Passive acoustic monitoring with AI-based detection and identification reveal Sooty Grouse hooting patterns in western Oregon

Passive acoustic monitoring with AI-based detection and identification reveal Sooty Grouse hooting patterns in western Oregon | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL Wildlife Biology This is a preprint and has not been peer reviewed. Data may be preliminary. 28 April 2025 V1 Latest version Share on Passive acoustic monitoring with AI-based detection and identification reveal Sooty Grouse hooting patterns in western Oregon Authors : Kelly Walton 0009-0003-4851-4692 [email protected] , Sarah Frey 0000-0002-4343-0700 , Cara Christensen 0009-0006-1629-9416 , Mikal Cline , Lauren Gramberg , and Jonathan Dinkins Authors Info & Affiliations https://doi.org/10.22541/au.174585482.29885531/v1 Published Wildlife Biology Version of record Peer review timeline 579 views 265 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Many bird species are monitored using auditory point count surveys during the breeding season. Advances in passive acoustic technology have enabled the use of autonomous recording units (ARUs) alongside point-count surveys, improving survey methodologies. However, automated song/call identification and manual review of recordings are required to assess accuracy of data collected from ARUs. We evaluated the feasibility and accuracy of PNW-Cnet, an AI-based detection application, for identifying the hooting song of male Sooty Grouse (Dendragapus fuliginosus). From 2020–2023, we deployed ARUs at 149 locations in western Oregon, USA near known hooting males. We used PNW-Cnet to identify hoots and the accuracy of the detector was calculated by manually verifying 10,000 detections. Accuracy was near perfect relative to false detections. Once hoots were identified, we used generalized additive models with random effects to examine seasonal (across the breeding season) and daily (relative to time since sunrise) hooting patterns. Model results indicated hooting rates peaked in late-April, providing guidance on the optimal timing of point count surveys based on the estimated number of hoots that would be heard per survey. Daily patterns revealed a rapid increase in hooting 30 minutes before sunrise, then leveling at a relatively constant hooting rate up to at least six hours after sunrise (latest time our ARUs were recording). Thus, our results suggest males continue to vocalize throughout the entire morning allowing effective surveys to be conducted beyond the early morning. By integrating ARUs and an AI-based detection application, we gained detailed information about hooting patterns that will allow improvements to future data collection, increasing survey efficiency, and ultimately leading to a more efficient population monitoring. \papertype Original Article INTRODUCTION Large-scale, multi-species bird surveys are an efficient way to estimate abundance and trends for monitoring populations (Sauer et al. 2017, Bibby et al. 2000; Farnsworth et al. 2002, Ziolkowski et al. 2023). Locations for these surveys are spatially distributed throughout many vegetation types, representing the habitat of multiple species; thus, these surveys are able to count many species through visual and auditory detections. One example of a broad-scale point count survey is the Breeding Bird Survey (BBS), which not only has a long-term dataset starting in 1966 and population change data for ~424 species, but also has transects across the USA, Canada, and Mexico (Robbins et al. 1986, Sauer et al. 2017). The BBS occurs in late spring during the early morning hours when most passerine activity is high (Ziolkowski et al. 2023). However, some species have low detection probability using these survey protocols because of low abundance in surveyed areas, minimal overlap of range or habitat within surveyed areas (Sauer et al. 2013), and timing of the survey. In addition, some vegetation types, such as dense forests, and landscape features can reduce acoustic detectability and the distance from which singing birds can be detected (i.e., sound attenuation in thick vegetation or complex terrain; Winiarska et al. 2024). For species that breed earlier or later in the year, these surveys do not align with peak vocalization or their breeding season. In addition, BBS survey routes in most areas of the lower 48 states of the USA are mainly constrained to easier-to-access roads (Sauer et al. 2013). Therefore, BBS routes do not sample areas with poor accessibility at a high enough rate to detect trends; these areas are often critical habitat for many species. Many crepuscular, nocturnal, or cryptic species are not accurately captured during broad-scale point count surveys and separate species-specific surveys are often necessary to record sufficient detections to estimate a population trend (Conway and Gibbs 2011, Martin et al. 2014, Lima et al. 2020, Zuberogoitia et al. 2020, Knight et al. 2021, Hannah et al. 2022). Sooty Grouse ( Dendragapus fuliginosus ) is a montane forest grouse species that is found primarily in rugged, high-elevation conifer forests in western North America (Zwickel and Bendell 2004, Zwickel and Bendell 2020). Sooty Grouse inhabits a variety of forest stand ages, ranging from edges of timber cuts to old-growth forests, although they often utilize openings created by natural or human-caused disturbances (Zwinkel and Bendell 1972, Redfield 1973, Doerr et al. 1984, Niederleitner 1987, Zwickel and Bendell 2020, Zwinkel and Bendell 2024). Males defend territories and attract females using a hooting song, which begins in late-March to early-April, increasing through April and mid-May, and then declines (Zwinkel and Bendell 2004). Throughout the range of Sooty Grouse in the Pacific Northwest, most forested areas have been impacted by human development, climate change, wildfire frequency, and forest management practices (USDA and USDI 1994, Halofsky et al. 2020), and the long-term impacts of these changes to Sooty Grouse abundance is unknown. The State of the Birds watchlist identified Sooty Grouse as a species in need of urgent conservation attention to reduce