Abundance and occupancy trends of Sooty Grouse in western Oregon: determining best modeling practices by comparing observed and simulated data

Abundance and occupancy trends of Sooty Grouse in western Oregon: determining best modeling practices by comparing observed and simulated data | 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 Search Search ADVANCED SEARCH SCROLL Wildlife Biology This is a preprint and has not been peer reviewed.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. 23 April 2025 V1 Latest version Share on Abundance and occupancy trends of Sooty Grouse in western Oregon: determining best modeling practices by comparing observed and simulated data Authors : Sarah Frey 0000-0002-4343-0700 [email protected] , Kelly Walton 0009-0003-4851-4692 , Mikal Cline , Lindsey Sanders , and Jonathan Dinkins Authors Info & Affiliations https://doi.org/10.22541/au.174540808.83675545/v1 Published Wildlife Biology Version of record Peer review timeline 499 views 211 downloads Contents Abstract Supplementary Material Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract [1]¿p#1 Sooty Grouse (Dendragapus fuliginosus) is a large game bird that occupies montane forests in the Pacific Northwest, USA. These forests have been disturbed by human activities, which has been documented to have positive and negative impacts on populations. The North American Breeding Bird Survey (BBS) indicated population declines for this species across their range (-1.2 % per year [95% CI = -3.0 – 0.25 % per year], 1966 – 2022). However, Sooty Grouse is inadequately represented along BBS routes due to little overlap with habitat, survey timing, low population density, and low detectability. We developed a monitoring protocol specialized for Sooty Grouse to better evaluate population trends for this species. We surveyed Sooty Grouse from 2011 – 2024 along 119 10- to 20-km survey routes across western Oregon. We estimated abundance and occupancy trends utilizing hierarchical models that simultaneously address the observation and ecological processes of monitoring wildlife populations. We did this in five common modeling frameworks, including exponential growth, Poisson linear regression, and logistic regression using JAGS, ubms, and unmarked (program R). Trend estimates varied across approaches, Poisson linear regression models displayed the most precise trend estimates, indicating that Sooty Grouse populations declined 2.9 % (95% CI = 1.4 – 4.5 %) annually over the span of our study. Given differences among frameworks, we simulated data to test which provided the most accurate trend estimate. Occupancy models did not perform well estimating trends on simulated data, whereas abundance models yielded robust results, particularly when the dataset did not contain missing data. Detection probability also varied across models, with occupancy models producing higher estimates (mean = 0.84) than abundance models (mean = 0.46). Our results confirm Sooty Grouse trends have declined in the recent past and warrant a more detailed assessment to determine what factors are driving this pattern. Abundance and occupancy trends of Sooty Grouse in western Oregon: determining best modeling practices by comparing observed and simulated data ABSTRACT Sooty Grouse ( Dendragapus fuliginosus ) is a large game bird that occupies montane forests in the Pacific Northwest, USA. These forests have been disturbed by human activities, which has been documented to have positive and negative impacts on populations. The North American Breeding Bird Survey (BBS) indicated population declines for this species across their range (-1.2 % per year [95% CI = -3.0 – 0.25 % per year], 1966 – 2022). However, Sooty Grouse is inadequately represented along BBS routes due to little overlap with habitat, survey timing, low population density, and low detectability. We developed a monitoring protocol specialized for Sooty Grouse to better evaluate population trends for this species. We surveyed Sooty Grouse from 2011 – 2024 along 119 10- to 20-km survey routes across western Oregon. We estimated abundance and occupancy trends utilizing hierarchical models that simultaneously address the observation and ecological processes of monitoring wildlife populations. We did this in five common modeling frameworks, including exponential growth, Poisson linear regression, and logistic regression using JAGS, ubms, and unmarked (program R). Trend estimates varied across approaches, Poisson linear regression models displayed the most precise trend estimates, indicating that Sooty Grouse populations declined 2.9 % (95% CI = 1.4 – 4.5 %) annually over the span of our study. Given differences among frameworks, we simulated data to test which provided the most accurate trend estimate. Occupancy models did not perform well estimating trends on simulated data, whereas abundance models yielded robust results, particularly when the dataset did not contain missing data. Detection probability also varied across models, with occupancy models producing higher estimates (mean = 0.84) than abundance models (mean = 0.46). Our results confirm Sooty Grouse trends have declined in the recent past and warrant a more detailed assessment to determine what factors are driving this pattern. Keywords : abundance, hierarchical models, occupancy, simulations, Sooty Grouse, population trends INTRODUCTION Accurate estimation of population trends are foundational components of conservation and management of wildlife (Rosenberg et al. 2019). Trends rely on abundance or occupancy quantified across space and time with raw data collected as annually repeated count or detection/non-detection surveys (Royle 2004, MacKenzie et al. 2006, Royle and Dorazio 2008). Without reliable knowledge of these metrics, managers are unable to detect population changes and subsequently identify threats to populations. Population fluctuations are typically