Research Paper

INTRODUCTION

Declines in migratory aerial insectivorous birds are significant and long-term (Spiller and Dettmers 2019). The declines across many species within the guild of aerial insectivores have been driven by a myriad of stressors including breeding habitat loss, climate change linked weather patterns, and environmental contaminants (Hallmann et al. 2014, Weegman et al. 2017, Imlay et al. 2018, Spiller and Dettmers 2019, Cox et al. 2020). Identifying a single cause of population decline for the entire guild is unlikely because drivers are likely large scale and region or species specific (Smith et al. 2015, Michel et al. 2016).

Among aerial insectivorous birds, nightjars (Family: Caprimulgidae) are often understudied because their nocturnal behavior makes them challenging to detect and is temporally misaligned with traditional dawn bird surveys (Wilson and Watts 2006). Nightjars are most active between dusk and pre-dawn, or when moonlight can provide a backlit sky for them to forage (Cleere 1998). Two nightjars with breeding ranges in Illinois, USA, Eastern Whip-poor-will (Antrostomus vociferus; hereafter: Whip-poor-will) and Chuck-will’s-widow (A. carolinensis), have experienced estimated declines in their breeding range of 53% and 48%, respectively, between 1993 and 2019 according to the Breeding Bird Survey (BBS; Sauer et al. 2020). Given that the BBS is a diurnal survey, it may not be the most accurate population estimate for nocturnal species, but the negative directionality is significant. No specific drivers have been identified as a cause for these declines and knowledge gaps associated with their environmental preferences, such as their relationship to landscape composition and configuration, exist across many areas of their breeding range. Furthermore, widespread species such as these may not select the same environmental conditions throughout their entire range thus requiring region specific research (Sagarin et al. 2006, Pease et al. 2022a, 2022b)

Whip-poor-will and Chuck-will’s-widow are classified as Species of Greatest Conservation Need and the latter is listed as threatened in the state of Illinois (Illinois Department of Natural Resources 2015). Both species are of management interest in the Forest and Woodlands Campaign of the Illinois Wildlife Action Plan in part because of their associations with open woodland conditions, which have steadily declined since the early 20th century (Nowacki and Abrams 2008, Illinois Department of Natural Resources 2015, Habel et al. 2022). Open woodlands are forest communities defined by high canopy openness and poorly developed woody understory, conditions that can be important for both species for its increased light on the forest floor and proximity to gaps used for foraging (Spiller et al. 2022). Interest in maintaining and expanding open woodland conditions has gained support of managers in Illinois and nightjars may benefit as a result (Illinois Department of Natural Resources 2015). There are established surveys targeting nightjars that exist at both the state and national level that follow the road-based survey methods of the Breeding Bird Survey but lack smaller scale spatial and temporal coverage in southern Illinois (Wilson 2008, Beveroth 2017, Pardieck et al. 2019). Research aimed at identifying the distribution or environmental relationships within southern Illinois is likely due to the need for targeted methods to detect Whip-poor-will and Chuck-will’s-widow. The most recent large-scale monitoring effort, focused on censusing all birds in Illinois, was the Illinois Breeding Bird Atlas (IBBA), which occurred between 1986 and 1991 (Kleen et al. 2004). Largely through volunteer effort, surveyors reached nearly 1000 survey blocks in a gridded design for cumulative survey effort of over 32,000 hours across the entire state of Illinois. Over the last century, land cover in southern Illinois has undergone increases in forest cover and decreases in agriculture cover with the government purchase of large portions of land for the establishment of the Shawnee National Forest. Since the initial IBBA, forest cover has increased ~6% (approximately 58,700 hectares) across southern Illinois with noted changes within the forest structure and composition as species shift from oak-dominated to shade tolerant species, a change consistent with forests across the Midwest (Ozier et al. 2006, Ruffner and Groninger 2006, Hanberry and Abrams 2018, Pease et al. 2019). Both the Whip-poor-will and Chuck-will’s-widow were detected in the IBBA, however, the regional changes and length of time that has lapsed since this survey means it likely does not represent their current status in southern Illinois.

We deployed autonomous recording units (ARUs) in a uniform design using the same survey areas as the IBBA in the southern 11 counties of Illinois to detect Whip-poor-will and Chuck-will’s-widow (Fig. 1). Our primary objective was to estimate their current distribution within southern Illinois, along with identifying ecological correlates, and whether these ecological relationships varied by species. Additionally, we were interested in how these distributions may have changed over the last 30 years when compared with the initial IBBA surveys. We investigated species-environmental relationships of Whip-poor-will and Chuck-will’s-widow separately using competing model sets of landscape composition, forest patch configuration, and forest disturbance history to determine what best predicted their occupancy (Fig. 2).

METHODS

Study area and sampling design

We surveyed the southernmost 11 counties of Illinois (Jackson, Williamson, Saline, Gallatin, Union, Johnson, Pope, Hardin, Alexander, Pulaski, and Massac; Fig. 1) during May–July 2022. The 9812 km² area encompasses large tracts of public land managed by both state and federal agencies including the Shawnee National Forest, Crab Orchard National Wildlife Refuge, Cache River State Natural Area, and several Illinois State Parks. We followed a uniform gridded sampling scheme adapted from the IBBA to resample blocks previously established (Kleen et al. 2004). The IBBA survey design was based on the United States Geological Survey 7.5-minute quadrangle maps that were subdivided into six approximately equal portions. They surveyed the third of the six blocks, hereafter called the primary block, at least once during the five years of the IBBA. We further subdivided the IBBA primary blocks into nine equal subblocks that were approximately 2.25 km², reflecting the largest estimates of Whip-poor-will home range to minimize overlapping detections and relate the sampling design to a biologically significant value (Wilson 2003, Hunt 2016). We then randomly selected three of the nine subblocks for a total of 156 sampling sites of which 60 were located on private land. Private landowners were contacted by telephone and mail using publicly available property records. If we were unable to gain permission from private landowners in the randomly sampled subblock, landowners in the neighboring subblock were contacted.

Bird surveys

Whip-poor-will and Chuck-will’s-widow cryptic and nocturnal behavior presents a challenge for traditional bird surveying methods (e.g., point counts). They are well camouflaged, usually occur in isolated populations, and are most likely to be detected during a relatively short time window, just after dusk and before dawn (Cink et al. 2020, Straight and Cooper 2020). To overcome these logistical challenges, we used ARUs, which are programmable devices that record sound at specified times, allowing for greater survey effort with fewer surveyors in the field, and not requiring surveyors to be in the field at night (Shonfield and Bayne 2017). Whip-poor-will and Chuck-will’s-widow primarily vocalize during times when few other birds are calling thereby reducing call overlap and acoustic complexity, further making them good candidates for passive acoustic surveys.

