Influence of fine-scale habitat characteristics on sage-grouse nest site selection and nest survival varies by mesic and xeric site conditions

ABSTRACT Resource managers and scientists across western U.S. agencies seek methodologies for identifying environmental attributes important to both wildlife conservation and broad-scale land stewardship. The Greater Sage-Grouse (Centrocercus urophasianus; hereafter, sage-grouse) exemplifies a species in need of this broad-scale approach given widespread population declines that have resulted from loss and degradation of habitat from natural and anthropogenic disturbances. These include agricultural land conversion, conifer expansion, energy development, and wildfire coupled with ecological conversion by invasive plants such as cheatgrass (Bromus tectorum). Development of habitat assessments and conservation actions for sage-grouse benefit from studies that link demographic responses to habitat selection patterns. To address this, we examined nest survival of sage-grouse in relation to fine-scale habitat patterns (i.e., field-based habitat measurements) that influenced nest site selection, using data from nests of telemetered females at 17 sites over 6 years in Nevada and northeastern California, USA. Importantly, sites spanned mesic and xeric average precipitation conditions that contributed substantially to vegetation community structure across cold desert ecosystems of the North American Great Basin. Vegetative cover immediately surrounding sage-grouse nests was important for both nest site selection and nest survival, but responses varied between mesic and xeric sites. For example, while taller perennial grasses were selected at xeric sites, we found no evidence of selection for perennial grass at mesic sites, indicating a functional response to availability of habitat features between hydrographic regions. Furthermore, perennial grass height and forb height both had positive effects on nest survival at xeric sites, but we found varying effects at mesic sites. We emphasize that precipitation conditions driving ecosystem productivity vary regionally among sagebrush communities, shaping vegetation structure and suitable habitat conditions for nesting sage-grouse. How to Cite Brussee, B. E., P. S. Coates, S. T. O'Neil, M. A. Ricca, J. E. Dudko, S. P. Espinosa, S. C. Gardner, M. L. Casazza, and D. J. Delehanty (2022). Influence of fine-scale habitat characteristics on sage-grouse nest site selection and nest survival varies by mesic and xeric site conditions. Ornithological Applications 125:duac052. LAY SUMMARY Effective conservation and management for sensitive species requires maintenance of habitat conditions that promote demographic success and persistence of populations. We measured field-based fine-scale vegetation characteristics at nests of Greater Sage-Grouse across 17 sites within California and Nevada, USA, during 2012–2017, and we examined associations with nest selection and nest survival among female sage-grouse nesting in mesic and xeric sagebrush-steppe environments. We demonstrate strong associations with fine-scale features and variation in these associations across precipitation conditions. Our results suggest differences in the regional availability of important vegetation components across xeric and mesic conditions that influence sage-grouse occurrence and reproductive success and highlight the importance of ecological context and long-term average precipitation when developing habitat management prescriptions. Variation in the relative influences of herbaceous cover among habitat types within our study helps to elucidate discrepancies observed among studies of grass-related variables affecting selection or survival of sage-grouse nests. RESUMEN Los administradores de recursos y los científicos de las agencias del oeste de EEUU buscan metodologías para identificar atributos ambientales importantes tanto para la conservación de la vida silvestre como para la gobernanza de la tierra a gran escala. Centrocercus urophasianus es un ejemplo de una especie que necesita este enfoque a gran escala dada la disminución generalizada de la población como resultado de la pérdida y degradación del hábitat por disturbios naturales y antropogénicos. Estos incluyen la conversión de tierras agrícolas, la expansión de coníferas, el desarrollo energético y los incendios forestales, junto con la conversión ecológica por plantas invasoras como Bromus tectorum. El desarrollo de evaluaciones de hábitat y acciones de conservación para C. urophasianus se beneficia de estudios que vinculan las respuestas demográficas con los patrones de selección de hábitat. Para abordar esto, examinamos la supervivencia de los nidos de C. urophasianus en relación con patrones de hábitat a escala fina (i.e., mediciones de hábitat basadas en el campo) que influyeron en la selección del sitio de anidación, utilizando datos de nidos provenientes de hembras medidas con telemetría en 17 sitios durante 6 años en Nevada y el noreste de California, EEUU. Es importante destacar que los sitios abarcaron condiciones de precipitación promedio mésicas y xéricas que contribuyeron sustancialmente a la estructura de la comunidad de la vegetación en los ecosistemas desérticos fríos de la Gran Cuenca de América del Norte. La cobertura vegetal alrededor de los nidos de C. urophasianus fue importante tanto para la selección del sitio de anidación como para la supervivencia del nido, pero las respuestas variaron entre los sitios mésicos y xéricos. Por ejemplo, aunque se seleccionaron pastos perennes más altos en sitios xéricos, no encontramos evidencia de selección de pastos perennes en sitios mésicos, lo que indica una respuesta funcional a la disponibilidad de las características del hábitat entre regiones hidrográficas. Más aún, la altura del pasto perenne y la altura de las hierbas tuvieron efectos positivos en la supervivencia del nido en los sitios xéricos, pero encontramos efectos variables en los sitios mésicos. Enfatizamos en que las condiciones de precipitación que impulsan la productividad del ecosistema varían regionalmente entre las comunidades de artemisa, dando forma a la estructura de la vegetación y a las condiciones del hábitat adecuadas para la anidación de C. urophasianus.