quantified via distributional (spatial) and temporal changes. Counted animals within these surveys are typically unmarked, which leads to observer detection biases (Gibson 2011, Kéry 2011, Veech et al. 2016). Although all surveys have some level of imperfect detections, these errors can be reduced with a well-designed survey and by using robust modeling techniques (Dahlgren et al. 2021). Recent advances in hierarchical modeling and standardization of annually repeated counts allows for the parallel estimation of the ecological processes (annual abundance or occupancy) and observation processes (detection probability, Royle and Dorazio 2008). Accounting for imperfect detection when estimating abundance, occupancy, and trends can reduce bias from observation error (Hostetler and Chandler 2015, Veech et al. 2016, Monroe et al. 2019). There are many broad-scale monitoring programs in place to track avian populations in North America. One of the largest and most well-known monitoring programs is the North American Breeding Bird Survey (BBS), where every bird seen or heard at each stop along a survey route is recorded. While the BBS aims to be all-encompassing and include as many species as possible, it is not a practical or appropriate means of surveying for all species. Some species are better surveyed with specialized survey techniques, either because timing of peak vocalizations is outside the BBS survey window, or their habitat is inaccessible or under-surveyed on BBS routes (Sauer et al. 2013, Hoerner 2018, Itescu and Meiri 2021). This is the case for many upland game bird species, which need accurate trend estimates for harvest management but often have low encounter rates on BBS routes. Specialized surveys that are commonly used for upland game birds include spring male vocalization counts, lek counts for lekking grouse species, drumming surveys for Ruffed Grouse ( Bonasa umbellus ), summer brood counts along driving routes, surveys of hunter effort and harvest success, and parts-collection surveys to identify age, sex, and wing molt patterns of harvested individuals (Dahlgren et al. 2018). Data from these surveys can then be used to estimate trends as well as reproductive success and harvest rates to give a better understanding of population dynamics. Sooty Grouse ( Dendragapus fuliginosus ) is a cryptic upland game species that occur in conifer forests of western North America from sea level to the subalpine zone (Zwickel and Bendell 2004). Males display and vocalize during the breeding season with a series of low-frequency hoots to attract females. In 2014, the North American Bird Conservation Initiative placed Sooty Grouse on their State of the Birds Watch List as a species at risk of becoming threatened or endangered without conservation action, giving the species a “Concern Score” of 14 out of 20 (Rosenberg et al. 2014, North American Bird Conservation Initiative 2016). This score was based on a relatively small global breeding population size, restricted distribution, threats to their habitat, and declining population trends. Understanding the magnitude of population decline and the environmental and human-caused factors contributing to these declines is critical for the conservation and continued management of this species. Threats to forest grouse include habitat change and loss from anthropogenic development, logging activities, wildfire and change in wildfire regimes, and changes in forest succession due to forest management practices (Storch 2000). Previous efforts to assess trends for Sooty Grouse have come from localized studies (summarized in Zwickel and Bendell 2004), and as part of upland game bird brood counts, harvest and parts collection surveys conducted by some state wildlife agencies. Brood counts are commonly used to monitor upland game bird abundance and productivity by counting the number of adults and chicks seen along driving routes in the late summer, but the cryptic and arboreal nature of Sooty Grouse results in a low encounter rate on these surveys. Upland game bird harvest often follows population trends, with higher harvest reported when population levels are up; however, the exact correlation is not known (Dahlgren et al. 2018). In addition, many other factors can influence harvest, such as variation in hunting access, hunter effort, and the number of active hunters. The little that is known about the status of Sooty Grouse populations is mainly from the BBS, which indicates the population is declining range-wide (-1.2 % per year [95% CI = -3.0 – 0.25 % per year, 1966 – 2019), and even more so in Oregon (-1.56 % per year) however the estimates had 95% CIs that overlap 0 (-3.63 – 0.41 % per year; Sauer et al. 2020). It is unclear how well this population trend adequately represents Sooty Grouse population status, given their high-elevation habitat, low densities, and low detectability on BBS routes. As a result, encounter rate in existing landscape-scale avian surveys is insufficient for Sooty Grouse population monitoring, which is necessary for science-based management. In Oregon, Sooty Grouse is managed as a game bird species with a liberal hunting season from 1 September through 31 January and a daily bag limit of three Blue Grouse (e.g., three total Dusky ( Dendragapus obscurus ) or Sooty Grouse, Oregon Department of Fish and Wildlife [ODFW] 2024). Given concerns about the population status of this species, and the unsuitability of existing survey methodologies, ODFW implemented a targeted survey protocol in 2011 to assess the current state of the Sooty Grouse population in western Oregon (Fox et al. 2011). Our primary objective was to use a targeted species-specific survey combined with hierarchical modeling techniques to estimate Sooty Grouse population trends, which could be used to develop and inform management strategies in Oregon. To meet this objective, we compared trends in abundance and