the risk of becoming threatened or endangered (Rosenberg et al. 2014). Watchlist species were determined using the Partners in Flight Species Assessment Database, which ranked Sooty Grouse as vulnerable due to a declining population and threats to their western forest habitats (North American Bird Conservation Initiative, U.S. Committee 2009, Rosenberg et al. 2014). However, this assessment relied heavily on population trends calculated from BBS data. BBS trends may not be representative of the true population trend of Sooty Grouse, especially assessing at smaller spatial scales, because BBS surveys often take place after the peak breeding period and have low encounter rates for Sooty Grouse. In addition, BBS routes do not survey much high-elevation breeding habitat that Sooty Grouse occupy. Prior to 2011, some information existed from historical surveys in Oregon, including counts from summer brood routes and opportunistic spring counts of hooting grouse along forest roads, but no formal survey protocol with consistent methods had been followed (ODFW unpublished data). In 2011, we implemented an annual Sooty Grouse hooting survey from 1 April – 15 May across the breeding range in western Oregon, USA using a consistent protocol. The survey consists of transects with 15–25 listening stops where 3-minute auditory surveys are conducted during the early morning (a half hour before sunrise to 3 hours after sunrise). The goal of these surveys was to monitor annual abundance and estimate population trends (Fox et al. 2011, Frey et al. In review ). After over a decade of hooting surveys, challenges still existed with the current survey methodology due to lack of knowledge about Sooty Grouse’s local breeding ecology. Peak breeding activity across the range in Oregon likely varies with environmental conditions and geography, such as weather, vegetation, elevation, and latitude. Other studies have shown geographic variation in time of peak hatch is likely related to weather and/or indirect interactions with vegetation (Zwickel 1977, Zwickel and Bendell 2002). Ideally surveys are conducted during the peak breeding period to maximize detection and include temporal replication to account for variability in detection. The accessibility of survey routes due to snow and spring weather conditions often limits when surveys can be conducted and whether all stops on a route are surveyed each visit. In addition, since surveys take place in remote mountainous areas, travel time to routes limits how many surveys can be conducted spatially and temporally, making it important to maximize efficiency. With advances in passive acoustic technology for monitoring species, there are new options for improving survey techniques while reducing personnel field visits, such as the use of autonomous recording units (ARUs). ARUs are relatively inexpensive, easy to deploy, capture data over long time periods, can reduce human bias, and provide a permanent acoustic recording (Shonfield and Bayne 2017, Digby et al. 2023). Advances in automated recognition software have made it possible to identify vocalizations of interest without having to manually review all recordings, although some manual validation is still necessary (Priyadarshani et al. 2018, Stowell et al. 2019, Ruff et al. 2023). ARUs may not detect as many vocalizations as a human observer depending on the distance an ARU can pick up a vocalization, which varies based on the ARU model, microphone (Rempel et al. 2013, Turgeon et al. 2017), characteristics of the vocalization, and barriers that may exist in the landscape (e.g., topography, vegetation). ARUs can be used in conjunction with in-person auditory surveys to better understand when a species starts vocalizing, the frequency and consistency of vocalizations, and potential factors that may influence when a species vocalizes, which can all help increase the efficiency of surveys. ARUs are ideal for determining occupancy and estimates of density and abundance can be obtained for some species (Shonfield and Bayne 2017, Yip et al. 2017a, b). We implemented hooting surveys designed to monitor Sooty Grouse populations in western Oregon starting during the 2011 breeding season. To address limitations of these surveys, we used ARUs to examine daily and seasonal hooting activity and investigate influencing factors. We utilized an AI-based detection application to identify Sooty Grouse hooting vocalizations in our recordings. First, we evaluated the effect of day of year, elevation, latitude, daily precipitation, minimum daily temperature, and study region on seasonal hooting patterns. Second, we examined how time of day (time since sunrise), precipitation, and minimum daily temperature affected daily hooting patterns. We hypothesized that both seasonal and daily hooting activity would increase following the start of the breeding season or day, have one peak followed by a decrease in hooting in the late-spring or late morning, respectively. We expected hooting activity would vary due to weather conditions with hooting activity decreasing with colder temperatures and greater amounts of precipitation. Finally, we predicted that elevation and latitude would influence the timing of hooting patterns across western Oregon with peak hooting later in the season for higher elevations and latitudes. Our results provide valuable behavioral information relative to hooting activity during the breeding season, while also informing our current in-person hooting survey protocol, thereby, improving efficiency and accuracy of data collection. A well-designed monitoring protocol that can be used to understand population trends is an important component to management of this upland game bird species in Oregon. \papertype Original Article MATERIALS AND METHODS Study Area We stratified our study area by four distinct terrestrial physiographic provinces in western Oregon—Coast Range, West Cascades, East Cascades, and Klamath—and hereafter refer to them as study regions. The Coast Range and West Cascades are wetter than the East Cascades and Klamath as moisture decreases north to south and the West Cascades create a rain shadow for