We used Open Acoustics AudioMoth 1.2.0 (Open Acoustic Devices 2020) ARUs because they are cost effective and have comparable performance to other available hardware (Hill et al. 2018, Toenies and Rich 2021). We deployed ARUs for five consecutive evenings programmed to record beginning 30 minutes before sunset for four hours to capture the hours of highest calling activity while balancing battery and data capabilities of the ARUs (Wilson and Watts 2006). We deployed ARUs throughout May and June 2022. Dates and locations of deployments were determined based on landowner availability. We set ARUs to custom mode and programmed them to record continuously through each sampling period at a 48 kHz sampling rate and medium gain (30.6 dB). We housed ARUs in heavy duty, 4 mil plastic bags with silica packets to reduce moisture then placed inside cloth camouflage bags; previous research suggests little to no loss in performance using this waterproofing technique (Lapp et al. 2023). Cloth bags had openings in the corner at the microphone so that sound was not muffled by the cloth. At each sampling location, we affixed ARUs to a tree located at least 15 m from major roadways that was less than 127 mm in diameter at breast height to reduce sound attenuation (Lapp et al. 2023). While placement in the centroid of a subblock is ideal, the location of ARUs were determined by permissions to access granted by private and public landowners. Locations were greater than 200 m apart and no preference was given to any land cover type.

Data extraction and verification

We extracted detections from the acoustic data using a call classification software BirdNet v2 (Kahl et al. 2021), a freely available machine learning model able to identify 984 North American and European bird species. BirdNet is a deep artificial neural network trained to identify birds by spectrogram. Spectrograms are a visual representation of sound based on frequency (Hz) and intensity (dB). BirdNet annotates the spectrogram at 3-second intervals and reports up to three of the most probable species detected during an interval. Following this, BirdNet creates a text document for each 4-hour recording with bird species name, time stamp of vocalization detection event, and confidence level associated with the estimated likelihood of correct detection. We used BirdNet in Python version 3.7.13 (Vanrossum PYTHON) using default settings and refined species list consisting of the 342 birds most likely to be detected in deployment areas based on the time of year and their distributions (Billerman et al. 2022).

Although BirdNet has documented drawbacks, such as variable accuracy of confidence scores between species and across studies, Whip-poor-will and Chuck-will’s-widow typically call repeatedly, hundreds of times in a single night while there is very little activity at similar frequencies (Stoner 1920, Pérez-Granados 2023). The focal species’ calling behavior makes the manual verification of their presence from detections produced by the BirdNet software relatively straightforward. To verify BirdNet detections of each species, we manually verified any night with less than 100 unique BirdNet detections to determine if Whip-poor-will or Chuck-will’s-widow was calling, as fewer than this number of calls does not align with their calling behavior. Preliminary data analysis indicated that there were no nights that had between 100 to 250 BirdNet detections as all nights with detections ranged from 250 to 3500 BirdNet detections, justifying the 100-detection threshold for further analysis. We also manually verified all nights at any site where both Whip-poor-will and Chuck-will’s-widow were detected because both species call at similar frequencies, which may lead to misidentification. Manual verification involved listening to the exact timestamp that BirdNet marked a detection to determine if the target species was calling. To create nightly detection history at each sampling location, we listened to all the BirdNet detections within a single recording sequentially, until we identified a calling individual, resulting in a confirmed detection for that survey night, or if all BirdNet detections in the nightly recording were listened to without confirmation, the species was marked as not detected for that survey. We then used our species-specific detection histories for modeling species occurrence throughout the 11 counties.

Drivers of current distribution

We created three primary model sets describing potential drivers of current distributions: landscape composition, forest patch configuration, and forest disturbance history (Fig. 2). We chose to explore these drivers because they have been deemed important to nightjars in previous research across multiple regions in North America, including the Midwest, southern Canada, and the eastern United States, but were untested in this region (Akresh and King 2016, Slover and Katzner 2016, English et al. 2017, Thompson et al. 2022). The landscape composition model set compares the relative effect of land cover types, forest cover, pastureland, and water sources, on Whip-poor-will and Chuck-will’s-widow distribution. The forest configuration model set compares how the shape, size, and connectedness of forest patches, and the landcover conditions neighboring forest patches influence species’ distribution. Last, we tested a suite of models exploring the driver of forest disturbance on the occurrence of nightjars because Whip-poor-will have been noted to use recently disturbed areas (Akresh and King 2016).

Landscape composition

We calculated landscape composition covariates to quantify the relative amount of landcover types in each subblock using the National Land Cover Database (NLCD) 2019 30x30 m resolution product (Dewitz and U.S. Geological Survey 2021). Using R package “exactextractR” (Baston et al. 2022), we calculated the proportion of pastureland (PAST), water (WAT), conifer cover (CONI), and total forest canopy cover (CANC) in each subblock. We included pastureland because both species require open areas for foraging in addition to forests for nest establishment (Cink et al. 2020, Straight and Cooper 2020). We hypothesized that pastureland cover would be important because it is a common open landcover type in our study region. We chose not to include cropland cover because the configuration across the study area is highly grouped and largely the inverse to forested areas leading to high correlation. We also combined the four forest classifications (mixed, conifer, deciduous, and woody wetland) into a forest/non-forest layer to characterize the percent of total forest canopy cover. Woody wetlands was included in our canopy cover calculation because this NLCD class was included in the 2016 NLCD Canopy Tree Canopy Cover product (Coulston et al. 2012). Furthermore, several sites in our study area along the Mississippi River, such as Oakwood Bottoms in the Shawnee National Forest, were categorized as woody wetland type and have significant areas of canopy cover. Although forest understory structure is likely an important vegetative characteristic, we were unable to find data to summarize this across a large scale (i.e., averaged understory structure of forests within an entire subblock).