occupancy from five different hierarchical modeling frameworks, including exponential growth, Poisson linear regression, and logistic regression in JAGS, ubms, and unmarked in program R. We then used simulations to determine which approach was most accurate for estimating Sooty Grouse population trends. Ideally trend estimates would be consistent across modeling frameworks and not sensitive to the statistical approach. METHODS Study area From 2011–2024, ODFW conducted Sooty Grouse hooting surveys throughout western Oregon; surveys ranged between 275 – 1,225 m. a. s. l. (Fig. 1), which we used survey data for our analyses. The study area included the Coast Range, Cascade Range, and the Klamath regions (Fig. 1). ODFW constrained their study area to montane forests within the known Sooty Grouse distribution. Variability in forested landscapes in western Oregon has resulted from historical disturbances such as timber harvest, fire, and anthropogenic development. The landscape was shaped by timber harvest, which has resulted in many distinct blocks of even-aged forest ranging from early-seral to mature forests. Only small patches of remnant old-growth forest remain on the landscape. The climate in western Oregon is highly influenced by moist air from the Pacific Ocean moving inland across the Coast and Cascade Ranges. Forests closer to the Coast tend to have a more maritime climate with mild temperatures and high levels of precipitation falling as rain. At mid to high elevations in the Cascade Range winters are cold with high levels of snow and relatively warm, dry summers. Field survey methods Surveys routes were stratified by study region (Coast Range, Klamath, West Cascades, and East Cascades) and then randomly placed within suitable forested habitat inside the study area. Survey starting points were located on the nearest drivable road to the random survey placement, as surveys required driving access. Snow often prevents access to high elevation sites in the Cascades in spring; thus, an upper elevational bound was placed on the surveys to limit routes to areas accessible in most years (1,225 m. a. s. l.). Routes were established to be variable in length from year to year depending on access, but up to 15-km long. We conducted surveys approximately every 800 m along each route, resulting in 15 – 25 survey stops per route. Route length was intended to maximize the survey area given road accessibility. Surveys were conducted in the early morning from 30 minutes before sunrise to three hours after sunrise from 1 April – 31 May. At each survey stop along a route, we conducted a 3-minute listening survey for hooting male Sooty Grouse. Surveys were generally conducted from low to high elevation along a route due to logistical constraints, resulting in high elevation sites rarely getting surveyed early in the morning. Finally, since these surveys rely on auditory detections, primarily hooting males were counted, with visual encounters being rare. Starting in 2019, some routes were repeat surveyed one to five times throughout the season to allow for estimation of detection probability within a year. For additional details on the survey methodology see Cline et al. (2024). Clustering survey stops The exact location of any given survey location on a route varied slightly within and between years. This variation posed an issue because the models we used require consistency in spatial location for repeated visits to sites. To rectify this, we clustered survey stop locations within a year (for repeat visits) and then between years. Our goal was to identify unique center locations (for each cluster of survey stops) approximately 400 m apart along a route that represented the same small areas across years and within year repeat surveys. First, we clustered within a year to account for repeat surveys in a single survey season, then across all years resulting in one spatial location (represented by one set of coordinates) for each survey point. We used the hclust () command in the R package stats in base R (R Core Team 2023) employing the ‘average’ method of clustering. This method measures the pairwise differences between two clusters and then averages them and allows for a distance threshold to be set for this pairwise difference. For the within-year clusters, we used a distance threshold of 525 m. Next, we clustered stops from all years and assigned them to the centroids of these clusters. We used a threshold of 700 m after comparing spacing at threshold settings from 525 – 700 m (see Figs. S1 and S2 for examples of clustering of points by threshold on a curvy and linear route). The centroid locations developed through this clustering method were adjusted slightly to the nearest location a vehicle could safely be stopped and to avoid areas with high background noise level that would interfere with surveys (i.e., close to a stream or culvert). The centroid location was then defined as the unique permanent survey point location to be used on hooting surveys conducted in the future. When multiple survey stops fell within a cluster, we used the maximum Sooty Grouse count and associated survey information, if surveys had the same counts, then we randomly selected one. Statistical analyses We estimated Sooty Grouse population trends using hierarchical models that account for imperfect detection when estimating abundance or occupancy dynamics. We compared trend estimates (and 95% CIs), measured as lambda, the finite rate of increase per unit time (in this case per year) from the following modeling frameworks: The Dail-Madsen model with exponential growth (EGM). This is a dynamic state-space model that describes exponential growth in an N-mixture framework (Dail and Madsen 2011, Hosteler and Chandler 2015). It models latent abundance as a Poisson process and changes in abundance are determined by recruitment and survival. The observed counts are linked through