the eastern slopes of the Cascade Mountains. The Coast Range is characterized by steep mountain slopes and ridges from sea level to 450–750 m. a. s. l. Cool, moist air from the ocean creates a temperate forest dominated by Douglas-fir ( Pseudotsuga menziesii ), western hemlock ( Tsuga heterophyll ), and western red cedar ( Thuja plicata ). Major disturbances include wildfire and timber harvest, which have created fragmentation of older forest stands. The West Cascades extends from the foothills on the western slope of the Cascade Mountains to the crest over 3,000 m. a. s. l. Forests are dominated by conifers, primarily Douglas-fir and western hemlock below 1,200 m. a. s. l. At higher elevations, dominant tree species include Pacific silver fir ( Abies amabilis ), mountain hemlock ( Tsuga mertensiana ), and subalpine fir ( Abies lasiocarpa ). In the southern forests, ponderosa pine ( Pinus ponderosa ), sugar pine ( Pinus lambertiana ), and incense cedar ( Calocedrus decurrens ) often are found with Douglas-fir at lower elevations. The Klamath region in southwestern Oregon includes the Umpqua and Siskiyou Mountains and the interior valleys and foothills east to the Cascade Range. This area is diverse in plant and wildlife species and dominated by mixed conifer and mixed conifer/hardwood forests with high fire frequency. The East Cascades extends from the Cascade crest to the warmer, drier high desert of central and eastern Oregon. It is dominated by mixed conifer, pine, and by true fir and mountain hemlock forests at higher elevations (USDA and USDI 1994). Autonomous Recording Units (ARUs) Acoustic recordings were collected using Swift ARUs (first generation Swift with one omnidirectional microphone, Cornell University, New York, USA). ARUs were programmed to record from 5 am to 12 pm (local daylight savings time) daily at a sampling rate of 16 kHz. ARUs were mounted to a tree or stump approximately 1.5 m above ground at preselected locations or locations where Sooty Grouse were heard (see Selection of Sites section below). ARUs were placed at 14–66 locations per year from 2020–2023 (2020 n = 14, 2021 n = 66, 2022 n = 49, and 2023 n = 20). ARUs were deployed at locations identified from previous surveys and as close to mid-February as logistically possible. Data was recorded daily until the unit was retrieved in early-July. Selection of Sites We selected sites for ARU placement opportunistically to maximize the likelihood of a male Sooty Grouse being present to record vocalizations as our study focused on the hooting behavior. Sites were selected in two ways: (1) using hooting survey data from the previous five years and (2) opportunistically as hooting birds were detected while conducting hooting surveys. Sooty Grouse have high site fidelity to territories with the same tree or patch of trees being occupied from year to year. Thus, we used previous survey data, including survey location, number of birds detected, and compass bearing and proximity to hooting birds were used to identify locations where birds were heard consistently in at least two of the past five years. We also selected locations based on accessibility, elevation, and distribution across the breeding range in Oregon, while avoiding streams and features that might interfere with the quality of recordings. All ARUs were placed at least 800 m apart to minimize recording the same male on two ARUs. If a male was actively hooting when the ARU was deployed, then the ARU was placed as close to the bird as possible. If hooting was not heard, the ARU was placed at the pre-selected location. Data Processing Hooting consists of a five to six syllable song, which is typically repeated one to 10 times per minute (Stirling and Bendell 1970). The hooting sound is relatively distinct from other bird songs because of its very low frequency (≤ 150 Hz, Hjorth 1970). We used the PNW-Cnet v4 with a shiny app graphical interface (Ruff et al. 2023) to automate the detection of hooting males in our recordings. PNW-Cnet uses a deep convolutional neural network to classify images and was developed for recognition of Northern Spotted Owl ( Strix occidentalis caurina ) vocalizations but has been trained to recognize the calls of 37 bird and mammal species in the Pacific Northwest, including Sooty Grouse (see Ruff et al. [2023] for information on software functionality and performance metrics). We set PNW-Cnet parameters to generate non-overlapping 12-second spectrogram segments with an upper frequency of 4000 Hz (to eliminate vocalizations that were not of interest). The application then generated class scores for each spectrogram segment and produced a file with the number of detections above a defined threshold for each recording (Ruff et al. 2023). Since this application was originally developed for automating the detection of owl vocalizations, we decided to evaluate the minimum threshold acceptable for Sooty Grouse by comparing the results for detection thresholds of score classes above 0.95, 0.75, 0.50, and 0.25. Determining which score class threshold to use is a balance between margin of error and detection probability. After comparing results from different score class thresholds, we choose to use a 0.25 score class threshold to be more inclusive of detected Sooty Grouse. We then manually verified 10,000 apparent Sooty Grouse detections with Kaleidoscope Pro 5 (Wildlife Acoustics, Inc., Maynard, Massachusetts, USA) and calculated precision as the number of true positives divided by apparent detection made by PNW-Cnet for a sample of 10,000 apparent detections. Since the application determines detections in 12-second segments, we acknowledge that the true number of hoots may be different than the number of detections identified by PNW-Cnet, because a hoot may be counted twice if split between two segments or two hoots may be counted once if they fall in the same 12-second segment. This was uncommon and occurred at random within our dataset; thus, it should have a minimal impact on our findings. Analysis and Model Selection We modeled both seasonal (across the spring season) and daily (relative to time since sunrise) hooting patterns