Introduction

Widespread degradation of sagebrush (Artemisia spp.) ecosystems in western North America poses an immense conservation challenge (Knick et al., 2003; Mac et al., 1998; Noss et al., 1995). Few of these landscapes remain intact post Euro-American settlement (Miller and Eddleman, 2001; West, 1996). Fragmentation has accelerated in past decades, driven by invading annual grasses, encroaching conifer, cultivation of native rangelands, and expanding rural urban and industrial development (Miller et al., 2011). It is unlikely sagebrush systems are to return to presettlement condition as scale and magnitude of these changes greatly exceeds available financial and logistical resources (Miller et al., 2011). To curb future loss, conservation strategies should focus on remaining intact sagebrush landscapes to maximize return on limited conservation investments (Bottrill et al., 2008; Hobbs and Kristjanson, 2003; Joseph et al., 2009).

Efforts to conserve sagebrush landscapes are driven largely by investments that benefit greater sage-grouse (Centrocercus urophasianus; hereafter, “sage-grouse”) populations and their habitats, with the aim of precluding the need for an Endangered Species Act (ESA; Meinke et al., 2009) listing. Sage-grouse is a gallinaceous species endemic to sagebrush communities of western North America (Schroeder et al., 1999). Degradation and loss of sagebrush shrublands have contributed to extirpation of the species from half its historic range and declines in many remaining populations (Schroeder et al., 2004). In 2010, heightened concern over the species population status resulted in a designation of warranted but precluded by the US Fish and Wildlife Service (USFWS) under the ESA (USFWS, 2010). Reevaluation of this designation in 2015 helped to stimulate more than 1.2 million ha of additional conservation to reduce primary threats impacting sage-grouse (USFWS, 2015). These actions were a contributing factor in the 2015 decision not to list sage-grouse as threatened or endangered (USFWS, 2015).

Conservation of sagebrush ecosystems continues to accelerate on private (NRCS, 2015a) and public lands (USFWS, 2015) in anticipation of the USFWS revisiting their sage-grouse listing decision in 2020. The Bureau of Land Management (BLM) and US Forest Service (USFS) have committed additional resources to conservation on federal lands encompassing more than half the current occupied range of sage-grouse (BLM and USFS, 2015). Conservation investments are being steered toward landscapes that support high-abundance sage-grouse populations (USFWS, 2015). Federal, state, and private land managers anticipate that broad-scale reduction in primary threats will benefit a suite of sagebrush-associated taxa (Copeland et al., 2014; NRCS, 2015a; Stiver et al., 2006; USFWS, 2013).

Extending benefits of single-species conservation to multiple taxa assumes spatial correlation across species (Andelman and Fagan, 2000; Caro and O'Doherty, 1999; Prendergast et al., 1993; Simberloff, 1998). Sage-grouse distributions are well known and exhibit highly clustered patterns (Knick and Hanser, 2011) that concentrate 75% of breeding populations within a quarter of the species range (Doherty et al., 2010b). Their broad distribution and obligate status have implicated sage-grouse as an indicator species of sagebrush ecosystem health (Rich et al., 2005). However, limited understanding of their spatial coincidence with other species (e.g., common sagebrush lizard [Sceloporus graciosus], pigmy rabbit [Brachylagus idahoensis]) has restricted implementation of multispecies conservation efforts within sagebrush ecosystems (Rowland et al., 2006).

Passerine species endemic to sagebrush ecosystems—Brewer's sparrow (BRSP; Spizella breweri), sagebrush sparrow (SAGS; Artemisiospiza nevadensis), and sage thrasher (SATH; Oreoscoptes montanus)—are among the fastest declining bird groups in North America (NABCI, 2014). Like sage-grouse, these sagebrush-obligate songbirds (hereafter “songbirds”) are considered important predictors of ecosystem condition because of their sensitivity to local and landscape-scale habitat change (Knick et al., 2003; Rotenberry and Wiens, 2009). The same threats facing sage-grouse are also linked to declines in songbirds; that is, conifer expansion (Knick et al., 2014), wildland fire (Knick et al., 2005), cultivation of grazing lands (Vander Haegen, 2007), invasion of exotic annual grasses (Earnst and Holmes, 2012), and energy development (Gilbert and Chalfoun, 2011; Mutter et al., 2015). Range overlap among breeding songbirds and sage-grouse is extensive but has proven to be a poor correlate of co-occurrence (Rich et al., 2005).