INTRODUCTION
Identifying habitat factors that influence demographic performance is critical for the conservation and restoration of species (Gaillard et al. 2010, Matthiopoulos et al. 2015). Yet, ecological complexity causes variation in the expression of species' habitat preferences that influence distribution and demographic performance (Chalfoun and Martin 2007). Such complexity complicates efforts to generalize habitat requirements for management purposes because localized patterns in species' habitat use and measures of success often deviate from broadly observed trends. Differences in environmental conditions among geographic regions influence the general availability of core habitat features, leading to habitat functional responses (habitat selection depends on broader habitat availability; Mysterud and Ims 1998, Paton and Matthiopoulos 2016, Holbrook et al. 2019. In the American Intermountain West, environmental heterogeneity comprises gradients in elevation, climate (e.g., average precipitation), soil aridity, and subsequent herbaceous plant communities that span a variety of shrubland and sagebrush ecosystems providing habitat for numerous wildlife species, including the imperiled Greater Sage-Grouse (Centrocercus urophasianus, hereafter sage-grouse; Crawford et al. 2004. A wide range of threats to native species and their habitats exist within this region (Knick et al. 2003, Davies et al. 2011), yet consensus for management actions has often been elusive. In addition to the landscape's environmental complexity, stakeholder differences (Boyd and Svejcar 2009), policy and land use changes (Maestas et al. 2003), and other socioecological forces (Svejcar et al. 2017) shape management across vast spatial extents. The sage-grouse is an indicator species within western sagebrush ecosystems, emblematic of both the loss and degradation of habitat integrity within these systems (Knick and Rotenberry 2000, USFWS 2015, while serving as an example of the conservation chal-lenges faced by western land management agencies tasked with safeguarding core habitats for at risk species (Doherty et al. 2022).
Management efforts for sage-grouse populations are focused on reversing long-term population decline and contraction of habitat , Aldridge et al. 2008, Coates et al. 2021, Doherty et al. 2022. Identifying critical habitat components are central to these land stewardship and conservation efforts (Doherty et al. 2010, Knick and), yet defining habitat remains complicated by landscape heterogeneity and associated habitat functional responses. When considering populations that occupy heterogeneous landscapes and biomes, a prevailing view is that it is unrealistic to expect a "one-size-fits-all" management solution that is ecologically generalizable or for habitat guidelines to be broadly applicable across an entire species distribution (Wiens 1989, Morrison 2012. Recent management frameworks for sage-grouse have recognized that monitoring habitat requires an understanding of environmental conditions encompassing the species' distributional range, and the corridors, distinct sub-populations, individual home ranges, seasonal patterns, and daily foraging and movements nested within the broader delineations (e.g., context-dependence; Stiver et al. 2015). For example, while sage-grouse are widely regarded as a sagebrush obligate species, female sage-grouse select versatile environmental characteristics during the critical nesting period, that optimize cover from aerial and terrestrial predators Delehanty 2010, Conover 2007), provide a thermal environment that moderates fluctuating cold and hot weather conditions (Hansell andDeeming 2002, Anthony et al. 2021), benefit incubation constancy (percent of time spent on nest in 24 hr;  and provide access to suitable forage during incubation recesses (Dudko et al. 2019). In water-limited cold semidesert ecosystems (e.g., the Great Basin region, ~25% of total sage-grouse range), the types and distribution of resources available to Brussee et al. Fine-scale habitat influences sage-grouse nest selection and survival 3 nesting sage-grouse are influenced by soil water and nutrient holding capacity (Miller et al. 2013, Chambers et al. 2014) as well as localized gradients of precipitation, which shape plant community composition and structure. Variation in precipitation across sagebrush ecosystems, and subsequent habitat functional responses, are generally not considered in sage-grouse habitat selection and nest survival analyses (Hagen et al. 2007, Smith et al. 2020, complicating the ability to scale up the influences of habitat characteristics and sometimes prompting uncertainty about the broader application of management based on localized vegetation and habitat measurements (Smith et al. 2020). While remotely sensed land cover products have been conventionally adopted for widespread use in studies of sage-grouse habitat, estimates of specific environmental requirements for sage-grouse were traditionally obtained with greater precision based on direct, ground measurements in the field ("fine-scale measurement;" Connelly et al. 2000Connelly et al. , 2004. However, despite the finer grain of evaluation, species' responses to fine-scale measurements are still affected by context dependence and potential habitat functional responses, particularly in the heterogeneous Great Basin region of the American West. We examined the influences of habitat characteristics from fine-scale measurements on nest site selection and nest survival of sage-grouse at a wide range of study sites in Nevada and California, USA. Our objectives were, to better understand habitat functional responses driven by variation in coarse moisture gradients defined by long-term precipitation conditions across sagebrush ecosystems. Importantly, this study region encompasses a diverse range of environmental conditions and site productivity (Miller et al. 2013), within which detailed fine-scale measurements of sites used by sagegrouse sub-populations were carried out. In this study, we emphasize the importance of nest site habitat features associated with nest fate and consider that selection patterns may vary by hydrographic regions that were predominantly mesic or xeric based on long-term precipitation patterns (Coates et al. 2017). We also highlight the important variation in fine-scale vegetation composition and structure that exists between mesic and xeric regions. Understanding sage-grouse responses to variability in vegetation characteristics across mesic and xeric precipitation conditions can improve knowledge of the factors influencing localized population growth or decline, thereby furthering conservation efforts for this indicator species and the imperiled sagebrush ecosystem.