with two separate generalized additive mixed models (GAMMs) using the mgcv package (Wood 2006) in Program R (R Core Team 2023). For both models, we used a negative binomial distribution and set k , the upper limit on the complexity of the spline basis for each smooth term, to 10. To model seasonal hooting patterns, we used the number of hoots detected per day as our response variable with an offset term for number of recording hours for that day. We assessed differences in peak hooting date among years by using day of year and an interaction between day of year and year. We looked at the influence of elevation, latitude, daily precipitation (Daymet; Thornton et al. 2022), minimum daily temperature (Daymet; Thornton et al. 2022), and study region on hooting frequency by including an interaction between day of year and the factor of interest. We included site as a random effect. To model daily hooting patterns, we used the number of hoots per 10-minute interval as the response variable. We were interested in how hooting rate changed within a day (from 5 am to 12 pm) and whether this varied among years. Additionally, we were interested in what influence time since sunrise and weather variables, daily precipitation and minimum daily temperature (Thornton et al. 2022), had on daily hooting patterns. For this analysis, we included random effects for site, study region, and day of year. For both seasonal and daily analyses, we used backward model selection, eliminating variables that were unprecise (p > 0.15) until all model parameters were precise. For weather variables, we also compared models with these variables included as fixed effects and models where these variables were in a spline. The model with the greatest amount of deviance explained was selected for each analysis. RESULTS On average, we deployed ARUs are 37 locations annually (range: 14–66). The median deployment date was 9 April in 2020, 13 April in 2021, 25 March in 2022, and 29 March in 2023. ARUs recorded a total of 54,839 hours, with more hours of recording in 2021 (22,345 hours) and 2022 (17,435 hours) when more ARUs were deployed compared to 2020 (7,018 hours) and 2023 (8,041 hours). Forty-three percent of the recording time was in the Coast Range, 35% in the West Cascades, 16% in Klamath, and 6% in the East Cascades study region. We successfully used PNW-Cnet to automate recognition of Sooty Grouse hoots. Our calculated precision for a sample of 10,000 apparent detections with a threshold for class score ≥0.25 that were manually reviewed was 0.9997 (see Fig. 2 for example spectrograms with varying class scores). Due to the high precision, we proceeded with our analysis using class scores ≥0.25 as our detection threshold. PNW-Cnet detected 2,667,534 hoots at this threshold. The number of hoots per hour of recording detected ranged from 44 hoots per hour in 2022 to 63 in 2020 with a relatively consistent number (mean = 47) of hoots per hour in 2021 and 2023 when a higher number of ARUs were deployed. The overall mean number of hoots per hour of recording was highest in the West Cascades (54 hoots per hour), followed by the Coast Range (49 hoots per hour), Klamath (47 hoots per hour), and East Cascades (19 hoots per hour). Our seasonal GAMM model describing hooting frequency included precipitation and four interactive terms with day of year including study region, year, latitude, and elevation. This model explained 18.2% of the deviance for the seasonal model. In our model selection process, the minimum daily temperature was not informative for describing seasonal hooting frequency in a spline or as a fixed effect. We determined that precipitation in a spline was a better fit than fixed effects (18.1% deviance explained compared to 18.2%, or increased deviance explained by 0.1%). Hooting frequency was low at the beginning and end of the breeding season with a peak hooting frequency between approximately 25–30 April (Figs. 3 and 4). This peak varied by study region but was similar for the Coast Range, East Cascades, and Klamath regions (range = 23–25 April) and was slightly later for the West Cascades (30 April; Fig. 3). This peak in hooting was also relatively consistent among years (Fig. 4). The effect of latitude interacting with day of year on predicted hooting frequency across the breeding season indicated a relatively consistent peak hooting date across latitudes. However, the lower latitudes had a higher magnitude in the peak number of hoots per hour (Fig. 5). The effect of elevation on date indicated peak hooting occurred at a similar date at most elevations, except at the highest elevations where the peak date range was wider (Fig. 5). The latitude and elevation results did not conform exactly to our expected relationships. As daily precipitation increased, the predicted number of hoots per hour steadily decreased (Fig. 6), as we anticipated. Our daily model describing hooting frequency included three interactions with year: time since sunrise, precipitation, and temperature. This model explained 22.3% of the deviance for the daily model. All predictor variables included in the model selection process were informative. We determined that precipitation and temperature in a spline were a better fit than as fixed effects (with fixed effects for these variables explaining 22% of the deviance). Since hooting frequency varied by year, daily results showed number of hoots in relation to time since sunrise (in minutes) for multiple dates between 1 April – 15 May for each of the four years. In general, hooting frequency rapidly increased from 30 minutes before sunrise to sunrise and then leveled out before slightly increasing six hours after sunrise. This same pattern was documented April to mid-May, but the number of hoots detected per 10 minutes varied in magnitude across the one and a half months (Fig. 7). The effect of precipitation indicated a lower number of hoots per 10 minutes as the amount of daily precipitation increased (Fig. 8). The effect of minimum daily temperature followed a different pattern with most hoots between approximately 3 – 15 degrees Celsius followed by a decline at higher temperatures (Fig. 8). DISCUSSION Proper timing of auditory bird surveys is important for monitoring population trends. Surveys that occur before or after the seasonal peak activity period may take place before all individuals have arrived at their breeding grounds . Morelli et al. (2022) found detection rates for different avian species changed throughout the day indicating inference from multi-species surveys needs to be carefully interpreted. This can result in low detectability creating variation and noise in the count data and result in biased abundance and trend estimates (Johnson and Krohn 2001, Nebel and McCaffery 2003, Harms and Dinsmore 2014). Bird detectability depends on many factors, including species life history and behavior, individual characteristics that vary by age and sex, environmental factors, survey methodology, and observer skill (Sólymos et al. 2018, Zamora-Marín et al. 2021). Single-species surveys are able to take into consideration the species’ ecology, which can lower bias and improve detectability. Sooty Grouse occur at relatively low densities in Oregon with a high proportion of zeros, or non-detections, during our hooting surveys (Walton and Cline 2023). PNW-Cnet was able to identify Sooty Grouse hooting with high precision (0.9997), meaning that 99.97% of the apparent detections made by PNW-Cnet that were verified, were actual Sooty Grouse hoots. Ruff et al. (2023) also calculated high precision (0.981), although at a threshold of ≥ 0.95. Recall can also gauge the detector’s accuracy, which is the number of true positives divided by the number of available positive examples. We did not calculate recall, but Ruff et al. (2023) had a recall of 0.572 at a 0.95 threshold. The precision and recall values reported by Ruff et al. (2023) at a 0.95 threshold indicate that apparent detections made by PNW-Cnet tend to be actual Sooty Grouse, but many actual detections are missed by the detector. By lowering the threshold, we suspect our recall was much higher than 0.572. As we lowered the threshold, we noticed hoots that were fainter, slightly obscured, and less distinct were included, with virtually no reduction in precision. As we hypothesized, hooting rate across the breeding season exhibited a unimodal (bell-shaped) pattern with a steep increase around early- to mid-April, peaking in late-April and then starting to decline in early-May, which aligns with males establishing breeding territories and becoming more vocal when most females are mating (Zwickel and Bendell 2004, Zwickel and Bendell 2020). Across the four years of ARU deployment, hooting peaked similarly in late-April, although a slightly later peak was observed in 2023. This later peak was likely due to measurable late-season snow and colder-than-normal temperatures in March and April of 2023 in western Oregon (USDA 2023). Differences in peak hooting across the four study regions was expected as geology, plant communities, and climate varies greatly north to south and west to east across this species’ range in Oregon. Hooting peaked earliest in the Coast Range and latest in the West Cascades study area. The Coast Range has milder winter temperatures with precipitation typically in the form of rain; whereas, the West Cascades, Klamath, and East Cascades have colder temperatures and at high elevations winter precipitation is primarily in the form of snow though early-April. These patterns in hooting frequency across the season and study regions can be used to inform study region-specific survey timing to increase the likelihood of detection. In addition, an observer is more likely to detect a Sooty Grouse if multiple hoots are heard during the 3-minute auditory survey (i.e., a male hooting only once or twice during a survey can easily be missed or the observer may not be confident a hoot was heard). Using predictions of mean number of hoots per 3-minutes can guide decision making regarding when surveys should be most effective. For example, if management wanted to survey when individual male Sooty Grouse were on average hooting three times per 3-minute, the duration of hooting surveys, then our results suggest the survey period should be approximately 15 April to 5 May compared to the original protocol’s survey dates of 1 April – 15 May (Figs. 3 and 4). Our assessment also allows managers to recognize which annual weather conditions could lead to low hooting rates; thereby, resulting in greater difficulty in detecting hooting males and potentially biasing counts low. Most bird surveys during the breeding season are conducted in the early morning. Vocal activity for many species of birds peaks early in the morning and then declines (Robbins 1981, Thompson et al 2017), as we expected to see with Sooty Grouse. BBS starts 30 minutes before sunrise and concludes by mid-morning (Sauer et al. 2013) to target this time of peak activity. Our hooting survey protocol was designed with the same assumption that hooting would be highest during early morning hours. Our analyses suggest the number of hoots per 10 minutes was steeply increasing 30 minutes before sunrise to sunrise, which indicates starting hooting surveys before sunrise compared to at sunrise may result in lower detection, especially early and late in the breeding season (Fig. 7). A similar daily pattern in hooting activity is followed throughout the season, although the number of hoots per 10 minutes is lower on 1 April and 15 May compared to the core of the breeding season from 10 April – 6 May. Based on the 10 April – 6 May daily results, Sooty Grouse continue to hoot at a high frequency well past our current morning survey window, which ends three hours (180 minutes) after sunrise. This was contrary to our hypothesis that Sooty Grouse hooting activity was predicted to plateau by mid-morning and exhibit a decline at the end of our recording timeframe—up to 400 minutes after sunrise (Fig. 7). Ideally, we would have set the ARUs to record into the afternoon to determine if and when during the day hooting rate decreases, but we had to compromise between an increase in recording time with battery life. We recommend expanding the time window when hooting