The unprecedented scale of sage-grouse conservation provides the opportunity for community-level benefits in other sagebrush-obligate species (Boyd et al., 2014; Davies et al., 2011). Sage-grouse planning efforts have assumed broader ecosystem and multiple species benefits (NRCS, 2015b; USFWS, 2015) but lack much of the empirical evidence needed to test assumptions of the range-wide strategies proposed (Rowland et al., 2006). We addressed this information gap by 1) identifying spatial patterns of sagebrush-obligate songbird abundance in relation to sage-grouse distribution across their range and 2) evaluating the potential for community-level benefits derived from targeted sage-grouse conservation. We first modeled spatial variability in relative abundance for the songbirds BRSP, SAGS, and SATH using count data from the North American Breeding Bird Survey (BBS; Pardieck et al., 2015). We then evaluated the dependence of songbird abundance on known sage-grouse distributions using point pattern analysis and summarized results within Western Association of Fish and Wildlife Agencies Sage-Grouse Management Zones I–VII (hereafter “management zones”; Stiver et al., 2006, Fig. 1). Lastly, we evaluated patterns of relative songbird abundance to coincidence of sage-grouse habitat restoration (invasive conifer removal) and landscape protection actions (i.e., Wyoming's Sage-Grouse Core Area Strategy [see Copeland et al., 2013] and Sage-Grouse Fire and Invasive Assessment Tool priority areas [FIAT; BLM, 2014]).

Methods

Modeling Spatial Variability in Sagebrush-Obligate Songbird Abundance

We defined our modeling extent (1.7 million km²; see Fig. 1) using a moving 50-km² sample frame to identify areas containing ≥ 1% sagebrush land cover (i.e., sagebrush shrubland) in the western United States, similar to methods used by Knick et al. (2003). Sagebrush extent was derived from LANDFIRE 90 m existing vegetation type dataset (LANDFIRE, 2012). Six sagebrush communities were used to define sagebrush extent: Great Basin xeric mixed sagebrush shrubland, Intermountain basins big sagebrush shrubland, Columbia Plateau low sagebrush steppe, intermountain basins big sagebrush steppe, intermountain basins montane sagebrush steppe, and Columbia Plateau silver sagebrush seasonally flooded shrub-steppe (Nature Serve Explorer, 2012). Sagebrush shrublands in northern Arizona and New Mexico were omitted, as there was no potential for overlap within the range of sage-grouse.

We characterized patterns of relative abundance for BRSP, SAGS, and SATH during the breeding season. These species are ideal to test the ability of sage-grouse conservation to provide community-level benefits as they are 1) sufficiently ubiquitous to develop range-wide models of abundance, 2) broadly encompassed within the sage-grouse range such that conservation actions have the potential for overlap, and 3) identified as species of conservation concern (USFWS, 2008).

Count data from BBS (Pardieck et al., 2015)were used as an index to variation in abundance for each species. Counts occur in the spring on routes typically 40 km in length along secondary roadways. Volunteers conduct 3-min point counts annually at 50 sites spaced ∼800 m apart along the route. Routes that did not meet the standards of BBS trend analyses (e.g., surveys conducted in inclement weather, occurring outside allotted time period) were omitted from analyses.

A total of 523 BBS routes were encompassed in our sampling frame. To reduce the effects of annual variation in bird abundance measured, we averaged total counts for each route by species across a 10-yr period (2004–2014). These 10-yr means were then used as a response variable for model-based analyses to provide an index to relative abundance (hereafter, “abundance”). Routes not surveyed within this period were omitted. A 10% subset of routes used in the analysis (n = 52)were randomly selected and withheld for model evaluation.

Songbird abundance is structured along multiscale ecological gradients that converge along patterns of land cover, landform, climate, and human disturbance (Knick et al., 2008). Using available spatial data, we applied a suite of landscape predictor variables to model distributions of BRSP, SAGS, and SATH. While each songbird species relies on intact sagebrush ecosystems during the breeding season, their habitat needs differ in structural and vegetative conditions at more local scales (Wiens et al., 1987).We accounted for this by including large-scale climate and productivity predictors that may give rise to habitat variation and observed differences in breeding songbird abundance within sagebrush ecosystems (Table 1).