Study Area
Our study took place across 17 sites that facilitated data collection from sage-grouse sub-populations within northeastern California and Nevada (42.0023°-37.5550°N, 114.7196°-120.7547°W) during 2012-2017 ( Figure 1). Elevation ranged from ~1,100 m to ~3,700 m across all sites. The most southwestern sites were within the Bi-State Distinct Population Segment (Oyler-McCance et al. 2014), which was along the border of California and Nevada and bounded by the Sierra Nevada Mountains on the west and non-sagebrush habitat on the east. The northern and western sites, including those in the Bi-State, typically experience greater precipitation and are characterized as sagebrushsteppe. Conversely, the southeastern sites consist of warmer and drier soils characteristic of sagebrush semidesert (West and Young 2000). Sagebrush species included mountain big sagebrush (Artemisia tridentata spp. vaseyana) at high elevations and Wyoming big sagebrush (Artemisia tridentata ssp. wyomingensis), black (Artemisia nova), and low (Artemisia arbuscula) sagebrush below 2,100 m. Rabbitbrush (Chrysothamnus ssp.), Mormon tea (Ephedra viridis), snowberry (Symphoricarpos ssp.), western serviceberry (Amelanchier alnifolia), and antelope bitterbrush (Purshia tridentata) are examples of common non-sagebrush shrubs. Herbaceous vegetation was comprised of non-native annual grasses, such as cheatgrass (Bromus tectorum) and medusahead rye (Taeniatherum caput-medusae), and native perennial grasses including needle and thread (Hesperostipa comata), Indian ricegrass (Achnatherum hymenoides), and squirreltail (Elymus elymoides).

Field Methods
Grouse marking and tracking. We captured sage-grouse using spotlighting techniques annually during March through November. Captured sage-grouse were outfitted with battery powered, necklace-style very high frequency (VHF) transmitters (~20 g; A4060; Advanced Telemetry Systems, Isanti, Minnesota, USA). Additionally, from 2012 to 2016, a subset of sage-grouse (~16%) were equipped with a rump-mounted GPS-PTT (22 g; GT-22GS-GPS Geotrak, Apex, North Carolina, USA) device with an attached small ancillary VHF transmitter to allow for ground radio tracking and recovery of GPS-PTT unit. No differences in nest survival were detected between rump-mounted GPS and necklace-style VHF transmitters . A complete description of transmitters and attachments has been described in Severson et al. (2019).
To locate sage-grouse nests and monitor nest success, we conducted intensive on-the-ground tracking of sage-grouse prior to and throughout the nesting season. Specifically, we used telemetry to locate marked female sage-grouse at least twice per week during the months of March through May. We obtained visuals on each location, or circled grouse at a radius of ~50 m if a visual location could not be confirmed. We verified nests visually after females were found in the same position on two consecutive observations (Coates and Delehanty 2010), using triangulation to initially locate nests. On subsequent nest visits, we used a circling technique to avoid flushing sage-grouse and prevent observer-induced nest abandonment (Gibson et al. 2015). We assessed female status as on or off nest during visits from a 50-m distance using radio-telemetry. If the female was on the nest, we assumed the nest was active and avoided flushing females. If a female was off the nest, we visually inspected the nest and counted eggs. VHF transmitters were equipped with mortality signals that activate after 8 hr of inactivity which allowed us to assess whether females were alive or dead during each check. We monitored each nest 2 times or more per week until nest fate, hatch, fail, or nest abandonment was determined. The frequency of nest checks ensured that each nest was visited within about 2-3 days of the predicted hatch date. After egg incubation ended, we visually inspected each nest to assess nest fate. We considered nests successful if one or more chicks hatched, as determined by eggshell remains or the presence of one or more chicks in the nest bowl (Coates and Delehanty 2010). We considered nests unsuccessful when the entire clutch failed to hatch following any combination of predation or abandonment.