Forest patch configuration

We used the binary forest layer described in the landscape composition model set to calculate forest patch metrics within each subblock using R package “landscapemetrics„ (Hesselbarth et al. 2022), a reimplementation of FRAGSTATS (McGarigal and Marks 1995). We chose metrics which reflected the configuration and availability of forest patches on the landscape—total core area (COR) and patch contiguity (CONT)—as described in FRAGSTATS (McGarigal and Marks 1995). Total core area sums the area of all forest cells that exclusively have other forest cells as neighbors thus it describes overall forest amount while accounting for shape because narrow patches will result in lower values than wide patches. Patch contiguity is a metric assessing the spatial connectedness of cells in patches; the larger the value from 0 to 1, the more highly connected forests cells in each subblock are to each other. Finally, we wanted to explore the effect that forest adjacency to selected landcover types had on species’ distribution. Distance between foraging and breeding habitats impacts the behavior and population dynamics of some nightjar species (Evens et al. 2018). To quantify this adjacency, we calculated the proportion of forest edge cells that were neighboring pastureland (PAN), cropland (AGN), and urban (URN) areas within each subblock.

Forest disturbance history

To test the effect of disturbance on the focal species, we calculated the forest-based disturbance history between 2006 and 2022 in the study area using the Google Earth Engine (GEE) implementation of the LandTrendr algorithm (Kennedy et al. 2018). The GEE LandTrendr algorithms use a time series set of Landsat satellite imagery to detect spectral change, or disturbance, in landscapes over time. The calculations produce a raster layer containing the year, magnitude, and duration of the disturbance in a 30x30m pixel. The magnitude value represents the spectral change in a satellite image pre and post a disturbance event and is interpreted as the intensity of a disturbance. We used the Normalized Burn Ratio images between 2006 and 2022 and all default settings within the GEE LandTrendr guided user interface for our calculations. We included the 15-year period to capture potential time-lagged effects following a disturbance (Latif et al. 2020). In a forest-based disturbance, the vegetative structure can change over this 15-year period, particularly with extreme changes like clear-cutting, because revegetation will create shifting conditions that the focal species may be sensitive to (Tozer et al. 2014). The LandTrendr algorithm is not specific to landcover types so changes caused by urban construction registers as a disturbance but will likely not be important to these species. To account for this, we used the Land Change Monitoring System (LCMS) Land Use layer that includes yearly, 30x30 m resolution, remotely sensed, forest cover data to retain changes only in areas that were forested at the time of the disturbance thus creating a forest-based disturbance history layer (Housman et al. 2022). To summarize the forest disturbance history, we calculated the mean magnitude (MAG) and duration (DUR), the proportion covered by disturbance (CF), and the years since the most recent disturbance (YSD) in each subblock.

Modeling approach

Because both species call at frequencies and times that few other organisms vocalize, identifying and extracting detections via machine learning methods is easier because of low sound interference or confusion with other bird calls. This accuracy in nocturnal and low activity soundscapes has been documented in owls, as well as other nightjars (Rognan et al. 2012, Knight et al. 2022). Although false positives were unlikely, false negatives may still be present because of the nature of ARUs, which depend on individuals vocalizing to be detected; it is possible that a species was present at a nightly survey but not detected by the ARU. To account for the imperfect nightly detection, we used single-season, single-species occupancy models to estimate the probability of occurrence (ψ) of a given species while simultaneously estimating the probability of detection (p), which accounts for the false-negative measurement error (MacKenzie et al. 2002). As such, this approach requires fitting two submodels, an observation submodel and an ecological submodel. Although we are most interested in the ecological model, by accounting for the imperfect detection of wildlife, occupancy models can more accurately estimate the effect of ecological covariates resulting in more reliable inference (MacKenzie et al. 2017). As described below, we first fit a suite of models identifying which detection covariates best explain variation in the observation process and following the identification of these covariates, we switch to identifying which ecological variables best explain the distribution of the nightjar species.

To model the probability of detecting a focal species given its occurrence, we tested four covariates thought to influence the detection of the focal species by ARUs. Similar to point counts, precipitation, wind, and vegetation are all documented to affect the detection process of ARUs (Pacifici et al. 2008, Yip et al. 2017, Priyadarshani et al. 2018, Kreh et al. 2023). To account for general weather conditions, we calculated daily precipitation (PRECIP) from Climate Hazards Group InfraRed Precipitation with Station data on GEE at each sampling location for each survey night (Funk et al. 2015). Additionally, extreme temperatures can impact nightjar activity, so we included daily temperature maximums (TEMP) as a factor affecting detection; we did this using observations from seven weather stations in the study area and assigned the temperatures of the closest weather station to each sampling location for each survey night (Knight et al. 2022, Schaaf et al. 2023). Nightjars require some light to see prey and a positive relationship between Whip-poor-will activity and moonlight has been documented (Wilson and Watts 2006). For each survey night, we calculated the moon phase (MP) using R package “suncalc” (Thieurmel and Elmarhraoui 2022). Moon phase signifies the stage of the moon with 0 interpreted as a new moon, 0.5 as a full moon and 0.75 is the last quarter waning moon. We chose moon phase because it encompasses both moon illumination and the possible activity changes based on waning versus waxing moon phases, and added a quadratic effect term to represent a parabolic relationship allowing for the probability of detection to increase or decline (Wilson and Watts 2006). Finally, as deployments occurred throughout the season, we included a parabolic relationship of the day of year (DOY) of each survey to reflect the decline in calling activity associated with end of breeding season (Cink et al. 2020; Table A1.1). We tested combinations of each detection covariate using single-season, single-species occupancy models and used model comparison via Akaike’s Information Criterion correction (AICc) to identify the covariate(s) best describing variation in the observation process (Table A1.2). The best fitting species-specific model was then carried forward to identify which ecological variables best explained the occurrence of each species.

We treated the three potential drivers of current distribution as three candidate model sets to identify which driver best explained occurrence of the nightjars. Importantly, we combined covariates describing each potential driver with the best performing detection model covariates from the previous modelling step, resulting in a combination of covariates across each submodel. Each potential driver model set included a global model that contained all covariates in that model set along with two other models that split the covariates based on predicted relationships (Table A1.2). The model sets were only ranked against models within their own set to test the driver-specific relative importance of covariates.

