INTRODUCTION

Many bird survey protocols involve counting birds through passive acoustic listening. Generally, the timing of survey visits should coincide with the peak of the breeding season when detectability is highest and identifications are most reliable (Wilson and Bart 1995, Amrhein et al. 2002, Wilson and Watts 2006). Peaks in rates of male singing are often linked to key stages in the breeding cycle, such as when a male is attempting to attract a female to his territory or when stimulating a female to breed (Catchpole 1973, Nowicki and Searcy 2004, Upham-Mills et al. 2020). In most passerines, daily vocal activity peaks at or around sunrise and declines throughout the morning (Krebs and Kacelnik 1983, Staicer et al. 1996). As a result, many avian monitoring protocols coincide with this peak in passerine activity, starting just prior to sunrise and continuing for several hours while vocal activity for many species remains high (e.g., North American Breeding Bird Survey (BBS); Sauer and Link 2011).

Some crepuscular or nocturnal bird species are more active outside of typical diurnal survey periods, posing a challenge for general large-scale monitoring programs (Wilson and Watts 2006, Conway and Gibbs 2011, Martin et al. 2014). Given this mismatch, targeted surveys addressing species-specific detection covariates have been developed for certain marsh birds (Conway and Gibbs 2011, Martin et al. 2014) and for owls and nightjars (Lima et al. 2020, Zuberogoitia et al. 2020, Knight et al. 2021a). Although desirable, dedicated surveys for rare or elusive species can be challenging and costly to develop, especially across large spatial and temporal scales (Thompson 2013, Zuberogoitia et al. 2020). The recent proliferation of autonomous recording units (ARUs) offers some potential advantages and efficiencies for testing species-specific detection covariates (Digby et al. 2013, Shonfield and Bayne 2017, Gibb et al. 2019). Further, automated recognition approaches have increased the efficiency of processing time and improved data precision of large acoustic datasets (Priyadarshani et al. 2018, Stowell et al. 2019, Knight et al. 2020). Such data can inform refinements to surveys for many rare and elusive species, which is important for the ongoing success of monitoring programs and management efforts (e.g., Schroeder and McRae 2020).

The Common Nighthawk (Chordeiles minor) is a migratory aerial insectivore with an extensive breeding range in North America (Brigham et al. 2011; Fig. 1). Considered largely crepuscular, it is most active before sunrise and after sunset (Brigham and Fenton 1991, Brigham et al. 2011). Typical breeding habitat consist of areas of extensive open ground, including gravel beaches, exposed rocky outcrops, early post-fire forest, and other disturbed sites (Brigham et al. 2011, Farrell et al. 2017, Vala et al. 2020), with wetland areas being important for foraging (Knight et al. 2021b). The combination of highly cryptic plumage and crepuscular activity makes it a difficult species to detect visually, so most encounters with this species represent audible detections. In 2007, the Committee on the Status of Endangered Wildlife in Canada (COSEWIC) assessed the species as Threatened, following documented population declines of 68% in southern Canada since 1970 (Brigham et al. 2011). In 2018, the species’ status was reassigned to Special Concern, which is a lower risk category, based on evidence of recent slowing in the rate of population decline and apparent availability of ample suitable habitat in the boreal forest (COSEWIC 2018). Given the likely continuation of perceived conservation risk and that a large proportion of the species’ breeding range is remote and poorly surveyed, improving survey designs to maximize detectability and efficiency is desirable.

Despite increased recent monitoring attention, survey protocol recommendations for this species are variable across jurisdictions, with few consistent guidelines. Current understanding of how activity patterns in this species vary across space and time, based on passive surveys, is limited and often anecdotal. In Saskatchewan, Common Nighthawk surveys suggested that daily activity peaked 30–45 min after sunset and peaked seasonally between 23–27 June (Wedgwood 1973, Wedgwood 1991). In British Columbia, protocols suggest Common Nighthawk surveys should begin at sunset, continue until the end of the dusk crepuscular period, and be limited to local civil twilight periods from mid– to late–June (Resources Inventory Committee 1998). The recently developed Canadian Nightjar Survey protocol recommends surveying from 30 min prior to sunset for approximately two h, between mid–June and mid–July (Knight et al. 2019), though this survey includes other, more nocturnal nightjar species. Generally, little effort has gone into identifying species-specific detection covariates that could inform or improve large-scale surveys in this wide-ranging species.

Sampling effort recommendations for the Common Nighthawk are also variable. The British Columbia protocol recommends several five-minute surveys (Resources Inventory Committee 1998), and the Wisconsin protocol recommends three 10-minute surveys (Viel et al. 2020), whereas the Canadian Nightjar Survey and Nightjar Survey Network require one six-minute survey anytime during the survey period (Knight et al. 2019; Nightjar Survey Network, http://www.nightjars.org/). Selecting the appropriate protocol for population monitoring or environmental assessment is hindered by the fact that the effect of survey duration and number of surveys on Common Nighthawk detection probability is unknown.

The goal of our study was to use a temporally intensive and spatially extensive acoustic dataset of vocal (peents) and non-vocal (booms) acoustic signals to inform species-specific surveys for the Common Nighthawk. Our first objective was to model the effects of season, time of day, and varying latitude on vocal and non-vocal activity to determine the best timing for surveys. Our second objective was to determine whether the twilight period was effective for constraining survey timing. Our third objective was to use occupancy modeling to understand the effect of sampling effort (survey duration and number of surveys) on detection probability. We hypothesized that, given strong latitudinal gradients in the seasonal distribution of twilight periods, Common Nighthawk activity would differ across the study area (Mills 2008). Specifically, we predicted that activity at lower latitude sites would be highly constrained compared to higher latitude sites, given the relatively short twilight duration.