Fine-Scale Habitat Sampling
We measured fine-scale habitat characteristics in the field at each nest and at random locations (Supplementary Material Appendix A and Appendix B Table 1). We sought to collect habitat measurements within 3 days of nest fate. However, habitat measurements were often taken beyond this date for successful nests, or sometimes before an estimated hatch date for nests that failed (Gibson et al. 2016). Therefore, we used statistical methods to correct misalignments in vegetation characteristics related to timing of field measurements (see plant phenology analysis below). We created site boundaries using the 100% minimum convex polygon within ArcGIS (ArcGIS 10.6.1, ESRI 2018, Redlands, CA) surrounding all telemetry locations of sage-grouse at the site, including outlying movements, and were thought to represent all area available to sage-grouse at each site. To characterize site level availability, we generated one random location for every nest within each site MCP. If a random location was inaccessible, such as on those on private property, we generated a new random location within the site MCP. We measured vegetation characteristics at the nest or random locations and along 4 (2012-2013) or 3 (2014-2017) 10-m transects extending from the nest bowl or randomly generated locations. For random locations, the nearest shrub served as the center point. For shrub vegetation directly over the nest or random point, we recorded maximum crown width (cm), maximum width of the shrub perpendicular to the first measurement, and maximum shrub height (cm). Additionally, we measured horizontal and vertical cover at the nest bowl using a visual obstruction method (Jones 1968). At the end of each 10-m transect, we used the Daubenmire method (Daubenmire 1959) to measure herbaceous understory vegetation cover at a 20 × 50 cm subplot. Understory groups consisted of perennial grass, annual grass, perennial forb, annual forb, residual cover, litter cover, and bare ground (Supplementary Material Appendix A). Observers estimated the cover class of each understory group from 1 to 7 where 1 = 0-5% cover; 2 = 6-15% cover; 3 = 16-25% cover; 4 = 26-50% cover; 5 = 51-75% cover; 6 = 76-95% cover; and 7 = 96-100% cover. For analysis, we converted cover classes to midpoint percentage values (e.g., cover class 1 = 2.5%, cover class 2 = 10.5%, etc.) and averaged values across all subplots to derive continuous values for vegetation characteristics at a 10-m scale.
We used the line-intercept method (Canfield 1941) along the same 10-m transects to measure overstory shrub canopy for multiple shrub communities. Shrub communities included tall sagebrush (i.e., big sagebrush), dwarf sagebrush (i.e., little sagebrush, and black sagebrush species), and non-sagebrush shrubs that included montane and lowland shrub types. For each shrub community, we derived percent shrub cover within 10 m of the nest by summing the total cover along all of the transects and dividing by the total length of transects (30 or 40 m).
We measured vegetation heights for each of 5 cover categories using the nearest plant to the center point of the transect line: sagebrush, non-sagebrush shrubs, perennial grass, perennial forb, and residual grass. We averaged heights for each cover category across all transects to derive mean vegetation heights for each cover category at the 10-m scale. Additionally, we measured heights of 10 randomly selected shrubs along the transect line, and we averaged heights from all transects to derive average shrub height within 10 m of nests or random points. For more detailed description of finescale habitat collection protocol, see Supplementary Material Appendix A and Appendix B Table 1.

Defining hydrographic regions.
Vegetation composition across the Great Basin varies with respect to soil temperature and moisture gradients (Miller et al. 2013, Chambers et al. 2014) driven by broad climatic pat-terns such as precipitation, and can be partitioned into mesic and xeric hydrographic zones (Chambers et al. 2014). We characterized sites as mesic or xeric within the Great Basin, using broad precipitation patterns to serve as a tractable metric for managers to use when evaluating local habitats and site conditions. To define sites as xeric or mesic, we first averaged annual precipitation across all sites (average: 35.0 cm; range: 24.9-47.7 cm) using PRISM Climate data (30-year normal; PRISM Climate Group, Oregon State University, Daly et al. 2008). We then characterized xeric sagebrush sites (n = 7) as those receiving < 35.0 cm average annual precipitation ( Figure 1). Sites with ≥ 35.0 cm precipitation were characterized as mesic (n = 10). The 35.0 cm cutoff aligned with the hydrographic delineation for seasonal habitat mapping in Coates et al. (2016a). We included descriptive summaries of fine-scale habitat measurements contrasted in xeric vs. mesic sites and used two-sided non-parametric tests (independent 2-group Mann-Whitney U tests) to demonstrate statistical evidence of differences (or similarities) in available vegetation characteristics between site condition type (Brussee et al. 2022a).

Plant phenology analysis.
Although fine-scale habitat surveys were conducted near predicted hatch date, we followed a modified date-adjustment recommendation by Gibson et al. (2016) and Smith et al. (2018b) to further prevent plant phenology (Hausleitner et al. 2005) from confounding differences between successful and failed nests. Specifically, we adjusted measurements of heights and cover for herbaceous vegetation (i.e., perennial grass, annual grass, and perennial forb, annual forb) for both xeric and mesic sites based on calculated growth rates from generalized linear mixed effects models (GLMMs). We used the estimated coefficients from models to adjust the vegetation measurement to a calculated estimated hatch date. Given a 38-day nesting period, for both successful and failed nests, we adjusted vegetation measurements to 19 days after a nest was first located, which is the approximate midpoint of the laying and incubation period, when most nests are located.