a binomial detection process. To run this model we used pcountOpen () in the R package unmarked (Fiske and Chandler 2011) and selected the ‘trend’ dynamics option, which only models recruitment as a trend parameter (i.e., growth rate). For site i at time t , latent abundance is modeled as a Poisson process: N i,t ~ Poisson( µ i,t ) (1) where the expected abundance is specified using a log-linear regression: log (µ i,t ) = β 0 + t * log( λ ) (2) Likewise, the exponential growth process can be described recursively as: N i,t = N i,t-1 *λ (3) Where λ is the growth rate, also known as the finite rate of increase. Variability in the observation process is modeled with the detection probability parameter ( p i,t,j ). The observed count C i,t,j from replicate j at site i in year t is modeled via: C i,t,j ∼ Binomial( N i,t , p i,t,j ) (4) with the detection probability modeled using a logistic regression: logit( p i,t,j ) = β 0 + β 1 *X i,t,j (5) where β 1 is the effect of the covariate X i,t,j on detection. Here X i,t,j is the number of stops completed along a route during each survey. Random effects are not supported in pcountOpen () and were therefore not included. Maximum likelihood is used to estimate parameter values in this framework. Binomial N-mixture model with Poisson linear regression (PLR) . The open-population Binomial N-mixture model is a hierarchical model where the year effect on abundance is the trend via Poisson linear regression (Kery et al. 2009, Kery and Royle 2010). As with the EGM, latent abundance at site i in year t is Poisson distributed with mean µ i,t N i,t ~ Poisson( µ i,t ) (6) where the trend in abundance is described by a Poisson linear regression: log( µ i,t ) = β 0 + β 1 *year + η i (7) To account for similarity between sites, we included a site random effect η i in the abundance model. The observed counts ( C i,t,j ) are modeled as: C i,t,j ∼ Binomial( N i,t , p i,t,j ) (8) where the detection probability is modeled via logistic regression: logit( p i,t,j ) = β 0 + β 1 *X i,t,j + θ i,j,t (9) where β 1 is the effect of the covariate X i,t,j on detection. As with the EGM, X i,t,j is the number of stops completed along a route during each survey. Here, we included an observation random effect term ( θ i,j,t ) in the detection model to account for any additional variability in the detection process not captured by X i,t,j . We ran the PLR in two modeling frameworks that employ Bayesian estimation methods. The first was the R package ubms (unmarked Bayesian models; Kellner et al. 2021), which was developed to be similar to the unmarked package but provides a user-friendly way to use Bayesian methods for estimating abundance and occupancy models. The second was the JAGS framework where models were specified in the JAGS language and then run in conjunction with the R packages runjags (Denwood 2016), rjags (Plummer 2024), and coda (Plummer et al. 2006). This framework provides more flexibility in modifying the model structure but also requires more expertise to do so appropriately. Occupancy trend model with logistic regression (OTM). This model uses detection/non-detection data to estimate occupancy (and its trend via a linear year parameter via logistic regression) and detection in a double Bernoulli process (Kery and Royle 2021). Occupancy is a binary outcome and the true occupancy state for site i in year t is modeled as a Bernoulli process: z i,t,j ∼ Bernoulli( ψ i,t ) (10) where the trend in occupancy probability ( ψ i,t ) follows a logistic regression: logit( ψ i,t ) = β 0 + β 1 *year + η i (11) where the effect of year ( β 1 ) is the occupancy trend (lambda) and η i is a site-specific random effect capturing heterogeneity among sites. Variability in the observation process is modeled with the detection probability parameter. The observed detections y i,t,j from replicate j at site i in year t is modeled via: y i,t,j ∼ Bernoulli( z i,t * p i,t,j ) (12) where z i,t is the latent occupancy state. Detection probability is modeled with logistic regression: logit( p i,t,j ) = β 0 + β 1 *X i,t,j (13) where β 1 represents the effect of the covariate X i,t,j on detection. Here X i,t,j expresses the number of stops completed along a route during each survey. We considered an observation random effect for this model. As with the PLR model, we ran this model in the two modeling frameworks: a) JAGS and b) ubms as described above. We assessed fit for all models as overdispersion in the data has been found to bias abundance estimates in N-mixture models (Knape et al. 2018). See Appendices S1 and S2 for the model structures in JAGS language and model fit methods and results. Simulations To identify the best model for our system, we conducted simulations to determine which of our trend estimates most accurately reflected patterns and dynamics of Sooty Grouse populations. We simulated abundance and occupancy data based on average parameter estimates from the models. The only parameter that varied was the trend estimate. The trend levels we used for the simulations were -3.0, 0.0 (no trend), and 3.0 % change per year. We used the same model frameworks to estimate both the abundance and occupancy trends on the simulated data for each of the trend levels. We did two runs in each of the model frameworks for each trend level, one with a dataset without any missing data and one with missing data that matched the real dataset. RESULTS During 2011–2024, a total of 119 transects were surveyed, 98 routes were surveyed in more than one year, and six routes were surveyed in every year of this study (however, this number increased to 12 when 2011, the pilot year, was not included). Over this period, we surveyed a total of 1,047 routes (31 – 124 per year) comprising 16,696 3-min surveys and 3,415 Sooty Grouse detections. Each year an average of 1,192.6 (445 – 2,006) 3-min surveys and 243.9 (93 – 442) Sooty Grouse detections were made