METHODS

Study Area

Our study included 23 survey sites at 12 locations across a 3000 km span of the breeding range in Canada, between Ottawa, Ontario (45°21’ N, 76°0’ W) and Yellowknife, Northwest Territories (62°42’ N, 116°6’ W; Fig. 2). Sites ran from southeast to northwest, so latitude and longitude were negatively correlated (r = -0.96). Seven of the locations had two survey sites, two locations had three survey sites, and three locations had a single survey site. All survey sites were ≥ 1.8 km apart to minimize the likelihood of double counting individual birds. We opportunistically selected sites that were likely to be occupied by Common Nighthawks to identify the most important factors affecting detection of birds that were present, instead of modeling presence itself. We selected survey sites at each location based on prior knowledge of nighthawk occurrence, apparent suitability based on mapped imagery, or apparent suitability determined by biologists in the field. We considered sites suitable if they contained extensive openings with gravel, sand, or rock substrate, or areas that were recently burned or logged (Brigham et al. 2011).

Vocal and non-vocal sounds

Most Common Nighthawk detections comprise two sounds, each with a different behavioral context. The first is a mid-frequency vocal peent (3–5 kHz) with a simple structure (Fig. 3). The peent has widespread use across the home range (Brigham et al. 2011), including a conspecific contact call during travel or foraging, but is most frequently made during territorial and courtship displays (E. C. Knight, unpublished data). The second sound is a low-frequency (0.4–1.0 kHz), non-vocal booming sound made by flexing the wings at the bottom of an aerial dive. Although not well studied, non-vocal booms are presumably only performed by males and appear to function as territorial displays or for guarding mates or nests (Weller 1958, Brigham et al. 2011; Knight et al. 2022). We quantified patterns in each of the sounds separately because each appears to have its own behavioral context, and the mechanisms driving the use and rate of each might differ.

Acoustic recording collection and processing

We collected acoustic recordings at each location using Song Meter ARUs (Model SM2+, Wildlife Acoustics Inc., Maynard, Massachusetts; firmware version 3.2.5). We deployed ARUs in late May or early June 2014 at all locations and retrieved from mid-July to mid-August 2014, based on logistics. Song meters were programmed to record continuously from 1 h prior to sunset to 1 h after sunrise, local time, at each site starting 1 June; the same recording program was repeated every 4 d until 31 July or until ARUs were retrieved. Given the large number of recordings required and the low frequency of typical nighthawk sounds, ARUs were programmed to record at a sampling rate of 16 kHz and a bit depth of 16 bits to conserve memory.

We used two partially automated approaches to obtain data from the large number of acoustic recordings collected. Peent detections were obtained using an automated time-frequency, band-limited energy detector (BLED; see Mills 2000) in Raven Pro (version 1.4; Charif et al. 2010). Frequency range and time durations used to parameterize the detector were defined based on a random sampling of 100 peents from our recordings. For the remaining detector parameters, we used either default settings or adjusted settings until we found the best configuration for isolating peent vocalizations. After running the detector, the resulting list of candidate peent detections was verified manually for accuracy. False positives were scored as 0 and true positives were scored as 1 in the resulting output file. Our detector, when evaluated for classification performance relative to a benchmark dataset, performed well, with a presence-absence recall of ~ 0.6 for five-minute recordings (Knight et al. 2017), suggesting that our resulting dataset was a reasonable representation of true peent activity.

We used a visual scanning approach to count non-vocal booms by viewing spectrograms in Raven Pro. Using the timed auto page-advance function, we could view and rapidly assess 1 min recording segments per second. This enabled us to process 1 h of acoustic recordings in approximately 1 min, depending on the rate of nighthawk booming. The analyst initially audibly confirmed each visually detected candidate boom until signals were verified using only visual detection. Subsequent booms were verified visually on a recording. Once a boom was detected, the timed auto-page advance function was paused and the signal was selected by drawing a box around it, capturing the start time, duration, and frequency range.

We then summarized the peent and boom detections from the continuous recording data into interval time periods to represent the range of durations seen in many bird surveys. We used the “lubridate” and “intrval” packages (Grolemund and Wickham 2011, Sólymos 2017) to separate each continuous recording into 2-minute intervals and determine the number of peent or boom detections in each interval. We repeated this process for the remaining interval lengths (every 2 min from 4–20).