surveys can be conducted each morning to six hours after sunrise, which will increase the efficiency of surveys as a large amount of staff time is spent travelling to and from survey routes. Our analyses verified that Sooty Grouse are likely underrepresented in BBS, solely from a mismatch in vocalization timing. BBS surveys are conducted for three-minutes, start 30 minutes before sunrise, and occur primarily in June (Ziolkowski et al. 2023), which our acoustic data shows as being well past the date of peak Sooty Grouse hooting. Peak hooting dates can be used in combination with other sources of information to better understand the breeding cycle of a species. Understanding the chronology of breeding and hatch which effects when broods break up later in the summer, can be used to make decisions on reducing the effects of human activities that cause disturbance and delaying hunting seasons until females and broods have dispersed and are less vulnerable. Average hatch date, calculated by looking at the replacement of primary feathers on hunter harvested Sooty Grouse wings, from the same time period (2020–2023) in western Oregon, indicated an average date of 12 June (range 6–19 June across the 4 years, ODFW unpublished data). A 5-year study using banded females at Comox Burn, British Columbia calculated peak copulation on 7 May, peak incubation the week of 14 May, and peak hatch the week of 11 June for adult Sooty Grouse (Zwinkel 1977). From what we learned about hooting activity in this study, as well as calculated hatch date from hunter harvested wings, Sooty Grouse in Oregon likely follow a similar timeline for breeding as in coastal British Columbia. ARUs provided valuable information for informing our survey protocol that would be difficult to get without having observers in the field for long periods of time. Other studies have looked at replacing point count auditory surveys with ARUs at each stop (such as Drake et al. 2021), but that was not our objective here as the effort to replace surveys with ARUs would be costly and time intensive to deploy and maintain ARUs at each survey stop. In addition, an observer can count the number of males heard at a stop, whereas an ARU typically provides presence/absence information, although this may change with advances in AI technology. We did not determine the detection radius of our ARUs, but we speculate it varies with topography, vegetation, and background noise. We have anecdotally noted that the detection radius for the ARUs appears to be smaller than for a human observer, which is supported by findings from other studies (Yip et al. 2017a, Darras et al. 2018). In summary, well-designed survey protocols are essential to effectively monitor wildlife species, particularly their population trends. However, many such protocols are not designed or evaluated with science-based information to ensure adequate and unbiased data has been collected to meet monitoring needs. Our study demonstrated how an assessment of a well-established survey protocol can confirm confidence in data quality related to the survey but also provide science-based information to improve the efficacy and efficiency of monitoring in the future. We recommend survey protocols be rigorously evaluated whenever possible to ensure the best inference. As a bonus, these evaluations may lead to learning new information about the ecology of the species being monitored. REFERENCES Bibby, C. J., D. A. Hill, N. D. Burgess, and S. Mustoe. 2000. Bird Census Techniques, 2nd Ed. London: Academic Press. 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 Darras, K., P. Batáry, B. Furnas, A. Celis‐Murillo, S. L. V. Wilgenburg, Y. A. Mulyani, and T. Tscharntke. 2018. Comparing the sampling performance of sound recorders versus point counts in bird surveys: a meta‐analysis. Journal of Applied Ecology 55:2575–2586. https://doi.org/10.1111/1365-2664.13229 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 4:675–683. https://doi.org/10.1111/2041-210X.12060 Doerr, J. G, C. L. Barescu, J. M. Brighenti Jr., and M. P. Morin. 1984. Use of clear-cutting and old-growth forests by male blue grouse in central southeast Alaska. Pp 309–313 in Meehan, W.R. T. H. Merell Jr., and T.H. Hanley (eds). Fish and wildlife relationships in old-growth forests, a symposium. American Institute of Fishery Research Biologists. Drake A., D. R. de Zwaan, T. A. Altamirano, S. Wilson, K. Hick, C. Bravo, J. T. Ibarra, and K. Martin. 2001. Combining point counts and autonomous recording units improves avian survey efficacy across elevational gradients on two continents. Ecology and Evolution. 2021:8654–8682. https://doi.org/10.1002/ece3.7678 Farnsworth, G. L., K. H. Pollock, J. D. Nichols, T. R. Simons, J. E. Hines, and J. R. Sauer. 2002. A removal model for estimating detection probabilities from point-count surveys. Auk 119:414–425. Fox, T. T., B. D. Dugger, and K. M. Dugger. 2011. Survey protocol for monitoring the relative abundance of sooty grouse ( Dendragapus fuliginosus fuliginosus) populations in western Oregon. Oregon State University, Corvallis, Oregon. 111 pp. Frey, S. J. K., K. M. Walton, M. L. Cline, L. Sanders, and J. B. Dinkins. In review . Abundance and occupancy trends of Sooty Grouse in western Oregon: determining best modeling practices by comparing observed and simulated data. Authorea . April 23, 2025 Halofsky, J. E., D. L. Peterson, and B. J. Harvey. 2020. Changing wildfire, changing forests: the effects of climate change on fire regimes and vegetation in the Pacific Northwest, USA. Fire Ecology 16:4. Harms, T. M., and S. J. Dinsmore. 2014. Influence of season and time of day on marsh bird detections. Wilson Journal of Ornithology 126:30–38. Johnson, C. M., and W. B. Krohn. 2001. The importance of survey timing in monitoring breeding seabird numbers. Waterbirds: The International Journal of Waterbird Biology 24:22–33. Knight, E. C., A. C. Smith, R. M. Brigham, and E. M. Bayne. 2021. 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 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 122:1–20. https://doi.org/10.1093/condor/duaa016 Martin, K., N. Koper, and R. Bazin. 