per year (Table 1). Trend estimates varied across modeling frameworks (Fig. 2). The PLR models revealed identical trend estimates, showing a 2.9 % annual decline in Sooty Grouse across this study period with high precision (95% CIs = -4.5 – -1.4 %). Trends from the OTMs also suggested large population declines, though with low precision as 95% CIs overlapped zero (-6.9 % [95% CIs = -14.8 – 0.8 %]). Due to the wide confidence intervals that overlapped zero, no trend could be inferred with the OTMs. Similarly, the trend estimate from the EGM indicated no trend, but in contrast the precision was high (-0.6 % [95% CIs = -2.9 – 1.8 %]). The yearly abundance estimates for the PLR models were precise (Fig. 3a-b), particularly compared to the EGM (Fig. 3c). The mean raw yearly counts were lower and more variable as they do not account for imperfect detection and the abundance estimates from the models follow the trend line (Fig. 3). Detection probability estimates varied widely across models (range = 0.12 – 0.89, Fig. 3a). The lowest detection estimates were produced using ubms to run the PLR (0.12 [95% CIs = 0.07 – 0.2]), whereas moderate detection estimates were produced for the same model using JAGS (0.4 [95% CIs = 0.27 – 0.52]). The OTMs produced the highest detection estimates: 0.89 (95% CIs = 0.86 – 0.93) and 0.78 (95% CIs = 0.65 – 0.88), with JAGS and ubms, respectively. All estimates had very narrow CIs, indicating high precision for this parameter across all models. In all the model frameworks, detection probability increased with the number of 3-min survey completed per route (Fig. 3b). The occupancy models performed poorly overall in the simulation tests (Fig. 5). The no trend simulations accurately tracked no-trend; however, simulations predicted trends much higher or lower than the known high (3.0%) and low (-3.0%) trend in the simulated data, which was amplified when missing data was included in the simulations (Fig.5). All abundance models were fairly accurate at estimating the trends in the simulated data, particularly when the simulated dataset did not contain any missing data and when the simulated was a negative trend (Fig. 6). However, trend estimates from the PLR models (both JAGS and ubms) were slightly more accurate and precise than the EGM (unmarked) indicating that they provided the most reliable trend estimates for Sooty Grouse. DISCUSSION We evaluated multiple statistical frameworks to estimate trends from a targeted species-specific monitoring protocol for Sooty Grouse and detected population-level declines. Given the large number of trend modeling frameworks available, determining the best statistical approach to use was complex. Only after direct comparison of various modeling frameworks and simulation tests did we determine which produced the most accurate population trend estimates for Sooty Grouse in western Oregon. Abundance models (specifically PLR) were clearly the preferred method for estimating Sooty Grouse trends (Fig. 3). Conversely, occupancy models resulted in imprecise and inaccurate trend estimates because they use detection/non-detection data providing less detail than counts when sampling across an entire survey route (Fig. 5). Similarly, Monroe et al. (2019) conducted simulations on multiple statistical frameworks to estimate Greater Sage-grouse ( Centrocercus urophasianus ) trends and found trend accuracy to vary among models. Statistical model and survey design were also found to influence population estimates for Dusky Grouse (Monroe et al. 2019, Leipold et al. In review ). Our findings echo Hess et al. (2001) in that the methods employed to estimate trends matter for trend detection and interpretation, and care should be taken when making this choice to ensure reliable population trend estimates. Beyond the purpose of our study, simulations can be used to determine minimum sample sizes for detecting a trend in abundance (Blomberg and Hagen 2020), which is useful when implementing monitoring programs. The declining trend we detected in the Sooty Grouse population in Oregon aligns with the negative BBS trend estimates for Sooty Grouse across its range (Sauer et al. 2020). However, through a specialized monitoring program, we were able to detect a more precise negative trend, suggesting that BBS is not capable of accurately detecting a near 3 % decline for this species. This should raise concerns about the precision of population trends throughout the Sooty Grouse range estimated with BBS analyses. In a related analysis, Feng and Che-Castaldo (2021) found that at broad spatial scales, multi-year trends of common birds were generally consistent among monitoring datasets and modeling methods. However, at finer spatial scales, trends did not align as well, highlighting not only the importance of the detail of the data source (counts versus detection/non-detection), but also the statistical model selected for estimating trends. The choice of model framework impacted our interpretation not only of the trend, but also the observation processes (i.e., detection rates differed depending on that choice). The positive relationship between detection and number of survey stops completed along a route during a visit was intuitive as a longer survey route covers more area, and thus would have higher probability of having a Sooty Grouse present. Similarly, the more 3-min surveys conducted along a route increased the chances of detecting a Sooty Grouse through increased survey area and listening time. Thompson and La Sorte (2010) compared five abundance trend modeling methods for six species of songbirds. They found consensus among methods; however, variability in detection probability impacted trend estimates for some methods, highlighting the utility of including detection probability in abundance trend analyses through the model used and/or the study design. Several studies on Greater Sage-grouse ( Centrocercus