Statistical analysis

Survey timing

We used generalized additive mixed models (GAMMs) in the gamm4 package (Wood et al. 2017) in R 4.0.3 (R Core Team 2020) to examine the influence of survey-specific covariates on peent and boom activity rate (number of peent detections and number of boom detections per survey). We were specifically interested in whether the effects of ordinal day and time since sunset in number of seconds (TSSS), calculated with the suncalc package (Agafonkin and Thieurmel 2017), influenced activity. We also included a categorical variable for latitudinal group (A = 45°–50° [6 sites], B = 50°–55° [11 sites], C = 55°–60° [2 sites], and D = 60°–65° [4 sites]), because the length of nighttime recordings and survey season was notably shorter at northern sites. We standardized TSSS and ordinal day prior to modeling and modeled the effect of latitude as a categorical rather than continuous variable because GAMMs do not model interaction terms but can be run separately for different categories. We did not use longitude because longitude and latitude were strongly correlated (sites ran roughly from southeast to northwest, r = -0.96) and because we were interested in testing hypotheses about twilight levels driven by latitude. We expected a stronger effect of latitude on length of night (shorter nights at more northern sites in the summer due to Earth’s axial tilt) and, therefore, on the relative activity of nighthawks from sunset to sunrise. We used the 10-minute interval datasets and built separate models with the detected number of either peents or booms as the response variable because each sound represents a different behavioral context. We ran a single GAMM for each dependent variable. We included site as a random effect to account for correlations due to the use of repeated recordings from the same site. Because variance in number of peents or booms per interval was consistently greater than the mean count in all sample data sets that we used when developing GAMMs, we used a negative binomial error distribution for our GAMMs, specifying the amount of overdispersion (θ) with the MASS package (Ripley et al. 2013). We specified 5 knots (change points) to allow for up to two nighttime peaks in nighthawk activity due to TSSS and a maximum number of 3 knots to allow for one peak of seasonal activity due to ordinal day. Knots were connected with cubic splines to create a single, continuous piecewise nonlinear function that could vary in form between knots (Wood et al. 2017).

We used bootstrapping with the boot package in R (Davison and Hinkley 1997, Canty and Ripley 2020) to validate our GAMMs and generate quantile estimates for the model coefficients. We first randomly withheld 20 intervals per site without replacement as test data. We generated bootstraps from the remaining training data, with each bootstrap consisting of 20 intervals drawn with replacement from the peent and boom data. We then fit the two GAMMs for peent and boom detections and used the fitted models to predict the numbers of peent and boom detections per 10-minute interval in the withheld test dataset. We bootstrapped this training data selection, model fitting, and model validation process 100 times. We used Spearman correlation coefficients to compare the predicted numbers of peent and boom detections to actual numbers across the 100 bootstraps.

Time period

We used generalized linear mixed models (GLMMs) in the lme4 package (Bates et al. 2015) to examine the influence of twilight period on peent and boom activity rate. We were interested in whether using twilight period was an effective way to constrain surveys and, if so, whether it varied by latitude. Therefore, we built two models for each of the detection types (peent and boom). The first was a non-interaction model that included only a categorical covariate for twilight period. The second was an interaction model that included twilight period, latitude, and the interaction between the two. We again used the 10-minute interval dataset, but only included observations between ordinal days 170–190 (14 June–4 July), which was the peak activity period according to the previous analysis. We also excluded the northernmost six of the total 23 sites because they had no astronomical twilight or true night. Each GLMM was modeled using a negative binomial error distribution with site as a random effect.

Similar to the process used for survey timing, we used bootstrapping to validate our GLMMs and generate quantile estimates for the model coefficients. We used the suncalc package to assign twilight period categories to each 10-minute interval (before sunset, civil twilight PM, nautical twilight PM, astronomical twilight PM, night, astronomical twilight AM, nautical twilight AM, civil twilight AM, and after sunrise). We then set aside 10 observations per twilight period per site as test data. We randomly drew 12 ten-minute intervals with replacement for each site as training data and fit the two GLMMs for each detection type (peent, boom). We used the fitted models to predict the numbers of peent and boom detections per 10-minute interval in the withheld test dataset. We bootstrapped this training data selection, model fitting, and model validation process 100 times. Because we had two models for each dependent variable but ran each model 100 times and because models often failed to converge for a bootstrapped sample dataset, we compared the number of times a given model converged and used Spearman correlation coefficients to compare each model’s predicted numbers of peent and boom detections to actual numbers across the 100 bootstraps.

Survey design

We used an occupancy model in the unmarked package (Fiske and Chandler 2011) to determine the duration and number of recordings that would maximize Common Nighthawk detection probability. For this analysis, we used only sites with confirmed detections of non-vocal booms (n = 19 sites) to ensure we modeled true detection probability. We did not complete this analysis for call (peent) data because this signal is generally not associated with territorial behavior in this species (Knight et al. 2022). Furthermore, applying occupancy models to the call signal across increasing numbers of visits would potentially bias our parameter estimates by violating the assumption of closure (e.g., changes in occupancy) between subsequent recordings at the same site (Rota et al. 2009). To simulate the conditions of an actual survey, we restricted the data included to the survey covariates that the previous GAMM analysis suggested would maximize activity rates (TSSS defined as 1 h before to 1 h after sunset; ordinal day ranged from 14 June–4 July).

We bootstrapped our analysis to generate quantile estimates of detection probability on a given visit along with occupancy probability. We drew 100 samples for each combination of recording duration (every 2 min from 2–20) and number of recordings sampled per site (3–16, 20) and fit a null single-season occupancy model (i.e., no detection or occupancy covariates) to each of our bootstrap sample datasets. Finally, we calculated a logistic curve of detection probability from the bootstrapped predictions to explore how the duration and number of recordings affected detectability.

RESULTS

We detected a total of 61,597 vocal peents and 12,825 non-vocal booms on all our recordings throughout the breeding season. The number of peents varied from 0–25,096 per site (mean = 2678.13 ± 5727.79 SE) and the number of booms varied from 0–3087 per site (mean = 557.61 ± 713.65 SE).