Prior to model fitting, we calculated correlation coefficients for all pairwise combinations of the predictor variables. If any variables had a correlation coefficient greater than 0.6, they were not included in the same model. We fit all models using single-species, single-season occupancy models in “unmarked” (Fiske and Chandler 2011) in R 4.2.1 (R Core Team 2022) and compared them using AICc (MacKenzie et al. 2003, Burnham and Anderson 2004). We scaled each covariate so that the mean was centered on 0 prior to fitting models and used regression coefficients (β) and 95% confidence intervals to interpret the direction of relationship to covariates. We considered relationships significant if the 95% confidence intervals of regression coefficients did not intersect zero. The slope of the scaled coefficient estimates are interpreted as the expected change in occupancy probability of a single unit of the scaled covariate (Kéry and Royle 2020). We considered models within 2 AICc of the top model to have considerable support (Burnham and Anderson 2004)

Distributional change

To examine change in distribution over 30 years, we used LCMS Land Use layers to identify if the change in landcover explained changes in the distributions of either species during this time. Although sampling technique was different, the hours sampled in our resurvey of the IBBA primary blocks were comparable to the total hours sampled over the 5-year period of the IBBA. We summarized the past and current landscape composition within each block using the 1986 and 2022 LCMS Land Use layers for forest, cropland, developed land, and pastureland with identical resolution. We identified the direction and magnitude of change for each landcover type. We classified the occurrence changes for both species as “lost”> (detected in 1986 but not 2022), “gained” (detected in the 2022 survey but not the 1986 survey), or “not changed,” which included primary blocks with continuous detections and non-detections. We then compared the categorical status changes within each block as a response to trend of a single landcover type using a multinomial logistic regression, an analysis used for unordered categorical response variables, where the response was the status change, separated by species (Kwak and Clayton-Matthews 2002). We ranked models that included each landcover type individually and an intercept-only model via AICc.

RESULTS

Sampling summary

We surveyed 142 sites for a total of 20 hours each (5 nights of 4-hour recordings), resulting in a collective 170,000 minutes of recordings (2840 hours, or 118 days). We were unable to survey at 14 sites because of inability to contact private landowners in a primary block or refusal of permission to survey. We were unable to gain permission in three entire primary blocks (a total of 9 sampling locations), and the remaining five missed sampling locations were split between three primary blocks. We began sampling on 5 May 2022 and continued until 5 July 2022 and collected a total of 710 survey nights.

The number of BirdNet detections in one night ranged from a single vocalization to several thousand; Whip-poor-will and Chuck-will’s-widow had a maximum of 2326 and 3469 calls detected in one night at one location, respectively. Of the 710 survey nights, BirdNet detected Whip-poor-will vocalizations on 78 nights and Chuck-will’s-widow on 98 nights. We manually verified that Whip-poor-will were calling on all 78 nights, but Chuck-will’s-widow were only calling on 75 of the 98 nights, which BirdNet detected. Misidentifications of Chuck-will’s-widows were often the result of overlapping Whip-poor-will vocalizations, and 76% of the inaccurate survey nights were at Whip-poor-will occupied sites. Whip-poor-will were detected at 22 of the 142 subblocks (15%) and Chuck-will’s-widow were detected at 16 of the 142 subblocks (11%). Detections of both species occurred throughout the study area, although some detections were concentrated on the eastern portion of the study area close to large patches of forested, public land.

Whip-poor-will occupancy

The detection of Whip-poor-will was best explained by the global detection model that included daily temperature maximums, precipitation, day of the year, and moon phase (Table 1). In the landscape composition candidate model set, occupancy probability of Whip-poor-will was best explained by the global model but the model containing only conifer and canopy cover was also highly supported (ΔAICc < 2; Table 1). Within the global model, canopy cover had a strong positive effect, indicating that Whip-poor-will were most likely to occupy subblocks with over 70% forest coverage (= 6.49 ± 2.36 [SE]; Fig. 3; Table A2.1). We also documented a strong positive effect of pastureland coverage (= 2.16 ± 0.98; Fig. 3). Occupancy of Whip-poor-will in the forest patch configuration candidate set was best explained by the global model and a model with only total core area and patch contiguity (ΔAICc < 2; Table 1). There was a strong quadratic effect of total core area that peaked at just over 210 hectares of core forest available in a subblock (linear: = 4.49 ± 1.75, quadratic: = -1.57 ± -0.82; Fig. 3). The proportion of forest edges neighboring urban areas and cropland in the global model had slightly negative and positive trends, respectively, but were not significant. The forest disturbance history model, which combined the quadratic effects of magnitude and proportion of disturbance coverage, was the most supported model within the candidate set. The mean magnitude of disturbance had a significant negative effect with pronounced decreases after magnitudes of 250 (= -1.04 ± 0.43; Fig. 3). When comparing the top models from each candidate model set, the landscape composition model had the lowest AIC score. The forest patch configuration model received some support of the top model (ΔAICc < 5; Table 1). The forest disturbance history model received no support (ΔAICc > 20).

Chuck-will’s-widow occupancy

The detection of Chuck-will’s-widow was best explained by the day of the year and moon phase (Table 1). The most supported model across all candidate sets was the intercept-only model which contained no occupancy covariates. Although there was support for models that contained some covariates within each model set (ΔAICc < 5) there were no significant relationships between Chuck-will’s-widow occupancy probability and any covariates (Table 1).

Distributional change

During the IBBA, Whip-poor-will and Chuck-will’s-widow were detected at 24 and 13 primary blocks, respectively, with an average of 44 ± 91 (SD) effort hours per primary block. In 2022, we detected Whip-poor-will within 12 primary blocks and Chuck-will’s-widows in 11 primary blocks. Despite comparable survey effort in terms of duration, there were overall fewer detections across primary blocks for both species when compared to the IBBA. Notably, the number of primary blocks with Whip-poor-will detections dropped from 50% to 25% over the 30-year period. Whip-poor-will were detected in 9 of the same primary blocks across the 1986–1991 and 2022 surveys while Chuck-will’s-widow’s detections occurred in only one of the same primary blocks (Fig. 4). Multinomial regression analysis did not indicate any landcover covariates as key drivers of change between surveys. For Whip-poor-will, models that contained individual covariates associated with the change of landcover of each land type (pastureland, forest, developed, cropland) had little evidence of support (> Δ 3AICc; Table A2.2). Forest cover change for Chuck-will’s-widow showed some support (< Δ 2AICc) but the relationship was not significant. All other models showed less evidence of support (> Δ 3AICc).

DISCUSSION

Understanding the patterns of species distributions is foundational to recognizing threats and gaining knowledge of species ecology for effective management and recovery. Here, we resurveyed a study area after 30 years and estimated the environmental relationships driving the regional distributions of Whip-poor-will and Chuck-will’s-widow. Whip-poor-will were positively related to landscape characteristics consistent with findings in other regions, but we documented no significant relationships in Chuck-will’s-widow (Vala et al. 2020, Thompson et al. 2022). However, we observed differences in Chuck-will’s-widow distribution over time, and when compared to Whip-poor-will. These differences, likely due to unique ecological attributes (e.g., habitat preference), suggest that the two species may respond differently to conservation efforts, particularly at a finer scale. Future research should focus on understanding fine-scale ecology of Chuck-will’s-widow, especially in areas where they co-occur with Whip-poor-will. Research in areas where they co-occur should also emphasize how the species’ occurrence patterns differ in areas of with and without Whip-poor-will.