urophasianus ) have estimated trends using N-mixture models (McCaffery et al. 2016, Blomberg and Hagen 2020, Dinkins et al. 2021). Similar to our findings, McCaffery et al. (2016) and Monroe et al. (2019) found that N-mixture models (like our PLR) performed well to identify a real trend for Greater Sage-Grouse with variability in detection and a sparse dataset. These studies included simulations that allowed them to estimate predictive accuracy, including sample size and repeat survey requirements (more repeats improved estimates; McCaffery et al. 2016, Monroe et al. 2019). While N-mixture models have demonstrated advancement for estimation of the ecological and observation processes, simulations have also shown mixed results related to the precision of estimates (McCaffery et al. 2016, Veech et al. 2016, Monroe et al. 2019). Based on simulations run by Leipold et al. ( In review ) for Dusky Grouse, they identified the ideal number of survey sites and repeat visits for obtaining precise and unbiased abundance estimates from N-mixture models. When dealing with a species with lower detection rates, accounting for imperfect detection becomes even more critical because of the bias it creates in the ecological parameter ( Couturier et al. 2013). Ultimately the study design and analysis methods should be chosen based on focal species and research questions, but it is also important to understand how this choice could influence the population trend estimates. The coarser the resolution of the data, the less informative trend estimates are likely to be (i.e., detection/non-detection). In our occupancy models we observed very little variability of the occupancy of a route between years, because many routes had at least one grouse present each year. Our abundance models were much more sensitive to nuanced population changes, and thus ultimately more informative. The observed higher detection rates from our occupancy models were anticipated because the likelihood of detecting at least one grouse over an entire route with multiple survey stops is high. Whereas the count data was better able to capture the annual variability and detect changes which resulted in a more accurate and precise trend estimate. When drawing inference from our analyses, it is important that we consider the constraints of the dataset. In some years, fewer survey stops were conducted along any given route due to downed trees, landslides, snow blocking access, and other logistical constraints. This resulted in variability in the number of locations surveyed per route within and between years, which we accounted for by including it as a covariate on detection. Our study took place within a relatively narrow elevation band (between 275 – 1,225 m. a. s. l.) due to logistical constraints (access and snow at the time period of peak hooting). Trends could potentially differ at higher and lower elevations. For example, lower elevation habitat in the foothills has historically been more modified by humans for agriculture, development, and logging due to accessibility compared to higher elevation habitats that have likely seen less impact from anthropogenic activities. Limitations due to elevation could be potentially be addressed in the future with the use of autonomous recording units (ARUs) at sites that are otherwise inaccessible in the early spring. ARUs could also provide increased temporal resolution. States vary in their approach to managing upland game birds; however, the overall goal is to ensure population sustainability into the future while providing hunting opportunity (Dahlgren et al. 2021). Ideally management strategies take into consideration the life history of the species, which can be challenging when a species is under studied. Without population monitoring and trend information, harvest management is often not science-based and is reactive to perceived trends or trial-and error based on what has been done in the past. Common methods for managing harvest include adjusting bag and possession limits, season length and timing, open areas, and method of take (such as weapon restrictions). When population trend information is available, it is often used to reactively adjust future harvest or proactively in an adaptive management strategy to understand the impacts of harvest and other factors on populations and to adaptivity adjust as more information is collected (Dahlgren et al. 2021). The results of our study comprehensively assessed multiple methods for estimating population trends of a species of conservation concern over a large spatial extent and indicated a declining trend for Sooty Grouse in western Oregon, which is a valuable piece of information for developing conservation and management strategies (e.g., maintaining viable populations while simultaneously retaining harvest opportunity). Our results highlight the need for more detailed analyses to investigate how landscape variables and human disturbance (e.g., development and hunting) contribute to the declines we found. Consistent, targeted monitoring across 14 years allowed us to detect this decline providing wildlife managers with a path forward to better understand these declines and to seek solutions. The inference of this study goes beyond Sooty Grouse populations, as trends provide a widespread and practical tool for assessing the status of wildlife populations. Advances in statistical modeling techniques have resulted in a suite of methods available for estimating population trends and our research has demonstrated how and why that choice matters. LITERATURE CITED Blomberg, E. J. and C. A. Hagen. 2020. How many leks does it take? Minimum sample sizes for measuring conservation outcomes in greater sage‐grouse. Avian Conservation and Ecology 15:9. Cline, M., K. Walton, J. Dinkins, and S. J. K. Frey. 2024. Population assessment of western Oregon Sooty Grouse 2022–2023. Pitman‐Robertson Grant Annual Performance Report, Oregon Department of Fish and Wildlife. Couturier, S., P. Giraud, and F. Jiguet. 