Survey timing

Time since sunset (TSSS) had a stronger impact on both peent and boom activity than ordinal day (Table A1.1). At southern sites (Groups A, 45°–50° N and B, 50°–55° N), there were two peak times of peent activity around sunset and sunrise. For sites in Group A, the primary peak in peent activity occurred 1 h after sunset, and there was a secondary peak 1 h before sunrise. For sites in Group B, the primary peak occurred 0–1 h before sunrise and the secondary peak occurred 0–1 h after sunset. We had unusual results for the two sites in Group C (55°–60° N), in that peak peent activity was at 2 h after sunset and 1 h after sunrise, but we noted similar patterns in the number of peent detections relative to TSSS in the raw data. Finally at the most northern sites (Group D, 60°–65° N), there was one peak in peent activity from 1 h before to 2 h after sunset, but peent activity extended across a greater proportion of the nighttime period though it declined to zero by sunrise (Fig. 4).

We found different results for predicted boom activity, with a primary peak 2 h before sunrise and a secondary peak 0–1 h after sunset at all southern sites (Groups A and B). For the two sites in Group C, there was a primary peak in activity 2 h after sunset and a secondary peak 6 h after sunset, which extended past sunrise. In Group D, boom activity was highest 0–2 h after sunset and was relatively high before sunset as well but declined to zero by sunrise (Fig. 5).

Peent and boom activity were highest at the start of the season at the southernmost sites (Group A). Peent activity peaked around ordinal day 180 (June 29) at sites in Groups B and D but was highest at the end of the season for the two sites in Group C. Boom activity had a broad peak from ordinal days 175–195 (June 24–July 14) at sites in Group B, but peaked around ordinal day 170 (June 19) for sites further north (Groups C and D). For sites in Group C, there was a secondary increase in boom activity following the first peak, steadily increasing to the end of the season (Figs. 6 and 7).

Our model validation process showed moderately positive Spearman correlations between predicted and actual peent detections (median Spearman correlation = 0.22, 95% BCI = 0.10–0.28) and between predicted and actual boom detections (median Spearman correlation = 0.25, 95% BCI = 0.18–0.30).

Time period

There were more twilight periods of peak activity for peents than booms. Peent activity was highest from nautical twilight PM to civil twilight AM, whereas boom activity was highest in civil twilight PM and nautical twilight AM (Fig. 8). However, the peent model only converged for 38% of the bootstraps and the goodness-of-fit of the peent model was poor (median Spearman correlation = -0.07, 95% BCI = -0.05 to -0.10). In contrast, the boom model fully converged and had moderately positive performance (median Spearman correlation = 0.24, 95% BCI = 0.25–0.26). In contrast to TSSS in the generalized additive mixed models, there was little evidence that latitude had any effect on the relationship between twilight and Common Nighthawk activity. In fact, very few of the models with an interaction between twilight period and latitude converged (2% for peents, 27% for booms), suggesting little signal in the data (Table A2.1).

Survey design

Detection probability for booms during a given visit did not reach 1 for any of the combinations of number and duration of recordings (Fig. 9). Duration of recording substantially improved detection probability for booms, with detection probability greater than 0.5 with recording durations of 11 and 16 minutes, respectively.

Given that all sites analyzed with occupancy models were truly occupied, the probability of predicting occupancy increased with duration and number of recordings, but more longer recordings were required. The probability of predicting true occupancy based on boom detections increased with number and duration of recordings but was high (ψ ≥ 0.50) even using just three recordings of 2-min duration (Fig. 9).

DISCUSSION

Effective survey protocols are necessary to provide reliable information inputs to the management and conservation of birds. For species like the Common Nighthawk that have large geographic ranges, best monitoring practices and protocols may need to incorporate location-specific recommendations because detectability can vary across the range. We used a temporally intensive and spatially extensive dataset of recordings from autonomous recording units (ARUs) from across the northern portion of the breeding range to understand geographic variation in detectability and inform species-specific surveys. As hypothesized, we found that Common Nighthawk vocal (peent) and non-vocal (boom) activity rate relative to time since sunset varied among latitudinal groups; however, all other effects of latitude were minor. Despite variation in activity patterns with latitude, our results suggest that a single set of monitoring recommendations are suitable for the entire northern breeding range of the Common Nighthawk. We discuss the specifics of those recommendations below.

Survey timing

The most important predictor of Common Nighthawk peent and boom activity was time relative to sunset. Other researchers have found similar patterns, with more frequent activity after dusk and prior to dawn (Wedgewood 1973, 1991, Farrell et al. 2017, Farrell et al. 2019). However, our results were novel in that the pattern of activity relative to sunset and sunrise depended on latitude, as follows: southern sites (≤ 55° N) had two sharp peaks in activity, one relative to sunset and one to sunrise, whereas northern sites (> 60° N) had one broad peak relative to sunset. We also found Common Nighthawk boom activity was highest during the civil twilight period but did not vary with latitude. Given that the Common Nighthawk is a crepuscular species that relies on visual cues for foraging (Brigham and Barclay 1995), we suggest that the latitudinal differences we found can be attributed to differences in the length of the civil twilight period (Mills 2008). At low latitudes, where the civil twilight period is constrained, nighthawks may need to also make use of the dawn period for foraging and territory defense, resulting in two sharp peaks of activity. At high latitudes, where the civil twilight period is extended, nighthawks undertake one long period of activity.