2014. Optimizing repeat visit, call-broadcast nocturnal surveys for Yellow Rails (Coturnicops noveboracensis). Waterbirds 37:68–78. https://doi.org/10.1675/063.037.0109 Morelli, F., V. Brlik, Y. Benedetti, R. Bussière, L. Moudrá, J. Reif, and M. Svitok. 2022. Detection rate of bird species and what it depends on: tips for field surveys. Frontiers in Ecology and Evolution 9:671492. Nebel, S., and B. J. McCaffery. 2003. Vocalization activity of breeding shorebirds: documentation of its seasonal decline and application for breeding bird surveys. Canadian Journal of Zoology 81:1702–1708. Niederleitner, J. F. 1987. Use of early successional, midsuccessional, and old-growth forests by breeding blue grouse ( Dendragapus obscurus fuliginosus ) on Hardwicke Island, British Columbia. Canadian Journal of Zoology 65:151–155. North American Bird Conservation Initiative, U.S. Committee. 2009. The State of the Birds, United States of America, 2009. U.S. Department of Interior: Washington, DC. 36 pp. Oregon Department of Fish and Wildlife (ODFW). 1997. Game Bird Hunting in Oregon. Priyadarshani, N., S. Marsland, and I. Castro. 2018. Automated birdsong recognition in complex acoustic environments: a review. Journal of Avian Biology 49:01447. https://doi.org/10.1111/jav.01447 R Core Team. 2023. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. . Redfield, J. A. 1973. Demography and genetics in colonizing populations of blue grouse ( Dendragapus obscurus ). Evolution 27:576–592. Rempel, R. S., C. M. Francis, J. N. Robinson, and M. Campbell. 2013. Comparison of audio recording system performance for detecting and monitoring songbirds. Journal of Field Ornithology 84:86–97. http://dx.doi.org/10.1111/jofo.12008 Robbins, C. S. 1981. Effect of time of day on bird activity. Pages 275–286 in J. C. Ralph and M. J. Scott, editors. Studies in avian biology. Allen, Lawrence, Kansas, USA. Robbins, C. S., D. Bystrak, and P. H. Geissler. 1986. The Breeding Bird Survey: Its first fifteen years, 1965–1979. U.S. Fish and Wildlife Service Resource Publication 157. Rosenberg, K.V., D. Pashley, B. Andres, P. J. Blancher, G.S. Butcher, W.C. Hunter, D. Mehlman, A.O. Panjabi, M. Parr, G. Wallace, and D. Wiedenfeld. 2014. The State of the Birds 2014 Watch List. North American Bird Conservation Initiative, U.S. Committee. Washington, D.C. 4 pages. Ruff Z. J., D. B. Lesmeister, J. M. A. Jenkins, and C. M. Sullivan. 2023. PNW-Cnet v4: Automated species identification for passive acoustic monitoring. SoftwareX 23. https://doi.org/10.1016/j.softx.2023.101473 Sauer, J. R., K. L. Pardieck, D. J. Ziolkowski, Jr., A. C. Smith, M. R. Hudson, V. Rodriguez, H. Berlanga, D. K. Niven, and W. A. Link. 2017. The first 50 years of the North American Breeding Bird Survey. The Condor: Ornithological Applications 119:576–593. Sauer, J.R., W. A. Link, J.E. Fallon, K.L Pardieck, and D.J. Ziolkowski Jr. 2013. The North American Breeding Bird Survey 1966-2011: Summary Analysis and Species Accounts. North American Fauna 79:1–32. doi:10.3996/nafa.79.0001 Shonfield, J., and E. M. Bayne. 2017. Autonomous recording units in avian ecological research: current use and future applications. Avian Conservation and Ecology 12:14. Sólymos, P., S. M. Matsuoka, D. Stralberg, N .K. S. Barker, and E. M. Bayne. 2018. Phylogeny and species traits predict bird detectability. Ecography 41:1595–1603. doi: 10.1111/ecog.03415 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 10:368–380. https://doi.org/10.1111/2041-210X.13103 Thompson, S. J., C. M. Handel, and L. B. McNew. 2017. Autonomous acoustic recorders reveal complex patterns in avian detection probability. Journal of Wildlife Management 81:1228–1241. http://dx.doi.org/10.1002/jwmg. 21285 Thornton, M. M., R. Shrestha, Y. Wei, P. E. Thornton, and S-C. Kao. 2022. Daymet: Daily Surface Weather Data on a 1-km Grid for North America, Version 4 R1. ORNL DAAC, Oak Ridge, Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/2129 Turgeon, P. J., S. L. Van Wilgenburg, and K. L. Drake. 2017. Microphone variability and degradation: implications for monitoring programs employing autonomous recording units. Avian Conservation and Ecology 12:9. http://dx.doi.org/10.5751/ ACE-00958-120109 U.S. Department of Agriculture, Forest Service; U.S. Department of the Interior, Bureau of Land Management [USDA and USDI]. 1994. Final supplemental environmental impact statement on management of habitat for late-successional and old-growth forest related species within the range of the northern spotted owl. Portland, Oregon. U.S. Department of Agriculture (USDA), Natural Resources Conservation Services. 2023. Oregon water supply outlook report, May 1, 2023. Portland, Oregon. 36 pp. Walton, K. M. and M.L. Cline. 2023. Summary of 2023 Western Oregon Sooty Grouse Surveys. Oregon Department of Fish and Wildlife, Salem, Oregon. 10 pp. Welch, J., M. Pope, and J. A. Crawford. 2002. Comparative Ecology of Blue and Ruffed Grouse in Southwest Oregon. Annual Report January 2002. Oregon State University, Corvallis, Oregon. 29 pp. Winiarska, D., P. Szymański, and T. S. Osiejuk. 2024. Detection ranges of forest bird vocalisations: guidelines for passive acoustic monitoring. Scientific Reports 14, 894. https://doi.org/10.1038/s41598-024-51297-z Wood, S. N. 2006. Generalized additive models: an introduction with R. Chapman and Hall, Boca Raton, FL. Yip, D. A., E. M. Bayne, P. Sólymos, J. Campbell, and D. Proppe. 2017a. Sound attenuation in forested and roadside environments: implications for avian point count surveys. Condor 119:73–84. http://dx.doi.org/10.1650/CONDOR-16-93.1 Yip, D. A., L. Leston, E. M. Bayne, P. Sólymos, and A. Grover. 2017b. Experimentally derived detection distances from audio recordings and human observers enable integrated analysis of point count data. Avian Conservation and Ecology 12:11 Zamora-Marín, J. M., A. Zamora-López, J. F. Calvo, and F. J. Oliva-Paterna. 2021. Comparing detectability patterns of bird species using multi-method occupancy modelling. Scientific Reports. 11:2558. doi: 10.1038/s41598-021-81605-w Ziolkowski, D.J., M. Lutmerding, W. B. English, V. I. Aponte, and M-A.R. Hudson. 