Whip-poor-will in southern Illinois seem to follow similar broad relationships with landcover as they do in other regions of their breeding range, where the species is associated with high canopy cover and open areas (Akresh and King 2016, Spiller et al. 2022, Thompson et al. 2022). Much of the research on environmental relationships of Whip-poor-will has been in the Canadian boreal forest, and there are noted associations with wetland and marshland that may be analogous to the role that pasturelands play in southern Illinois, an open area to forage for insects and acquire mates (Vala et al. 2020, Billerman et al. 2022). The positive relationship with pastureland observed may also reflect the potential that pastureland, more so than cropland, has to support insect abundance, which is an important aspect of determining use of open landcover types (Møller et al. 2021). To add to this, a study in central Illinois observed a positive relationship between moth abundance and availability of forest core in the area (Souza-Cole et al. 2022). The combination of positive relationships to pastureland and forested areas suggests the juxtaposition of cover (i.e., forests) and open areas for foraging (i.e., pastureland) is an important landscape feature for the species.

Our results suggest that Whip-poor-will also prefer areas with low intensity disturbance, which has been reported in harvested forests in Canada and shrublands in the eastern United States, and more recently led to descriptions of Whip-poor-will as a disturbance-dependent species (Spiller and King 2021, Spiller et al. 2022). In the context of our study area, the disturbances varied from 100 to 1200 intensity, which measures the spectral change in satellite imagery. Although precise descriptions of intensity values vary, generally, disturbances like selective forest harvest registered as having magnitudes between 250 to 500 intensity while clear-cutting for development or pit-mining operations ranged from 900 to 1200 intensity (Fig. A3.1). Much of the less intense (< 250) forest disturbance in the study area was associated with the creation of edge, which would support hypotheses that they prefer such conditions while breeding. However, the vegetative conditions created by disturbance varies in suitability for Whip-poor-will nesting, which, even when occurring in open areas, requires some vegetative coverage (Akresh and King 2016). Furthermore, the growth of understory structure is impacted by more than the intensity of the disturbance, including deer density, invasive species, and management regimes (Rogers et al. 2008, Lettow et al. 2014, Russell et al. 2017). We lacked information on understory structure at the scale of our study, but it is likely an important aspect in their use of recently disturbed areas. Furthermore, reported variable relationships with stand age suggest Whip-poor-will are sensitive to fine-scale differences in disturbed forests that should be studied further (Tozer et al. 2014). We also expected larger negative impacts of cropland and urbanization, as found in studies in Canada and central Illinois, but this may be a result of low amounts of urbanization relative to the large patches of forest in the study area (Vala et al. 2020, Souza-Cole et al. 2022).

There were no significant predictors of Chuck-will’s-widows distribution, but trends suggested positive relationships to forest patch core and conifer cover. Basic breeding habitat requirements and foraging behaviors of Chuck-will’s-widows and Whip-poor-wills are alike, so the similarities of some trends, namely open areas to forage and forested land, were expected. Although their distribution in our study differed from Whip-poor-will, it is unclear how Chuck-will’s-widows’ preferred habitats vary, whether they may be in competition with Whip-poor-will or if they may be a generalist species capable of using a variety of conditions. Despite our inability to identify drivers of Chuck-will’s-widow distributions, the contrast in distributions when compared to Whip-poor-will is clear and the dynamics of these species should be further studied in regions where they co-occur.

Between 1986 and 2022, there was an overall gain in forest cover as well as a loss in cropland and pastureland throughout the study blocks. Despite the landcover changes, we did not observe a relationship between changes in landcover and either species distribution suggesting that landcover change may not be a prominent driver of their distributional change in southern Illinois. However, the distributional changes in both species over time are evident. When comparing to large scale surveys like the Breeding Bird Survey, where trends of decline are similar between species, the observed occupancy decline of Whip-poor-will detections in comparison to Chuck-will’s-widow are particularly notable (Sauer et al. 2020). Whip-poor-will detections trended negatively but they persisted in the same primary blocks over time while Chuck-will’s-widow detections shifted to new primary blocks. Although our inference is limited, the clear variation in patterns of detections over time requires further investigation to identify the variation between species. For example, though not available for use in this analysis, it is likely that changes in prey availability may better describe changes in their distribution through time (Spiller and Dettmers 2019, Raven and Wagner 2021). Furthermore, these pattern differences provide evidence that Whip-poor-will and Chuck-will’s-widow have variable responses to their environment and thus may require some species-specific management practices.

Finally, while passive acoustic monitoring is an evolving tool for bird surveying and has been effective for other nocturnal species over a large areas, it has not yet been applied to nightjars at this spatial scale (Wood et al. 2019, Duchac et al. 2020). In our study, the combination of ARUs and BirdNet provided rapid assessment of large amounts of data with high accuracy. The repetitive nature of their calling behavior makes manual verification for detection/non-detection data straight forward. Our results demonstrate that ARUs are a viable substitute for traditional nightjar monitoring methods and can be an easy addition to dawn-focused passive acoustic monitoring methods especially with the availability of open-source call identification software like BirdNet.

Our results may be limited by the homogeneous nature of sites where individuals were detected relative to sites where they were not. To capture broad trends of distributions throughout a large study area by revisiting previously surveyed sites, we chose a systematic sampling scheme rather than targeting specific landcover types; this choice likely came at the cost of a more refined understanding of environmental relationships. That is, the initial Illinois Breeding Bird Atlas was a systematic sampling grid aimed at detecting all breeding birds within the state, and here we were focused primarily on two species. Additionally, the relationships investigated here are closely linked to the availability of foraging habitat, but we did not gather insect abundance information that would be valuable to these models.

This study supports previous research that management actions for these species should include the maintenance of core forest area and adjacent open areas. Results also highlight management that encourages some disturbance and intermediate successional habitat would benefit these birds in southern Illinois. Although not explicitly addressed here, open areas with vegetation communities that are able to support high insect abundance may be particularly important to preserve. We did not specifically examine the relationships of these birds to open woodlands, but the affinity for low magnitude disturbances suggests support for management strategies that create open canopied areas. However, the affinity for high amounts of forest patch core area suggests that Whip-poor-will likely require a variation of forest canopies, both open and closed, across the landscape. Maintaining a matrix of open areas and mix-canopied forests with some disturbance is therefore likely beneficial to Whip-poor-will in southern Illinois. Although Chuck-will’s-widow relationships were not significant, some general trends followed those of the Whip-poor-will. However, the observed distributional changes between the IBBA and this survey provide support that Chuck-will’s-widow land use varies to some extent from Whip-poor-will. These observed differences in Chuck-will’s-widow environmental relationships warrant further investigation explicitly comparing these two species where they co-occur.