Preliminary variable reduction.
We considered 23 potential habitat predictors in our models for nest selection and survival (Supplementary Material Appendix B Table 1). Because we measured similar habitat characteristics using multiple methods, some covariates were subsets of others (e.g., tall sagebrush cover is a component of total sagebrush cover and total shrub cover), and were inherently correlated (|r| > 0.65). Therefore, we could not confidently fit models to the complete dataset without potential collinearity issues (Dormann et al. 2013).
To address this issue, we first identified nested variables or variables with high correlation coefficients, which fell into 3 groups: percent cover of various shrub types, shrub heights measured using 2 techniques, 2 measurements of shrub width at nests. We used Bayesian latent indicator scale selection (BLISS; Stuber et al. 2017) to identify the most influential variables among the three groups while maintaining other variables within the model (Supplementary Material Appendix B Table 1). For each group, the BLISS method estimates a latent categorical probability distribution, where each predictor variable within the group is assigned equal prior weight (Stuber et al. 2017). We then identified the predictor variables from each of the three groups that had the highest posterior probability of inclusion, and carried them forward for inclusion in final models of nesting habitat selection and survival. We also utilized a Bayesian indicator variable to eliminate groups or variables that were unimportant in explaining nest site selection or nest survival. Specifically, within each iteration, each variable k was multiplied by an indicator variable (w k ) that followed a Bernoulli distribution with prior probability set to 0.5 (O'Hara andSillanpää 2009, Converse et al. 2013). We used the mean of w k to estimate the posterior probability of each variable being included in the model. To ensure that variance was consistent across all iterations, regardless of how many variables were included in the model, we followed recommendations by Link and Barker (2006) and assigned total model variance a prior Gamma distribution with 3.29 and 7.8 as the parameters. We ran variable reduction procedures separately for nest selection and nest survival using model structures described in the following sections, and estimated coefficients and w k separately for each hydrographic region. We included variables within our final models of selection or survival if w k > 0.5 (posterior probability exceeded prior probability) for either hydrographic region. For each model incorporating Bayesian variable selection, we used Markov Chain Monte Carlo (MCMC) sampling with 3 chains of 20,000 iterations, a burn-in of 10,000 iterations, and retained every 5 th sample. Nest site selection model. For our final model of nest site selection, we included all finescale habitat variables identified during the variable reduction step. We used a GLMM framework within a Bayesian modeling environment to evaluate the effects of fine-scale habitat predictors on nest site selection. To allow for site-level variation that might otherwise confound our ability to detect differences between mesic and xeric sites, we implemented a random slopes model (Gillies et al. 2006, Gelman and Hill 2007, Kéry 2010, estimating an intercept (κ) and slope (β) for each site (s) in our data, which took the form: , X s β s are a vector of site-level selection coefficients (β s ) multiplied by the matrix of fixed covariates (X), and γ j is a random intercept effect for year. Site-level intercepts (κ s ) and coefficients (β s ) from each hydrographic region were assumed to be drawn from a multivariate normal distribution (MVN), with a grand mean intercept µ κregion and slope coefficient µ βregion estimated for each hydrographic region. Variance and covariance parameters, σ 2 κregion , σ 2 βregion , and σ κregionβregion , representing site intercept, slope, and slope-intercept covariance, respectively, were also estimated based on the hydrographic region assignment for each site (Kéry 2010). To facilitate estimation of the variancecovariance matrix, we specified the scaled-inverse Wishart prior distribution (Gelman and Hill 2007). The response observations Y followed a Bernoulli distribution, with y = 1 indicating a nest, and y = 0 indicating a random available location. Additionally, we estimated the difference in effects between hydrographic regions as measured by the difference between grand means µ βregion for each covariate. To estimate model parameters, we used MCMC sampling with 3 chains of 100,000 iterations following a burn-in of 50,000 iterations and retained every 100 th sample.
We evaluated model performance following Boyce et al. (2002), as our nest site selection results could be understood as a resource selection function (RSF). Briefly, we generated selection predictions for our data using the resource selection parameters estimated from our final model and an exponential link function (McDonald 2013). We created 10 bins of RSF habitat classes and calculated the number of observed locations within each bin. Using Spearman's rank ρ, we evaluated the agreement between observed and expected numbers of locations in each bin. Models with good fit have Spearman's ρ > 0.5 (Boyce et al. 2002).