2023. North American Breeding Bird Survey Dataset 1966–2022: U.S. Geological Survey data release, https://doi.org/10.5066/P9GS9K64. 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 26:1034–1050. https://doi.org/10.1111/ddi.13075 Zwickel, F. C. 1997. Local variations in timing of breeding of female blue grouse. Condor 79: 185-191. Zwickel, F. C. and J. F. Bendell. 1972. Blue grouse, habitat, and populations. International Ornithnol Congr, Proc 15:150–169. Zwickel, F. C. and J. F. Bendell. 2004. Blue Grouse: Their Biology and Natural History. NRC Research Press, Ottawa, Ontario, Canada. 284 pp. Zwickel, F. C. and J. F. Bendell. 2020. Sooty Grouse ( Dendragapus fuliginosus ), version 1.0. In Birds of the World, P. G. Rodewald, editor. Cornell Lab of Ornithology, Ithaca, New York, USA. Figure 1. Study area map in western Oregon showing locations of Sooty Grouse hooting survey routes and locations of autonomous recording units (ARUs) during the 2020–2023 breeding seasons. ARUs were deployed along hooting survey routes in all four study regions throughout western Oregon, USA. Figure 2. Four examples of spectrograms showing Sooty Grouse hoots and assigned class scores within 12-second segments recorded in western Oregon. Detections with class scores at a threshold of 0.25 or greater were used in our models. Example spectrograms include (a) a hoot from farther away with noise interference (b) a closer hoot with noise interference (c) a close proximity 5-syllable hoot, and (d) a close proximity 6-syllable hoot. Figure 3. Effect of day of year on the predicted seasonal hooting frequency stratified by study region using a generalized additive mixed model with a random site effect. Seasonal hooting frequency was defined as the number of hoots per recording hour. All other predictor variables were held at mean values. Horizontal dashed lines represent the average hoots per hour that would result in 1, 2, or 3 hoots occurring during a 3-minute in-person Sooty Grouse hooting survey. Data collected 2020–2023 in western Oregon. Figure 4. Effect of day of year on the predicted seasonal hooting frequency stratified by year using a generalized additive mixed model with a random site effect. Seasonal hooting frequency was defined as the number of hoots per recording hour. All other predictor variables were held at their mean values. Horizontal dashed lines represent the average hoots per hour that would result in 1, 2, or 3 hoots occurring during a 3-minute in-person Sooty Grouse hooting survey. Data collected 2020–2023 in western Oregon. Figure 5. Effects of latitude and elevation interacting with day of year on the predicted seasonal hooting frequency of Sooty Grouse using a generalized additive mixed model with a random site effect. All other predictor variables were held at their mean values. Day of year interaction with latitude represented as a (a) three-dimensional and (b) contour with contours depicting hoots per hour. Day of year interaction with elevation represented as a (c) three-dimensional (d) and contour with contours depicting hoots per hour. Seasonal hooting frequency was defined as the number of hoots per recording hour. Data collected 2020–2023 in western Oregon. Figure 6. Effect of daily precipitation on the predicted seasonal hooting frequency from a generalized additive mixed model with a random site effect. Seasonal hooting frequency was defined as the number of hoots per recording hour. All other predictor variables were held at their mean values. Data collected 2020–2023 in western Oregon. Figure 7. Effect of time since sunrise on the predicted daily hooting frequency stratified by year using a generalized additive mixed model with random effects of site, region, and day of year. Daily hooting frequency was defined as the number of hoots per ten-minute interval. All other predictor variables were held at their mean values. Current in-person Sooty Grouse hooting survey protocols are constrained between 30 minutes before sunrise and 3 hours after sunrise, as shown by the dashed lines in each panel. Dates correspond to key dates: original start to in-person surveys (1 April), modified start to in-person surveys (10 April), approximated peak hooting date from seasonal analysis (27 April), and end of in-person surveys (15 May). Data collected 2020–2023 in western Oregon. Figure 8. Effect of (a) daily precipitation in mm and (b) minimum daily temperature in degrees C on the predicted daily hooting frequency from a generalized additive mixed model with random effects of site, region, and day of year. Daily hooting frequency was defined as the number of hoots per 10-minute interval. All other predictor variables were held at their mean values. Data collected 2020–2023 in western Oregon. Information & Authors Information Version history V1 Version 1 28 April 2025 Peer review timeline Published Wildlife Biology Version of Record 30 Nov 2025 Published Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Wildlife Biology Keywords autonomous recording unit dendragapus fuliginosus pnw-cnet sooty grouse Authors Affiliations Kelly Walton 0009-0003-4851-4692 [email protected] Oregon Department of Fish and Wildlife View all articles by this author Sarah Frey 0000-0002-4343-0700 Oregon State University View all articles by this author Cara Christensen 0009-0006-1629-9416 Oregon State University View all articles by this author Mikal Cline Oregon Department of Fish and Wildlife View all articles by this author Lauren Gramberg Oregon State University View all articles by this author Jonathan Dinkins Oregon State University View all articles by this author Metrics & Citations Metrics Article Usage 579 views 265 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Kelly Walton, Sarah Frey, Cara Christensen, et al. Passive acoustic monitoring with AI-based detection and identification reveal Sooty Grouse hooting patterns in western Oregon. Authorea . 28 April 2025. DOI: https://doi.org/10.22541/au.174585482.29885531/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.174585482.29885531/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ff157f93ce48650',t:'MTc3OTM0MzA0Ng=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();