LITERATURE CITED

Akresh, M. E., and D. I. King. 2016. Eastern Whip‐poor‐will breeding ecology in relation to habitat management in a pitch pine-scrub oak barren. Wildlife Society Bulletin 40:97-105. https://doi.org/10.1002/wsb.621

Baston, D., ISciences, and LLC. 2022. exactextractr: Fast Extraction from Raster Datasets using Polygons.

Beveroth, T. A. 2017. Monitoring of owls and nightjars, MOON, in Illinois: 2015-2016 Report. Illinois Natural History Survey, Prairie Research Institute, University of Illinois, Champaign, Illinois, USA. http://hdl.handle.net/2142/106760

Billerman, S. M., B. K. Keeney, P. G. Rodewald, and T. S. Schulenberg, editors. 2022. Birds of the world. Cornell Lab of Ornithology, Ithica, New York, USA. https://doi.org/10.2173/bow

Burnham, K. P., and D. R. Anderson. 2004. Multimodel inference: understanding AIC and BIC in model selection. Sociological Methods & Research 33:261-304. https://doi.org/10.1177/0049124104268644

Cink, C. L., P. Pyle, and M. A. Patten. 2020. Eastern Whip-poor-will (Antrostomus vociferus), version 1.0. In P. G. Rodewald, editor. Birds of the world. Cornell Lab of Ornithology, Ithaca, New York, USA. https://doi.org/10.2173/bow.whip-p1.01

Cleere, N. 1998. Nightjars: a guide to the nightjars, nighthawks, and their relatives. Yale University Press, New Haven, Connecticut, USA.

Coulston, J. W., G. G. Moisen, B. T. Wilson, M. V. Finco, W. B. Cohen, and C. K. Brewer. 2012. Modeling percent tree canopy cover: a pilot study. Photogrammetric Engineering & Remote Sensing 78:715-727.

Cox, A. R., R. J. Robertson, W. B. Rendell, and F. Bonier. 2020. Population decline in tree swallows (Tachycineta bicolor) linked to climate change and inclement weather on the breeding ground. Oecologia 192:713-722. https://doi.org/10.1007/s00442-020-04618-8

Dewitz, J., and U.S. Geological Survey. 2021. National land cover database (NLCD) 2019 Products. U.S. Geological Survey, Sioux Falls, South Dakota, USA. https://doi.org/10.5066/P9KZCM54

Duchac, L. S., D. B. Lesmeister, K. M. Dugger, Z. J. Ruff, and R. J. Davis. 2020. Passive acoustic monitoring effectively detects Northern Spotted Owls and Barred Owls over a range of forest conditions. Condor 122:duaa017. https://doi.org/10.1093/condor/duaa017

English, P. A., J. J. Nocera, B. A. Pond, and D. J. Green. 2017. Habitat and food supply across multiple spatial scales influence the distribution and abundance of a nocturnal aerial insectivore. Landscape Ecology 32:343-359. https://doi.org/10.1007/s10980-016-0454-y

Evens, R., N. Beenaerts, T. Neyens, N. Witters, K. Smeets, and T. Artois. 2018. Proximity of breeding and foraging areas affects foraging effort of a crepuscular, insectivorous bird. Scientific Reports 8:3008. https://doi.org/10.1038/s41598-018-21321-0

Fiske, I., and R. Chandler. 2011. unmarked: An R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43:1-23. https://doi.org/10.18637/jss.v043.i10

Funk, C., P. Peterson, M. Landsfeld, D. Pedreros, J. Verdin, S. Shukla, G. Husak, J. Rowland, L. Harrison, A. Hoell, and J. Michaelsen. 2015. The climate hazards infrared precipitation with stations—a new environmental record for monitoring extremes. Scientific Data 2:150066. https://doi.org/10.1038/sdata.2015.66

Habel, J. C., M. Teucher, P. Gros, V. Gfrerer, and J. Eberle. 2022. The importance of dynamic open-canopy woodlands for the conservation of a specialist butterfly species. Landscape Ecology 37:2121-2129. https://doi.org/10.1007/s10980-022-01472-2

Hallmann, C. A., R. P. Foppen, C. A. van Turnhout, H. de Kroon, and E. Jongejans. 2014. Declines in insectivorous birds are associated with high neonicotinoid concentrations. Nature 511:341-343. https://doi.org/10.1038/nature13531

Hanberry, B. B., and M. D. Abrams. 2018. Recognizing loss of open forest ecosystems by tree densification and land use intensification in the Midwestern USA. Regional Environmental Change 18:1731-1740. https://doi.org/10.1007/s10113-018-1299-5

Hesselbarth, M. H. K., M. Sciaini, J. Nowosad, S. Hanss, L. J. Graham, J. Hollister, K. With, and F. Privé. 2022. landscapemetrics: Landscape metrics for categorical map patterns.

Hill, A. P., P. Prince, E. Piña Covarrubias, C. P. Doncaster, J. L. Snaddon, and A. Rogers. 2018. AudioMoth: evaluation of a smart open acoustic device for monitoring biodiversity and the environment. Methods in Ecology and Evolution 9:1199-1211. https://doi.org/10.1111/2041-210X.12955

Housman, I. W., L. S. Campbell, J. P. Heyer, W. E. Goetz, M. V. Finco, N. Pugh and K. Megown. 2022. U.S. Forest Service landscape change monitoring system methods Version 2021.7. GTAC-10252-RPT3. U.S. Forest Service, Geospatial Technology and Applications Center, Salt Lake City, Utah, USA. https://data.fs.usda.gov/geodata/rastergateway/LCMS/LCMS_v2021-7_Methods.pdf

Hunt, P. D. 2016. Extra-territorial movements by eastern Whip-poor-wills. North American Bird Bander 41(3):97-102.

Illinois Department of Natural Resources. 2015. Illinois wildlife action plan. Illinois Department of Natural Resources, Springfield, Illinois, USA.