Therefore, we suggest that surveys for this species should occur during civil twilight periods. This would allow for the latitudinal variation in timing of activity to be incorporated without complicating instructions or requiring region-specific protocols. Simple survey protocols are often most effective, particularly for citizen science programs like the existing Canadian Nightjar Survey (Parsons et al. 2011, McKinley et al. 2017). Twilight-based surveys would also provide surveyors with greater flexibility as opposed to a strict survey window relative to sunset. Given the relatively long twilight periods in the north, especially around the summer solstice, surveys could be conducted starting at dusk, running continuously until dawn. In the south, surveys could be conducted after sunset and then could resume during pre-dawn nautical twilight the following morning (Fig. 10).

Day of year also had an influence on Common Nighthawk activity, with substantial differences between the two detection types. Boom activity had a low, broad peak across the study area, with a later peak at northern sites than southern sites, presumably due to latitudinal differences in arrival timing or breeding phenology. However, we note that there is a confound between latitude and longitude in our dataset, and that longitude may also have an influence on arrival timing in this species (E. C. Knight, unpublished data). Booms tend to be territorial signals; therefore, the activity rate of booms would be expected to rise and fall with the breeding season (Catchpole 1973, Nowicki and Searcy 2004, Upham-Mills et al. 2020). Peents, on the other hand, are a more all-purpose signal (Brigham et al. 2011; E. C. Knight, unpublished data), so the activity rate of peents may increase throughout the season as new individuals (juveniles) join the population.

Acoustic signal

We suggest the behavioral significance of booms is therefore more useful than peents for surveying to infer Common Nighthawk breeding locations. Booms are better for population monitoring because, like passerine song, they usually indicate territoriality, are thought to be produced only by males, and are only produced by adults (Brigham et al. 2011, Knight et al. 2022). Common Nighthawk territories inferred by booms are likely a better representation of the breeding territory than areas used for feeding and other activities, as inferred by peents. As such, our results suggest that the number of recordings had less of an effect on estimates of detectability and occupancy for booms than for peents. Therefore, given the larger spatial context for peents, using this signal to define occupancy likely violates the closure assumption and may result in biased estimates of occupancy and detection probability (MacKenzie et al. 2002). We therefore suggest that, at a minimum, booms should be recorded when surveying for Common Nighthawks; however, survey data are most useful if both acoustic signals are recorded. Recording peents is particularly important if survey data are to be used to model habitat associations at the home range scale (Knight et al. 2021b). Luckily, Common Nighthawks vocalize every time they boom, which facilitates recording both types of acoustic signal.

The acoustic characteristics of booms also make them superior for breeding surveys because they have a larger survey radius. Booms are a lower frequency sound than peents, which causes them to attenuate less and travel further than peents. Peents can be detected with a recognizer on recordings made over 500 m away (Knight and Bayne 2019), so the detection radius for the boom is likely much greater. A larger survey radius allows effort to be spread more broadly across the landscape, which is an important advantage when the space to survey is large and the time windows for surveys are short (e.g., southern sites).

Survey design

Ultimately, best monitoring practices will vary with the intended use of the data, and this is particularly true for survey design. Imperfect detection can be accounted for statistically if the objective of surveys is population monitoring or habitat analyses (MacKenzie et al. 2002, Sólymos et al. 2013; Buckland et al. 2015). In fact, Common Nighthawk detectability is high enough (0.3) for a survey visit of only a few minutes to warrant using only two visits as in input to occupancy modeling (MacKenzie et al. 2002). Two other studies in northwestern Ontario, Canada have found similarly high detection probabilities: 0.6 at 1 h after sunset (Farrell et al. 2017) and between 0.4 and 0.6 up to 2 h after sunset and at sunrise (Farrell et al. 2019).

On the other hand, if the objective of surveying is to obtain strong confirmation of presence or absence (e.g., for environmental assessment or predevelopment site evaluations), then surveys should be designed to maximize detection probability. Duration of recordings had a greater effect than number of recordings on Common Nighthawk detectability. In these situations, we suggest surveying for booms and maximizing the duration of survey visits to achieve a detection probability as close to 1 as possible. We achieved detection probabilities between 0.4 and 0.6 for twenty-minute recordings but did not approach an asymptote. Longer duration surveys will likely result in even higher detection probability.

Survey type

The technology used to survey will likely also depend on the objectives of the survey. Recorders like the ARUs used here are an excellent tool to survey the Common Nighthawk. In particular, detection probability can be maximized by using ARUs to record at times of peak acoustic activity and collecting multiple recordings, as done in Shonfield and Bayne (2017) and Gibb et al. (2018). Autonomous recorders can also be deployed safely in remote locations during the day and programmed to ensure timing coincides with the crepuscular periods of peak activity found here. The resultant ARU recordings can be processed into Common Nighthawk detections quickly and reliably; signal recognition technology works well for this species because the call is simple, consistent, and frequent (Knight et al. 2017), and there are additional tools available to further improve processing efficiency and data utility (Knight and Bayne 2019, Yip et al. 2019, Knight et al. 2020). Given high unit costs, the broad spatial extent of the Common Nighthawk breeding range, and the need for temporal replication across years, ARUs may not be a practical solution for population monitoring at large scales. However, with the recent development of small, low cost, open-source passive acoustic recording devices, the potential scalability of acoustic monitoring may become a viable option in the future (Whytock and Christie 2017, Hill et al. 2018, Beason et al. 2018).