Nest survival model.
We used a Bayesian shared frailty model framework (Halstead et al. 2012, Severson et al. 2019) to estimate impacts of fine-scale habitat covariates on daily nestsurvival of sage-grouse. As was for nest site selection, the final model included habitat covariates that were identified during variable reduction. For nests where microhabitat data were not collected (n = 25 nests at mesic sites, n = 18 nests at xeric sites), we imputed values using a normal distribution with a mean of zero and standard deviation of one, based on standardized data. For each nest, we constructed encounter histories consisting of days when each nest was alive or censored throughout the study period. For each nest, the exposure period began the date the nest was found and ended at nest fate (i.e., hatched or failed). Thus, the number of exposure days for each nest varied depending on the stage of incubation when the nest was found and how soon nest fate occurred after being found. We then calculated the unit hazard (UH) for each nest exposure day as a Bernoulli trial (i.e., alive or censored), taking the form: where the subscripts h, i, j, and s reference individual nest, day, year, and site respectively, and X s β s are a vector of sitelevel coefficients β s on the matrix of fixed covariates X, which included fine-scale habitat predictors. We estimated random slopes and intercepts for each site utilizing the same MVN data structure as described for nest site selection models.

Brussee et al.
Fine-scale habitat influences sage-grouse nest selection and survival 7 Effects of age of the hen (yearling or adult), and nest initiation day of season (i.e., the day of year the nest was located relative to the mean day of nest locations for a given year) were also included in the model. Year was included as a random intercept (γ j ) in all models to account for unbalanced sampling and autocorrelation across time. We estimated the difference in effects between hydrographic regions as the difference between the grand means (µβ region ) for each covariate. We calculated cumulative nest survival for a 38-day egg-laying and incubation period (Batterson and Morse 1948, Patterson 1952, Nelson 1955) using: where CH is the cumulative hazard. We then calculated s, the cumulative survival rate, from the cumulative hazard, assuming constant hazard across the incubation and laying phases. Because we included categorical variables for hen age (i.e., adult and yearling) overall survival was calculated as a weighted average of each of 2 classes and was conditioned on all other covariates occurring at their mean values. To estimate model parameters, we used MCMC sampling with 3 chains of 100,000 iterations following a burn-in of 200,000 iterations and retaining every 100th sample. We calculated model goodness of fit using a Bayesian Pvalue (Gelman et al. 2013) by simulating encounter histories from our model and iterativly comparing them to observed encounter histories, following methods described in Schmidt et al. (2010). Values approaching 0 or 1 indicate poor model fit (Gelman et al. 2013).
All models (nest selection and survival, including preliminary variable reduction) were run using JAGS 4.2.0 (R version 3.4.3) and the package rjags (Plummer 2016). We checked for convergence of chains visually and verified convergence with Gelman-Rubin statistic (r <1.05; Gelman and Rubin 1992). For all parameters, we report median values of the posterior distribution and 95% credible intervals (CRI), unless otherwise stated. We considered strong support for covariates with probability of direction (i.e., pd) > 0.95, that is the proportion of the posterior distribution of the median's sign (Makowski et al. 2019), and moderate support for covariates for covariates with pd > 0.85. We provide tables of model results as an associated data release, available through USGS ScienceBase (see Acknowledgments), and code to fit models is provided (see Acknowledgments).

RESULTS
We tracked a total of 842 (site average = 84) and 757 (site average = 108) female sage-grouse at mesic and xeric sites, respectively. We located a total of 354 (site average = 33) and 433 nests (site average = 59) within mesic and xeric sites. We measured fine-scale habitat characteristics for 327 and 415 nests within mesic and xeric sites. Additionally, we measured fine-scale habitat characteristics at 324 and 420 random locations within mesic and xeric sites to characterize availability for each habitat characteristic (Brussee et al. 2022a; Figure 2). Differences in availability between regions were statistically evident for 14 of the 23 variables measured (Brussee et al. 2022a; Figure 2). Perennial grasses and forbs had greater availability in mesic sites, with both cover and height characteristics greater on average in mesic sites. Residual grass height, cover, and litter cover were also greater in mesic sites, while percent bare ground was greater on average in xeric sites. In contrast, dwarf sagebrush and total sagebrush cover were greater in xeric sites, but nonsagebrush shrubs had greater cover and height in the mesic sites. Shrub height was greater in mesic sites (Brussee et al. 2022a; Figure 2).

Nest Site Selection
Variable reduction identified 8 variables that were important for nest site selection (mean w > 0.5) at either mesic or xeric sites and were included in the final model. These included horizontal cover at the nest, shrub width at nest, shrub height at nest, total sagebrush cover, average shrub height, perennial forb cover, perennial grass height, and bare ground.  Figure 3.