Imlay, T. L., J. Mills Flemming, S. Saldanha, N. T. Wheelwright, and M. L. Leonard. 2018. Breeding phenology and performance for four swallows over 57 years: relationships with temperature and precipitation. Ecosphere 9:e02166. https://doi.org/10.1002/ecs2.2166

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

Kennedy, R. E., Z. Yang, N. Gorelick, J. Braaten, L. Cavalcante, W. B. Cohen, and S. Healey. 2018. Implementation of the LandTrendr algorithm on Google Earth Engine. Remote Sensing 10(5):691. https://doi.org/10.3390/rs10050691

Kéry, M., and J. A. Royle. 2020. Applied hierarchical modeling in ecology: analysis of distribution, abundance and species richness in R and BUGS: Volume 1. Academic, London, UK.

Kleen, V. M., L. Cordle, and R. A. Montgomery. 2004. The Illinois breeding bird atlas. Illinois Natural History Survey, Champaign, Illinois, USA.

Knight, E. C., K. C. Hannah, and J. DeMoor. 2022. In the still of the night: revisiting Eastern Whip-poor-will surveys with passive acoustic monitoring. Avian Conservation and Ecology 17(1):21. https://doi.org/10.5751/ACE-02080-170121

Kreh, C. D., B. S. Pease, and K. Pacifici. 2023. Efficacy of autonomous recording units to evaluate wild turkey gobbling chronology in North Carolina, USA. Wildlife Society Bulletin 47:e1433. https://doi.org/10.1002/wsb.1433

Kwak, C., and A. Clayton-Matthews. 2002. Multinomial logistic regression. Nursing Research 51(6):404-410. https://doi.org/10.1097/00006199-200211000-00009

Lapp, S., N. Stahlman, and J. Kitzes. 2023. A quantitative evaluation of the performance of the low-cost AudioMoth acoustic recording unit. Sensors 23:5254. https://doi.org/10.3390/s23115254

Latif, Q. S., J. S. Ivan, A. E. Seglund, D. L. Pavlacky, and R. L. Truex. 2020. Avian relationships with bark beetle outbreaks and underlying mechanisms in lodgepole pine and spruce-fir forests of Colorado. Forest Ecology and Management 464:118043. https://doi.org/10.1016/j.foreco.2020.118043

Lettow, M. C., L. A. Brudvig, C. A. Bahlai, and D. A. Landis. 2014. Oak savanna management strategies and their differential effects on vegetative structure, understory light, and flowering forbs. Forest Ecology and Management 329:89-98. https://doi.org/10.1016/j.foreco.2014.06.019

MacKenzie, D. I., J. D. Nichols, J. E. Hines, M. G. Knutson, and A. B. Franklin. 2003. Estimating site occupancy, colonization, and local extinction when a species Is detected imperfectly. Ecology 84:2200-2207. https://doi.org/10.1890/02-3090

MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248-2255. https://doi.org/10.1890/0012-9658(2002)083[2248:ESORWD]2.0.CO;2

MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. Bailey, and J. E. Hines. 2017. Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence. Elsevier, Amsterdam, The Netherlands.

McGarigal, K., and B. J. Marks. 1995. FRAGSTATS: spatial pattern analysis program for quantifying landscape structure. General Technical Report PNW-GTR-351. U.S. Forest Service, Pacific Northwest Research Station, Portland, Oregon, USA. https://doi.org/10.2737/PNW-GTR-351

Michel, N. L., A. C. Smith, R. G. Clark, C. A. Morrissey, and K. A. Hobson. 2016. Differences in spatial synchrony and interspecific concordance inform guild-level population trends for aerial insectivorous birds. Ecography 39:774-786. https://doi.org/10.1111/ecog.01798

Møller, A. P., D. Czeszczewik, E. Flensted-Jensen, J. Erritzøe, I. Krams, K. Laursen, W. Liang, and W. Walankiewicz. 2021. Abundance of insects and aerial insectivorous birds in relation to pesticide and fertilizer use. Avian Research 12:43. https://doi.org/10.1186/s40657-021-00278-1

Nowacki, G. J., and M. D. Abrams. 2008. The demise of fire and “Mesophication” of forests in the Eastern United States. BioScience 58:123-138. https://doi.org/10.1641/B580207

Ozier, T. B., J. W. Groninger, and C. M. Ruffner. 2006. Community composition and structural changes in a managed Illinois Ozark Hills Forest. American Midland Naturalist 155:253-269. https://doi.org/10.1674/0003-0031(2006)155[253:CCASCI]2.0.CO;2

Pacifici, K., T. R. Simons, and K. H. Pollock. 2008. Effects of vegetation and background noise on the detection process in auditory avian point-count surveys. Auk 125:600-607. https://doi.org/10.1525/auk.2008.07078

Pardieck, K., D. Ziolkowski Jr, M. Lutmerding, V. Aponte, and M. Hudson. 2019. North American breeding bird survey dataset 1966-2018. Patuxent Wildlife Research Center, Laurel, Maryland, USA.

Pease, B. S., E. J. Holzmueller, and C. K. Nielsen. 2019. Influence of forest structure and composition on summer habitat use of wildlife in an upland hardwood forest. Diversity 11:160. https://doi.org/10.3390/d11090160

Pease, B. S., K. Pacifici, and R. Kays. 2022a. Exploring spatial nonstationarity for four mammal species reveals regional variation in environmental relationships. Ecosphere 13:e4166. https://doi.org/10.1002/ecs2.4166

Pease, B. S., K. Pacifici, R. Kays, and B. Reich. 2022b. What drives spatially varying ecological relationships in a wide-ranging species? Diversity and Distributions 28:1752-1768. https://doi.org/10.1111/ddi.13594

Pérez-Granados, C. 2023. BirdNET: applications, performance, pitfalls and future opportunities. Ibis 165:1068-1075. https://doi.org/10.1111/ibi.13193

Priyadarshani, N., I. Castro, and S. Marsland. 2018. The impact of environmental factors in birdsong acquisition using automated recorders. Ecology and Evolution 8:5016-5033. https://doi.org/10.1002/ece3.3889

R Core Team. 2022. R: a language and environment for statistical analysis. R Foundations for Statistical Computing, Vienna, Austria.