2013. Estimating abundance and population trends when detection is low and highly variable: A comparison of three methods for the Hermann’s tortoise. Wildlife Biology 19:167–175. Dail, D. and L. Madsen. 2011. Models for estimating abundance from repeated counts of an open metapopulation. Biometrics 67:577–587. Dahlgren, D. K., E. J. Blomberg, C. A. Hagen, and R. D. Elmore. 2018. State management of upland and small game. In T. J. Ryder (ed.), State Wildlife Management and Conservation . Johns Hopkins University Press, pp. 96–115. Dahlgren, D. K., E. J. Blomberg, C. A. Hagen, and R. D. Elmore. 2021. Upland game bird harvest management. In K. L. Pope and L. A. Powell (eds.), Harvest of Fish and Wildlife: New Paradigms for Sustainable Management . CRC Press, pp. 213–220. Denwood, M. J. 2016. runjags: An R package providing interface utilities, model templates, parallel computing methods and additional distributions for MCMC models in JAGS. Journal of Statistical Software 71:1–25. Dinkins, J. B., K. J. Lawson, and J. L. Beck. 2021. Influence of environmental change, harvest exposure, and human disturbance on population trends of greater sage‐grouse. PLoS ONE 16:e0257198. Fedy, B. C. and C. L. Aldridge. 2011. The importance of within‐year repeated counts and the influence of scale on long‐term monitoring of sage‐grouse. Journal of Wildlife Management 75:1022–1033. Feng, X. and J. Che‑Castaldo. 2021. Comparing the reliability of relative bird abundance indices from standardized surveys and community science data at finer resolutions. Ecological Indicators 121:106–115. Fiske, I. and R. Chandler. 2011. unmarked: An R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43:1–23. 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, OR. 111 pp. Gibson, L. A. 2011. The importance of incorporating imperfect detection in biodiversity assessments: A case study of small mammals in an Australian region. Diversity and Distributions 17:613–623. Hess, G. R., et al. 2001. Linear trend analysis: A comparison of methods. Wildlife Biology 7:113–120. Hoerner, S. E. 2018. Improving survey methodology to monitor rare grassland birds in the Northern Great Plains. M.S. thesis, South Dakota State University. https://openprairie.sdstate.edu/etd/1687 Hostetler, J. A. and R. B. Chandler. 2015. Improved state‐space models for inference about spatial and temporal variation in abundance from count data. Ecology 96:1713–1723. Itescu, Y. and S. Meiri. 2021. North American Breeding Bird Survey underestimates regional bird richness. Ecosphere 12(7):e03925. https://doi.org/10.1002/ecs2.3925 Kellner, K., N. Fowler, T. Petroelje, T. Kautz, D. Beyer, and J. Belant. 2021. ubms: An R package for fitting hierarchical occupancy and N‑mixture abundance models in a Bayesian framework. Methods in Ecology and Evolution 13:577–584. https://doi.org/10.1111/2041-210X.13777 Kéry, M. and M. Schaub. 2011. Bayesian Population Analysis using WinBUGS: A Hierarchical Perspective . Academic Press, Burlington, MA. Kéry, M., R. M. Dorazio, L. Soldaat, A. van Strien, A. Zuiderwijk, and J. A. Royle. 2009. Trend estimation in populations with imperfect detection. Journal of Applied Ecology 46(6):1163–1172. doi:10.1111/j.1365-2664.2009.01724.x Kéry, M. and J. A. Royle. 2010. Hierarchical modelling and estimation of abundance and population trends in metapopulation designs. Journal of Animal Ecology 79(2):453–461. doi:10.1111/j.1365-2656.2009.01632.x Kéry, M. and J. A. Royle. 2021. Applied hierarchical modeling in ecology: Analysis of distribution, abundance and species richness using R and BUGS , Vol. 2: Dynamic and advanced models. Academic Press, Elsevier, New York, NY. Knape, J., P. de Valpine, R. A. Francis, and E. E. Crone. 2018. Sensitivity of binomial N‑mixture models to overdispersion: The importance of assessing model fit. Methods in Ecology and Evolution 9:2102–2114. Leipold, E. A., C. N. Gower, and L. McNew. 2025. Development of optimized methods for unbiased Dusky Grouse ( Dendragapus obscurus ) population monitoring using real and simulated data. Authorea , March 17, 2025. https://doi.org/10.22541/au.174219936.66651751 Leipold, E. A., C. N. Gower, and L. McNew. 2024. Using an ensemble approach to predict habitat of Dusky Grouse ( Dendragapus obscurus ) in Montana, USA. Avian Conservation and Ecology 19(2):7. https://doi.org/10.5751/ACE-02697-190207 MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. L. Bailey, and J. E. Hines. 2006. Occupancy estimation and modeling: Inferring patterns and dynamics of species occurrence . Academic Press, San Diego, CA. McCaffery, R., J. J. Nowak, and P. M. Lukacs. 2016. Improved analysis of lek count data using N‑mixture models. Journal of Wildlife Management 80:1011–1021. Monroe, A. P., G. T. Wann, C. L. Aldridge, and P. S. Coates. 2019. The importance of simulation assumptions when evaluating detectability in population models. Ecosphere 10:e02791. North American Bird Conservation Initiative. 2016. The state of North America’s birds 2016 . Ottawa, ON: Environment and Climate Change Canada. 8 pp. www.stateofthebirds.org Oregon Department of Fish and Wildlife. 2024. Oregon Game Bird Hunting Regulations . Salem, OR. 48 pp. Plummer, M., N. Best, K. Cowles, and K. Vines. 2006. CODA: Convergence diagnosis and output analysis for MCMC. R News 6:7–11. Plummer, M. 2024. rjags: Bayesian graphical models using MCMC. R package version 4-16. https://CRAN.R-project.org/package=rjags Rosenberg, K. V., A. M. Dokter, P. J. Blancher, J. R. Sauer, A. C. Smith, P. A. Smith, J. C. Stanton, A. Panjabi, L. Helft, M. J. Parr, and P. P. Marra. 2019. Decline of the North American avifauna. Science 366(6461):120–124. doi:10.1126/science.aaw1313 Royle, J. A. 2004. N‑mixture models for estimating population size from spatially replicated counts. Biometrics 60(1):108–115. Royle, J. A. and R. M. Dorazio. 