Nest Survival
In mesic sites, 161 nests hatched, and 188 nests failed. In xeric sites, 211 nests survived, and 205 nests failed. We excluded 5 nests from mesic sites and 17 nests from xeric sites with unknown fates from our nest survival analysis.
Variable reduction identified 11 variables that were important for nest survival (mean w > 0.5) at either mesic or xeric sites and were carried forward to the final model. These included horizontal cover, shrub width at nest measured perpendicular to maximum width, shrub height at nest, nonsagebrush shrub cover, perennial forb cover, perennial forb height, perennial grass height, annual forb cover, annual grass cover, residual cover, and residual grass height. Full variable reduction results are presented in Supplementary Material Appendix B Table 1. Based on final model results, 38-day nest survival was 0.24 (95% CRI: 0.08-0.43). We found evidence (pd = 0.97) that nest survival was higher for adult (0.27, 95% CRI: 0.09-0.50) compared to yearling sage-grouse (0.17, 95% CRI: 0.05-0.36), and support (pd = 0.89) for a positive effect (β init = -0.09, 95% CRI: -0.24 to 0.06) of initiation date on survival (Supplementary Material Appendix B Table 3).

DISCUSSION
While sage-grouse are widely regarded as a sagebrush obligate species, their occurrence and life history span heterogeneous sagebrush ecosystems (Crawford et al. 2004. Our results highlight how the effects of fine-scale habitat characteristics on sage-grouse occurrence and reproductive success in part depended on regional availability of vegetation components within sagebrush communities aggregated into mesic and xeric hydrographic regions based on long-term average annual precipitation. While selection for most fine-scale habitat characteristics, such as sagebrush cover, was consistent across hydrographic regions, we found the importance of grasses and forbs to vary across regions. For example, we identified stronger selection (i.e., larger coefficients) for tall perennial grasses in more xeric environments, where grass height is more limited, with resulting