Raven, P. H., and D. L. Wagner. 2021. Agricultural intensification and climate change are rapidly decreasing insect biodiversity. Proceedings of the National Academy of Sciences 118:e2002548117. https://doi.org/10.1073/pnas.2002548117

Rogers, D. A., T. P. Rooney, D. Olson, and D. M. Waller. 2008. Shifts in southern Wisconsin forest canopy and understory richness, composition, and heterogeneity. Ecology 89:2482-2492. https://doi.org/10.1890/07-1129.1

Rognan, C. B., J. M. Szewczak, and M. L. Morrison. 2012. Autonomous recording of Great Gray Owls in the Sierra Nevada. Northwestern Naturalist 93:138-144. https://doi.org/10.1898/nwn11-02.1

Ruffner, C. M., and J. W. Groninger. 2006. Making the case for fire in southern Illinois forests. Journal of Forestry 104:78-83. https://doi.org/10.1093/jof/104.2.78

Russell, M. B., C. W. Woodall, K. M. Potter, B. F. Walters, G. M. Domke, and C. M. Oswalt. 2017. Interactions between white-tailed deer density and the composition of forest understories in the northern United States. Forest Ecology and Management 384:26-33. https://doi.org/10.1016/j.foreco.2016.10.038

Sagarin, R. D., S. D. Gaines, and B. Gaylord. 2006. Moving beyond assumptions to understand abundance distributions across the ranges of species. Trends in Ecology & Evolution 21:524-530. https://doi.org/10.1016/j.tree.2006.06.008

Sauer, J. R., W. A. Link, and J. E. Hines. 2020. The North American Breeding Bird Survey, Analysis Results 1966 - 2019. U.S. Geological Survey, Laurel, Maryland, USA. https://doi.org/10.5066/P96A7675

Schaaf, A. A., M. Boullhesen, and M. S. Akmentins. 2023. Vocal activity of the Rufous Nightjar (Antrostomus rufus) in subtropical Yungas Andean forests of Argentina. Neotropical Biodiversity 9:59-63. https://doi.org/10.1080/23766808.2023.2185963

Shonfield, J., and E. M. Bayne. 2017. Autonomous recording units in avian ecological research: current use and future applications. Avian Conservation and Ecology 12(1):14. https://doi.org/10.5751/ACE-00974-120114

Slover, C. L., and T. E. Katzner. 2016. Eastern Whip-poor-wills (Antrostomus vociferus) are positively associated with low elevation forest in the Central Appalachians. Wilson Journal of Ornithology 128:846-856. https://doi.org/10.1676/15-156.1

Smith, A. C., M.-A. R. Hudson, C. M. Downes, and C. M. Francis. 2015. Change points in the population trends of aerial-insectivorous birds in North America: synchronized in time across species and regions. PLoS ONE 10:e0130768. https://doi.org/10.1371/journal.pone.0130768

Souza-Cole, I., M. P. Ward, R. L. Mau, J. T. Foster, and T. J. Benson. 2022. Eastern Whip-poor-will abundance declines with urban land cover and increases with moth abundance in the American Midwest. Ornithological Applications 124:duac032. https://doi.org/10.1093/ornithapp/duac032

Spiller, K. J., and R. Dettmers. 2019. Evidence for multiple drivers of aerial insectivore declines in North America. Condor 121:duz010. https://doi.org/10.1093/condor/duz010

Spiller, K. J., and D. I. King. 2021. Breeding habitat associations of Eastern Whip-poor-wills in managed forests. Journal of Wildlife Management 85:1009-1016. https://doi.org/10.1002/jwmg.22045

Spiller, K., D. King, and J. Bolsinger. 2022. Foraging and roosting habitat of Eastern Whip-poor-wills in the northeastern United States. Journal of Field Ornithology 93(1):6. https://doi.org/10.5751/JFO-00057-930106

Stoner, D. 1920. Whippoorwill calls. Wilson Bulletin 32:87-93.

Straight, C. A., and R. J. Cooper. 2020. Chuck-will’s-widow (Antrostomus carolinensis), version 1.0. in A. F. Poole, editor. Cornell Lab of Ornithology, Ithaca, New York, USA. https://doi.org/10.2173/bow.chwwid.01

Thieurmel, B., and A. Elmarhraoui. 2022. suncalc: Compute Sun Position, Sunlight Phases, Moon Position and Lunar Phase. https://doi.org/10.32614/CRAN.package.suncalc

Thompson III, F. R., M. C. Roach, and T. W. Bonnot. 2022. Woodland restoration and forest structure affect nightjar abundance in the Ozark Highlands. Journal of Wildlife Management 86(2):e22170. https://doi.org/10.1002/jwmg.22170

Toenies, M., and L. N. Rich. 2021. Advancing bird survey efforts through novel recorder technology and automated species identification. California Fish and Wildlife 107:56-70. https://doi.org/10.51492/cfwj.107.5

Tozer, D. C., J. C. Hoare, J. E. Inglis, J. Yaraskavitch, H. Kitching, and S. Dobbyn. 2014. Clearcut with seed trees in red pine forests associated with increased occupancy by Eastern Whip-poor-wills. Forest Ecology and Management 330:1-7. https://doi.org/10.1016/j.foreco.2014.06.038

Vala, M. A., G. W. Mitchell, K. C. Hannah, J. E. Put, and S. Wilson. 2020. The effects of landscape composition and configuration on Eastern Whip-poor-will (Caprimulgus vociferous) and Common Nighthawk (Chordeiles minor) occupancy in an agroecosystem. Avian Conservation and Ecology 15(1):24. https://doi.org/10.5751/ACE-01613-150124

Weegman, M. D., T. W. Arnold, R. D. Dawson, D. W. Winkler, and R. G. Clark. 2017. Integrated population models reveal local weather conditions are the key drivers of population dynamics in an aerial insectivore. Oecologia 185:119-130. https://doi.org/10.1007/s00442-017-3890-8

Wilson, M. D. 2003. Distribution, abundance, and home range of the Whip-poor-will (Caprimulgus vociferus) in a managed forest landscape. College of William and Mary, Williamsburg, Virginia, USA.

Wilson, M. D. 2008. The nightjar survey network: program construction and 2007 Southeastern nightjar survey results. CCBTR-08-01. Center for Conservation Biology Technical Report Series. College of William and Mary, Williamsburg, Virginia, USA.

Wilson, M. D., and B. D. Watts. 2006. Effect of moonlight on detection of Whip-poor-wills: implications for long-term monitoring strategies. Journal of Field Ornithology 77:207-211. https://doi.org/10.1111/j.1557-9263.2006.00042.x

Wood, C. M., R. J. Gutiérrez, and M. Z. Peery. 2019. Acoustic monitoring reveals a diverse forest owl community, illustrating its potential for basic and applied ecology. Ecology 100:e02764. https://doi.org/10.1002/ecy.2764

Yip, D. A., E. M. Bayne, P. Sólymos, J. Campbell, and D. Proppe. 2017. Sound attenuation in forest and roadside environments: implications for avian point-count surveys. Condor 119:73-84. https://doi.org/10.1650/CONDOR-16-93.1