bias in our study.First, we were unable to truly randomize deployment locations of ARUs within fields, as it was necessary to avoid cropped areas to minimize conflict with farming operations.Second, removal of recordings with poor observational conditions may have reduced estimates of species richness or affected occupancy modeling results.Last, we were unable to detect non-vocalizing individuals, which may have biased our estimates of richness, especially of wintering species, which may not increase vocalizations until reaching their breeding grounds.
We manually analyzed recordings in this study.Advances in automated species recognition, now commonly used in simpler soundscapes, will likely increase the efficiency of processing large amounts of acoustic data (Priyadarshani et al. 2018).As of now, software used to generate spectrograms and edit sound are largely unable to parse species-level detections from breeding season recordings, which are often generated in complex soundscapes (Potamitis et al. 2014, Ulloa et al. 2016).Methods for automated detection such as machine learning show promise for tracking arrival times of migratory birds and will make large-scale deployments of ARUs feasible for examining phenological shifts due to climate change and range expansions or contractions with land use change (Oliver et al. 2018).

CONCLUSION
Soundscape ecology is a burgeoning field of research (Gasc et al. 2016) enabled by technological improvements in acoustic recorders (Servick 2014).Autonomous recording units (ARU) allow researchers to easily repeat sampling, reduce observer bias and field time, and maintain a permanent record of surveys (Shonfield and Bayne 2017).We provided ecological information on grassland bird use of agricultural landscapes during spring, a period that remains vastly understudied for most birds (Vickery et al. 1999).We conclude that springtime deployment of ARUs can provide worthwhile investigation into spring bird communities and their dynamics.The technology can provide an important tool in monitoring shifts in avian phenology in response to global climate and land use change, phenomena that are already known to affect the spring dynamics of migrant birds.

Author Contributions:
JMM and JCG developed the idea for this study and collaborated with LAS to build research questions and design analyses.JMM and JCG analyzed the data.All authors wrote, edited, and gave final approval for the manuscript.
et al. 2015, West et al. 2016, Rigby and Johnson 2019, Vanausdall  and Dinsmore 2020), though we conducted more surveys.Several focal migratory species showed significant trends in springtime habitat associations.Our study supports expanded use of ARUs and other acoustic devices in the examination of springtime bird communities.

Table 2 .
Summary of mean land cover composition (standard deviation) surrounding autonomous recording unit (ARU) deployment sites in Iowa, USA, and distance from ARU to nearest road.

Table 3 .
Generalized linear mixed model analyzing relationship between site type and springtime bird species richness using autonomous recording unit (ARU) surveys in Iowa, USA, 2015-2019.Large patch prairie was used as reference level for site type variable.
nearby woody cover; Greater Yellowlegs (Tringa melanoleuca), Horned Lark (Eremophila alpestris), and Northern Bobwhite (Colinus virginianus) were detected in every site type but large patch prairie; Swamp Sparrows were only detected in a field with prairie strips in 2018; and Wilson's Snipe (Gallinago delicata) were only detected at sites with terraces.Several species of greatest conservation need were detected during three or fewer surveys.

Table 4 .
Date of first detection based on springtime autonomous recording unit (ARU) surveys in Iowa, USA, 2015-2019 and detection probabilities (and 95% confidence intervals) of five focal bird species.

Table 6 .
Candidate model sets sorted by Akaike's Information Criterion with small sample adjustment (AICc) for five focal species.Based on data from during springtime autonomous recording unit (ARU) surveys in Iowa, USA, 2015-2019.