10
Fine-scale habitat influences sage-grouse nest selection and survival Brussee et al. survival consequences. Namely, at xeric sites, nests associated with taller perennial grasses were more successful on average. These findings highlight the importance of considering how mesic vs. xeric site conditions influence underlying vegetation composition when developing management plans for at risk species across the sagebrush biome. Variation in species' response across hydrographic regions can be understood as a form of habitat functional response, where the selection of a local habitat type is a function of the regional prevalence of that habitat type when measured at a coarser scale (Mysterud and Ims 1998). In our study, sage-grouse selection patterns for several habitat covariates in part depended on their regional availability in xeric and mesic regions (Aarts et al. 2013, Holbrook et al. 2019. Mesic habitats within the Great Basin receive greater precipitation annually and thus have taller and more productive understories of grasses and forbs than xeric habitats. Differences in habitat availabilities were most notable for measurements of perennial grass and forb cover and height, as well as annual forb cover, where higher values were measured in mesic sites. On average, perennial grasses were approxi-mately 6.3 cm shorter, and forbs were ~4.1 cm shorter at xeric than at mesic sites ( Figure 2; Brussee et al. 2022a). Xeric sites also had greater bare ground (49.6 vs. 37.4%; Figure 2; Brussee et al. 2022a). Within xeric sites where grass cover is less and height is relatively short, our models indicated that sage-grouse selected nest sites with the tallest available understory grasses. This selection pattern was not clearly observed in mesic sites, likely because overall availability was greater, and sage-grouse presumably would not need to seek out taller grass understories to access adequate cover for nesting. Importantly, nest site selection by sage-grouse had implications for nest survival, where relationships between nest survival and fine-scale habitat characteristics in some cases exhibited similar deviations between mesic and xeric sites. The variation we found in effects of herbaceous cover within our study helps to explain inconsistencies in studies of grass-related variables on selection or survival of sage-grouse nests. For example, some studies throughout sage-grouse range have reported positive associations of grass-related variables on selection and/or survival of nests (Gregg et al. 1994, Sveum et al. 1998, Kirol et al. 2012, Doherty et al. 2014 whereas others have reported those associations to be weak (Holloran et al. 2005, Dinkins et al. 2016b or absent (Popham and Gutierrez 2003, Kolada et al. 2009, Coates and Delehanty 2010, Lockyer et al. 2015, Gibson et al. 2016). Inconsistencies among nest survival studies have since been attributed to variable confounding effects based on timing of measurement and plant phenology (Gibson et al. 2016, Smith et al. 2018b). However, here we accounted for plant phenology and found that grass effects varied between hydrographic regions and depended on grass species composition. Contradictory results may also be observed in studies that do not specify predominant herbaceous cover as annual grass (typically cheatgrass) or native perennial grass (e.g., squirreltail). For example, while we found strong influences of perennial grasses on nest survival, we did not find any influences of annual grass on nest survival across sites within xeric or mesic hydrographic regions. However, substantial negative impacts of annual grass on nest survival ) and population growth rates (Coates et al. 2016b), in association with previously burned areas, have been shown to occur when measured at a coarser resolution using remotely sensed measurements . These findings may be more fu7nctionally related to broad disturbances that result in state transition from shrub-dominated communities to those dominated by annual grassland (Chambers et al. 2014, Foster et al. 2019, Schuyler et al. 2022). The nuanced structural components of cover at fine scales could have consequences that become apparent across larger areas within the Great Basin region of the U.S. and warrant further investigation, especially with some environments undergoing shifts toward drier conditions with more substantial annual grass vegetation components (Blumenthal et al. 2016, Smith et al. 2021. Independent from vegetation cover type, visual concealment was consistently important for nest selection and survival across both mesic and xeric sites within the Great Basin, which could indicate that broad-scale management prescriptions for habitat characteristics may be appropriate in some contexts. For example, sage-grouse consistently selected greater overstory cover, specifically sagebrush cover, surrounding nests, and greater horizontal cover at the nest bowl, while avoiding bare ground. Furthermore, measurements of horizontal cover at the nest bowl were important for nest survival in both mesic and xeric habitats and greater shrub cover was associated with higher nest survival within mesic environments. Specifically, in mesic habitats, we found higher nest survival with greater non-sagebrush shrub cover, which was more prevalent at mesic sites than at xeric sites ( Figure 2). Sage-grouse nest survival in other, primarily mesic, locations have consistently shown positive responses to greater amounts of shrub cover (Gregg et al. 1994, Popham andGutiérrez 2003). Gaps in the canopy may have strong consequences for nest survival, as less structural horizontal and vertical closure allows increased detection by visually-cued predators such as Common Ravens (Corvus corax; Coates and Delehanty 2010, Dinkins et al. 2016a), and could perhaps reduce scent barriers for olfactory-cued terrestrial predators (Conover 2007).
Our models indicated that understory components comprising nest concealment had influences on survival that varied by hydrographic region. For example, within xeric environments, we found higher nest survival associated with taller perennial grasses. Relatively tall grasses and forbs off the nest but within the shrub interspace might help conceal females from predators as they move to and from their nests during incubation recesses , Dudko et al. 2019. Additionally, taller understory of perennial vegetation may help close the visible gap between understory and overstory vegetation, especially considering relatively short average understory vegetation heights and more bare ground in xeric environments. Within xeric habitats, reduced nest survival was associated with both increased annual forb cover and residual grass height. While areas with greater annual forbs may lack the structural characteristics of new perennial grass growth necessary to evade predation for successful nesting, the effect of residual grass height on nest survival may be temporally confounded by growing season. Specifically, in xeric habitats, early nests may benefit from residual grass prior to growth of new grasses. However, later in the growing season, as residual grasses decompose, new perennial grass growth provides structural cover and more effectively closes gaps in the canopy, resulting in increased nest survival. Collectively, our results indicate that inconsistencies in effects of understory vegetation may be at least partially attributed to differences between hydrographic regions, and varying vegetation productivity as expressed by the height and cover of understory, the availability of shrub overstory, and the growth form of sagebrush species as either tall and columnar or medium-short and spreading. A more thorough understanding of such interactions, along with ecosystem-and scale-dependent habitat responses will continue to provide critical information for conservation of species at risk within imperiled sagebrush ecosystems.
Our results highlight how moisture gradients influencing relative aridity may affect regional availability of vegetation components within habitat communities, while explaining subsequent variation or uniformity of their influences on sage-grouse habitat selection and reproductive success. Providing this context is particularly important given uncertainty regarding the importance of grass cover and height in promoting and maintaining quality nesting habitat for sagegrouse and other grassland species (Davies et al. 2006, Smith et al. 2018a, 2018b. Habitat selection patterns, and the consequences of habitat use, are known to be functionally dependent on regional habitat availability (Mayor et al. 2009, Laforge et al. 2016). However, these ecosystem complexities are rarely addressed in studies analyzing sage-grouse responses to fine-scale vegetation measurements (reviewed in Smith et al. 2020). Consequently, results that are highly generalized risk overlooking nuances in species-habitat relationships associated with regional variation, which may lead to management recommendations that are misaligned with local conditions. While we found that ecologically generalizable management solutions for habitat guidelines may be applicable for some vegetations characteristics, particularly sagebrush cover, we additionally demonstrate that coarse moisture gradients defined by long-term precipitation explain important differences in finescale vegetation requirements by sage-grouse. Such patterns cannot be detected when pooling data across broad landscapes due to spatial homogenization in modeling efforts. Furthermore, coarse-scale data using remotely sensed land cover characteristics are more likely to reflect generalized species requirements Brussee et al. Fine-scale habitat influences sage-grouse nest selection and survival 13 across broader regions, such as proportions of sagebrush cover, or to identify landscape changes such as conversion of shrubland to grassland. Our study supports the notion that fine-scale habitat measurements are important for more accurately characterizing understory conditions (i.e., grass height) that are influenced by moisture gradients and site aridity, subsequently affecting nest survival and perhaps population performance. Ultimately, this research addresses needs of land managers in developing management prescriptions for sage-grouse within state wildlife and federal land use plans using multi-scale habitat assessments (Stiver et al. 2015) and understanding sage-grouse habitat requirements across a range of site conditions.

Supplementary material
Supplementary material is available at Ornithological Applications online.