Recommendations summary

Optimizing detections of rare species, especially those of conservation concern, should be a goal of any targeted survey or monitoring program. Our results simplify previous survey guidance and protocols by providing clear and simple recommendations for future species-specific surveys of Common Nighthawks. Common Nighthawk surveys should be conducted within the civil twilight period, when possible. Using civil twilight timing greatly simplifies the design and logistics of species-specific surveys, especially because determining the start and duration of local civil twilight can easily be obtained from numerous sources (e.g., websites, smart phone applications). In southern Canada, the civil twilight period can be short in summer months, and there are narrow survey windows for nighthawks shortly after sunset and before sunrise. Further north, the window for nighthawk surveys during civil twilight is longer and should be closer to sunset than sunrise. In southern Canada (≤ 50° N), surveys should occur throughout June; surveys further north (≥ 50° N) should occur between mid-June and mid-July to coincide with the peak breeding activity period at higher latitudes. Given the strong associations of non-vocal booms with nesting sites and breeding activity, we recommend that surveys target these unique acoustic signals. Longer surveys of up to 20 minutes would increase the detection of nighthawks that are present at sites. This duration is longer than used in many nocturnal bird surveys but could be facilitated by using ARUs. Given the relatively high rate of activity and the simplicity of acoustic signals in this species, ARUs can continue to play an important role in any monitoring program for this species.

LITERATURE CITED

Agafonkin, V., and B. Thieurmel. 2017. suncalc: Compute sun position, sunlight phases, moon position, and lunar phase. R package version 0.34. http://suncalc.net

Amrhein, V., P. Korner, and M. Naguib. 2002. Nocturnal and diurnal singing activity in the nightingale: correlations with mating status and breeding cycle. Animal Behaviour 64(4):939-944. https://doi.org/10.1006/anbe.2002.1974

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.

Beason, R. D., R. Riesch, and J. Koicheva. 2018. AURITA: an affordable, autonomous recording device for acoustic monitoring of audible and ultrasonic frequencies. Bioacoustics 28(4):381-396. https://doi.org/10.1080/09524622.2018.1463293

Brigham, R. M., and R. M. Barclay. 1995. Prey detection by common nighthawks: does vision impose a constraint? Ecoscience 2(3):276-279. https://doi.org/10.1080/11956860.1995.11682294

Brigham, R. M., and M. B. Fenton. 1991. Convergence in foraging strategies by two morphologically and phylogenetically distinct aerial insectivores. Journal of Zoology 223(3):475-489. https://doi.org/10.1111/j.1469-7998.1991.tb04778.x

Brigham, R. M., J. Ng, R. G. Poulin, and S. D. Grindal. 2011. Common Nighthawk (Chordeiles minor). In A. F. Poole, editor. The birds of North America. Cornell Lab of Ornithology, Ithaca, New York, USA. https://birdsna.org/Species-Account/bna/species/comnig https://doi.org/10.2173/bow.comnig.01

Buckland S. T., E. A. Rexstad, T. A. Marques, and C. S. Oedekoven. 2015. Distance sampling: methods and applications. Springer, Heidelberg, Germany. https://doi.org/10.1007/978-3-319-19219-2

Canty, A., and B. Ripley. 2020. boot: Bootstrap R (S-Plus) Functions. R package version 1.3-25.

Catchpole, C. K. 1973. The functions of advertising song in the Sedge Warbler (Acrocephalus schoenobaenus) and Reed Warbler (A. scirpaceus). Behaviour 46:300-320. https://doi.org/10.1163/156853973X00067

Charif, R. A., A. M. Waack, and L. M. Strickman. 2010. Raven Pro 1.4 User’s Manual. Cornell Lab of Ornithology, Ithaca, New York, USA.

Committee on the Status of Endangered Wildlife in Canada (COSEWIC). 2018. COSEWIC assessment and status report on the Common Nighthawk Chordeiles minor in Canada. Ottawa, ON, Canada. https://www.canada.ca/en/environment-climate-change/services/species-risk-public-registry/cosewic-assessments-status-reports/common-nighthawk-2018.html

Conway, C. J., and J. P. Gibbs. 2011. Summary of intrinsic and extrinsic factors affecting detection probability of marsh birds. Wetlands 31:403-411. https://doi.org/10.1007/s13157-011-0155-x

Davison, A. C., and D. V. Hinkley. 1997. Bootstrap methods and their applications. Cambridge University Press, Cambridge, UK. https://doi.org/10.1017/CBO9780511802843

Digby, A., M. Towsey, B. D. Bell, and P. D. Teal. 2013. A practical comparison of manual and autonomous methods for acoustic monitoring. Methods in Ecology and Evolution 4(7):675-683. https://doi.org/10.1111/2041-210X.12060

Farrell, C. E., L. Fahrig, G. Mitchell, and S. Wilson. 2019. Local habitat association does not inform landscape management of threatened birds. Landscape Ecology 34:1313-1327. https://doi.org/10.1007/s10980-019-00843-6

Farrell, C. E., S. Wilson, and G. Mitchell. 2017. Assessing the relative use of clearcuts, burned stands, and wetlands as breeding habitat for two declining aerial insectivores in the boreal forest. Forest Ecology and Management 386:62-70. https://doi.org/10.1016/j.foreco.2016.11.026

Fiske, I., and R. Chandler. 2011. Unmarked: an R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43(10):1-23. https://doi.org/10.18637/jss.v043.i10

Gibb, R., E. Browning, P. Glover-Kapfer, and K. E. Jones. 2019. Emerging opportunities and challenges for passive acoustics in ecological assessment and monitoring. Methods in Ecology and Evolution 10:169-185. https://doi.org/10.1111/2041-210X.13101

Grolemund, G., and H. Wickham. 2011. Dates and times made easy with lubridate. Journal of Statistical Software 40(3):1-25.