2008. Hierarchical modeling and inference in ecology: The analysis of data from populations, metapopulations and communities . Academic Press, New York, NY. Sauer, J. R., J. E. Hines, W. A. Link, K. L. Pardieck, D. J. Ziolkowski Jr., and J. E. Fallon. 2013. The North American Breeding Bird Survey, 1966–2011: Summary analysis and species accounts. North American Fauna 79:1–32. https://doi.org/10.3996/nafa.79.0001 Sauer, J. R., W. A. Link, and J. E. Hines. 2020. The North American Breeding Bird Survey, analysis results 1966–2022. U.S. Geological Survey data release. https://doi.org/10.5066/P96A7675 Storch, I. 2000. Conservation status and threats to grouse worldwide: An overview. Wildlife Biology 6:195–204. Thompson, F. R. and F. W. La Sorte. 2010. Comparison of methods for estimating bird abundance and trends from historical count data. Condor 112:560–575. Veech, J. A., J. R. Ott, and J. R. Troy. 2016. Intrinsic heterogeneity in detection probability and its effect on n‑mixture models. Methods in Ecology and Evolution 7:1019–1028. Zwickel, F. C. and J. F. Bendell. 2004. Blue grouse: Their biology and natural history . NRC Research Press, Ottawa, ON. 284 pp. FIGURES Figure 1 | Map of western Oregon showing Sooty Grouse routes surveyed from 2011–2024. One hundred nineteen unique routes were surveyed. [1]¿p#1 Figure 2 | Sooty Grouse abundance trend estimates (measured as lambda, the finite rate of increase) and 95% credible/confidence intervals (CIs) for the five model frameworks investigated in this study. A lambda value of one represents a stable trend (indicated by the dotted gray horizontal line), below one indicates a declining trend, and above one indicates an increasing trend. Data collected in western Oregon 2011–2024. [1]¿p#1 Figure 3 | Mean abundance estimates (and 95% CIs) from the (a) PLR - JAGS, (b) PLR - ubms, and (c) EGM - unmarked abundance models by year (in blue) and mean raw counts (in red). Lambda (rate of growth per time year) is shown for each model in the upper left corner of each plot. Data collected in western Oregon 2011–2024. Figure 4 | a) Average estimated detection probability and 95% CIs for the five model frameworks used to estimate Sooty Grouse trends. b) The relationship between detection probability and number of stops completed along a survey route estimated from the Poisson linear regression model run in a JAGS framework (also referred to binomial N-mixture model). Data collected in western Oregon 2011–2024. Figure 5 | Occupancy trend (lambda) estimates (with 95% CIs) from the two model frameworks (OTM – JAGS [orange], OTM – ubms [blue]) using simulated data at three trend levels: (a) -3.0 % change per year, (b) 0.0 % change per year, no trend, (c) 3.0 % change per year). We ran the models using a dataset without any missing data (no NAs, filled circles) and including missing data that matched our Sooty Grouse dataset (with NAs, filled triangles). The dotted gray lines mark the three trend levels simulated for reference. Data collected in western Oregon 2011–2024. Figure 6 | Abundance trend (lambda) estimates (with 95% CIs) from the three abundance model frameworks (PLR – JAGS, PLR – ubms, EGM – unmarked) using simulated data at three trend levels (-3.0 [orange], 0.0 [green], 3.0 [blue] % change per year). We ran the models using a dataset without any missing data (no NAs, filled circles) and including missing data that matched our Sooty Grouse dataset (with NAs, filled triangles). The dotted black lines mark the three trend levels simulated for reference. Data collected in western Oregon 2011–2024. TABLES Table 1 | Summary of survey effort and raw counts of male Sooty Grouse (grouse) in western Oregon documented by year, route, and 3-min listening surveys. Total surveys column includes within-year repeat surveys. 2011 31 445 14.35 93 3.00 0.21 2012 56 881 15.73 142 2.54 0.16 2013 56 892 15.93 217 3.88 0.24 2014 53 856 16.15 159 3.00 0.19 2015 61 1014 16.62 269 4.41 0.27 2016 56 912 16.29 102 1.82 0.11 2017 52 813 15.63 158 3.04 0.19 2018 64 1036 16.19 202 3.16 0.19 2019 76 1257 16.54 294 3.87 0.23 2020 106 1665 15.71 406 3.83 0.24 2021 124 2006 16.18 442 3.56 0.22 2022 122 1971 16.16 383 3.14 0.19 2023 91 1432 15.74 233 2.56 0.16 2024 99 1516 15.31 315 3.18 0.21 Supplementary Material File (image3.emf) Download 1.01 MB File (image4.emf) Download 2.03 MB File (image5.emf) Download 723.26 KB File (image6.emf) Download 673.15 KB Information & Authors Information Version history V1 Version 1 23 April 2025 Peer review timeline Published Wildlife Biology Version of Record 16 Apr 2026 Published Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Wildlife Biology Keywords abundance hierarchical models occupancy population trends simulations sooty grouse Authors Affiliations Sarah Frey 0000-0002-4343-0700 [email protected] Oregon State University College of Agricultural Sciences View all articles by this author Kelly Walton 0009-0003-4851-4692 Oregon Department of Fish and Wildlife View all articles by this author Mikal Cline Oregon Department of Fish and Wildlife View all articles by this author Lindsey Sanders Oregon Department of Fish and Wildlife View all articles by this author Jonathan Dinkins Oregon State University College of Agricultural Sciences View all articles by this author Metrics & Citations Metrics Article Usage 499 views 211 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Sarah Frey, Kelly Walton, Mikal Cline, et al. Abundance and occupancy trends of Sooty Grouse in western Oregon: determining best modeling practices by comparing observed and simulated data. Authorea . 23 April 2025. DOI: https://doi.org/10.22541/au.174540808.83675545/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.174540808.83675545/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:'a0047765ef0f8e2e',t:'MTc3OTU0MzU2Mw=='};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())}}}})();