Hill, A. P., P. Prince, E. P. 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(5):1199-1211. https://doi.org/10.1111/2041-210X.12955

Knight, E. C., and E. M. Bayne. 2019. Classification threshold and training data affect the quality and utility of focal species data processed with automated audio-recognition software. Bioacoustics 28(6):539-554. https://doi.org/10.1080/09524622.2018.1503971

Knight, E. C., R. M. Brigham, and E. M. Bayne. 2021b. Specialist or generalist? It depends. Context-dependent habitat relationships provide insight into forest disturbance effects for a boreal bird species. Forest Ecology and Management 502:119720. https://doi.org/10.1016/j.foreco.2021.119720

Knight, E. C., R. M. Brigham, and E. M. Bayne. 2022. The Big Boom Theory: the Common Nighthawk wing-boom display delineates exclusive nesting territories. Ornithology 139(1):ukab066. https://doi.org/10.1093/ornithology/ukab066

Knight, E., K. Hannah, R. M. Brigham, J. McCracken, G. Falardeau, M.-F. Julien, and J.-S. Guénette. 2019. Canadian Nightjar Survey protocol. Birds Canada, Port Rowan, Ontario, Canada.http://wildresearch.ca/wp-content/uploads/2019/05/National-Nightjar-Survey-Protocol-WildResearch-2019.pdf

Knight, E. C., K. C. Hannah, G. Foley, C. Scott, R. Mark Brigham, and E. Bayne. 2017. Recommendations for acoustic recognizer performance assessment with application to five common automated signal recognition programs. Avian Conservation and Ecology 12(2):14. https://doi.org/10.5751/ACE-01114-120214

Knight, E. C., A. C. Smith, R. M. Brigham, and E. M. Bayne. 2021a. Combination of targeted monitoring and Breeding Bird Survey data improves population trend estimation and species distribution modeling for the Common Nighthawk. Ornithological Applications 123:duab005. https://doi.org/10.1093/ornithapp/duab005

Knight, E. C., P. Sólymos, C. Scott, and E. M. Bayne. 2020. Validation prediction: a flexible protocol to increase efficiency of automated acoustic processing for wildlife research. Ecological Applications 30(7):e02140. https://doi.org/10.1002/eap.2140

Krebs, J. R., and Kacelnik. 1983. The dawn chorus in the great tit (Parus major): proximate and ultimate causes. Behaviour 83:287-309. https://doi.org/10.1163/156853983X00200

Lima, K. A., E. M. Call, T. P. Hodgman, D. S. Potter, S. Gallo, and E. J. Blomberg. 2020. Environmental conditions and call-broadcast influence detection of eastern forest owls during standardized surveys. Condor 122(3):1-20. https://doi.org/10.1093/condor/duaa016

MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, Royle, A., 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

Martin, K., N. Koper, and R. Bazin. 2014. Optimizing repeat-visit, call-broadcast nocturnal surveys for Yellow Rails (Coturnicops noveboracensis). Waterbirds 37(1):68-78. https://doi.org/10.1675/063.037.0109

McKinley, D. C., A. J. Miller-Rushing, H. L. Ballard, R. Bonney, H. Brown, S. C. Cook-Patton, D. M. Evans, R. A. French, J. K. Parrish, T. B. Phillips, S. F. Ryan, L. A. Shanley, J. L. Shirk, K. F. Stepenuck, J. F. Weltzin, A. Wiggins, O. D. Boyle, R. D. Briggs, S. F. Chapin III, D. A. Hewitt, P. W. Preuss, and M. A. Soukup. 2017. Citizen science can improve conservation science, natural resource management, and environmental protection. Biological Conservation 208:15-28. https://doi.org/10.1016/j.biocon.2016.05.015

Mills, A. M. 2008. Latitudinal gradients of biologically useful semi-darkness. Ecography 31:578-582. https://doi.org/10.1111/j.0906-7590.2008.05346.x

Mills, H. 2000. Geographically distributed acoustical monitoring of migrating birds. Journal of the Acoustic Society of America 108(5):2582. https://doi.org/10.1121/1.4743594

Nowicki, S. and W. A. Searcy. 2004. Song function and the evolution of female preferences: why birds sing, why brains matter. Annals of the New York Academy of Sciences 1016(1):704-723. https://doi.org/10.1196/annals.1298.012

Parsons, J., R. Lukyanenko, and Y. Wiersma. 2011. Easier citizen science is better. Nature 471:37. https://doi.org/10.1038/471037a

Priyadarshani, N., S. Marsland, and I. Castro. 2018. Automated birdsong recognition in complex acoustic environments: a review. Journal of Avian Biology 49(5):01447. https://doi.org/10.1111/jav.01447

R Core Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/

Resources Inventory Committee. 1998. Inventory methods for nighthawk and poorwill. Standards for components of British Columbia’s biodiversity no. 9. Ministry of Environment, Lands and Parks, Victoria, British Columbia, Canada. https://www2.gov.bc.ca/assets/gov/environment/natural-resource-stewardship/nr-laws-policy/risc/poorw.pdf

Ridgely, R. S., T. F. Allnutt, T. Brooks, D. K. McNicol, D. W. Mehlman, B. E. Young, and J. R. Zook. 2012. Digital distribution maps of the birds of the western hemisphere, version 5.0. In BirdLife International and NatureServe, Bird species distribution maps of the world. Version 2.0. BirdLife International, Cambridge, UK and NatureServe, Arlington, USA.

Ripley, B., B. Venables, D. M. Bates, K. Hornik, A. Gebhardt, D. Firth, and M. B. Ripley. 2013. Package ‘MASS’. CRAN R 538:113-120. http://www2.uaem.mx/r-mirror/web/packages/MASS/

Rota, C. T., R. J. Fletcher, Jr., R. M. Dorazio, and M. G. Betts. 2009. Occupancy estimation and the closure assumption. Journal of Applied Ecology 46(6):1173-1181. https://doi.org/10.1111/j.1365-2664.2009.01734.x

Sauer, J. R., and W. A. Link. 2011. Analysis of the North American Breeding Bird Survey using hierarchical models. Auk 128:87-98. https://doi.org/10.1525/auk.2010.09220

Schroeder, K. M., and S. B. McRae. 2019. Automated auditory detection of rare, secretive marsh bird with infrequent and acoustically indistinct vocalizations. Ibis 162(3):1033-1046. https://doi.org/10.1111/ibi.12805

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

Sólymos, P. (2017). Relational operators for intervals. https://cran.r-project.org/web/packages/intrval/intrval.pdf

Sólymos, P., S. M. Matsuoka, E. M. Bayne, S. R. Lele, P. Fontaine, S. G. Cumming, D. Stralberg, F. K. A. Schmiegelow, and S. J. Song. 2013. Calibrating indices of avian density from non-standardized survey data: making the most of a messy situation. Methods in Ecology and Evolution 4(11):1047-1058. https://doi.org/10.1111/2041-210X.12106

Staicer, C. A., D. A. Spector, and A. G. Horn. 1996. The dawn chorus and other diel patterns in acoustic signalling. Pages 426-453 in D. E. Kroodsma and E. H. Miller, editors. Ecology and evolution of acoustic communication in birds. Cornell University Press, Ithaca, New York, USA. https://doi.org/10.7591/9781501736957-033

Stowell, D., M. D. Wood, H. Pamuła, Y. Stylianou, and H. Glotin. 2019. Automatic acoustic detection of birds through deep learning: the first Bird Audio Detection challenge. Methods in Ecology and Evolution 10(3):368-380. https://doi.org/10.1111/2041-210X.13103

Thompson, W. 2013. Sampling rare or elusive species: concepts, designs, and techniques for estimating population parameters. Island Press, Washington, D.C., USA.

Upham-Mills, E. J., J. R. Reimer, S. Haché, S. R. Lele, and E. M. Bayne. 2020. Can singing rate be used to predict male breeding status of forest songbirds? A comparison of three calibration models. Ecosphere 11(1):e03005. https://doi.org/10.1002/ecs2.3005

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

Viel, J. G., A. Donnelly, G. Fredlund, W. Mueller, and R. Brady. 2020. Common Nighthawk (Chordeiles minor) occurrence in urban areas Southeastern Wisconsin, U.S.A. American Midland Naturalist 183:233-245. https://doi.org/10.1637/0003-0031-183.2.233

Wedgwood, J. A. 1973. Nighthawks in the city. Blue Jay 31:82-88. https://doi.org/10.29173/bluejay3911

Wedgwood, J. A. 1991. Common Nighthawks in Saskatoon. Blue Jay 50(4):211-217. https://doi.org/10.29173/bluejay5092

Weller, M. W. 1958. Observations of the incubation behavior of a Common Nighthawk. Auk 75(1):48-59. https://doi.org/10.2307/4082060

Whytock, R. C., and J. Christie. 2017. Solo: an open source, customizable and inexpensive audio recorder for bioacoustic research. Methods in Ecology and Evolution 8(3):308-312. https://doi.org/10.1111/2041-210X.12678

Wilson, D. M., and J. Bart. 1985. Reliability of singing bird surveys: effects of song phenology during the breeding season. Condor 87:69-73. https://doi.org/10.2307/1367132

Wilson, M. D., and B. D. Watts. 2006. Effect of moonlight of detection of Whip-poor-wills: implications for long-term monitoring strategies. Journal of Field Ornithology 77(2):207-211. https://doi.org/10.1111/j.1557-9263.2006.00042.x

Wood, S., F. Scheipl, and M. S. Wood. 2017. Package ‘gamm4’. American Statistician 45:339.

Yip, D. A, E. C. Knight, E. Haave-Audet, S. J. Wilson, C. Charchuk, C. D. Scott, P. Sólymos, and E. M. Bayne. 2019. Sound level measurements from audio recordings provide objective distance estimates for distance sampling wildlife populations. Remote Sensing in Ecology and Conservation 6:301-315. https://doi.org/10.1002/rse2.118

Zuberogoitia, I., J. E. Martínez-Franco, J. A. González-Oreja, C. González de Buitrago, G. Belamendia, J. Zabala, M. Laso, N. Pagaldai, and M. V. Jiménez-Franco. 2020. Maximizing detection probability for effective large-scale nocturnal bird monitoring. Diversity and Distributions 26(8):1034-1050. https://doi.org/10.1111/ddi.13075