We aimed to summarize covariate values that characterized the hierarchical process of habitat selection inherent in birds (Wiens et al., 1987), so we summarized each variable at a local (120 m²) and landscape (6.4 km²) spatial scale. Focal mean and standard deviation of covariate values were calculated by summarizing representative raster data at each scale to capture the central tendency and heterogeneity of habitat variables; however, for climate variables we simply used the raw value, as the spatial resolution was typically large (N 1 km²; see Table 1). To append covariates to sample units, we buffered each route by 200 m, a distance for which each species had a nonzero probability of being detected using point counts (J. Tack unpublished data), and used the mean value of cells implicated within a buffered route.

Figure 1.

Study area encompassing>96% of sagebrush land-cover in western North America (>1.7 million km²). Lek locations are representative of greater sage-grouse distributions in 10 of 11 western states containing populations (California, Idaho, Montana, Nevada, North Dakota, Oregon, South Dakota, Utah, Washington, and Wyoming; USA).Western Association of Fish and Wildlife Agencies Sage-Grouse Management Zones (I-VII) are representative of geographical population segments containing common ecological setting and stressors (Stiver et al., 2006). Small portions of sagebrush lands in northern Arizona and New Mexico were omitted, as there was no potential for overlap within the range of sage-grouse.

We modeled the relationship between counts and environmental predictors for songbirds using Random Forest regression algorithms (Breiman, 2001). Random Forest uses a machine-learning approach and has demonstrated the ability to produce more accurate predictive models than generalized linear model (GLM) parameterizations, largely due to the ability to fit highly complex nonparametric interactions between predictors (Breiman, 2001). This modeling approach is less sensitive to many issues that confound inference from GLMs for use in species distributional models including collinearity among predictors while remaining robust to overfitting (Culter et al., 2007). We fit models using 5 000 trees following Random Forest model selection with the rf Utilities (parsimony value = 0.3) and Random Forest packages in program R (R Core Team, 2015). Spatial predictions were projected to 16.6-km² grid cells to generate continuous abundance surfaces for each songbird species. Grid cell size was roughly equivalent to sample unit area (200-m buffered BBS route). We evaluated the predictive capability of models by comparing the predicted to observed abundance classes using both model predictions and withheld data.

To aid data summary and data visualization, we grouped predicted songbird abundance surfaces into four classes. Class membership was determined by ranking songbird predictions (grid cells) from high to low and summarizing results into four bins, each containing 25% of the total predicted abundance. We classified these bins as “high,” “moderate,” “low,” and “sparse.” We used these classes for mapping products and as a metric to quantify co-occurrence estimates for sage-grouse and songbirds.

Dependence of Songbird Abundance on Sage-Grouse Distributions

Sage-grouse distributions were modeled as a first step in evaluating songbird dependence. Results applied as a predictor variable to evaluate spatial co-occurrence with BRSP, SAGS, and SATH abundance. Models were derived from the sage-grouse lek survey and location data. Lek surveys have been widely used by resource agencies to monitor trends in sage-grouse populations and are considered an index of relative distribution and abundance (Reese and Bowyer, 2007). Leks are sites where male and female sage-grouse congregate in the spring to breed and are typically the focal point for conservation actions targeting the species (Connelly et al., 2000). High fidelity to leks and surrounding nesting sites are typical in sage-grouse with birds congregating at the same location each year (Connelly et al., 2011). We presumed all lek locations to be fixed within the context of broad species patterns examined in this study but acknowledge that shifts in lek locations may occur due to persistent disturbance or alteration of vegetative cover (Hovick et al., 2015;Walker et al., 2007).

Table 1

List of predictor variables considered for predictive models of sagebrush obligate songbird counts. To characterize hierarchical selection of sagebrush-obligate birds, we summarized variables within a local (120-m) and landscape (6.4-km) scale. Variables were then measured using the focal mean (mean) or standard deviation (SD) of values within the sample unit of 200-m buffered Breeding Bird Survey routes

To account for high variability in survey intervals and sampling intensity, we used the average of annual maximum male counts for known lek locations over a 10-yr period (2004–2014) as the basis for estimating distributions and population abundance. Lek counts were collected and provided by 10 of 11 state wildlife agencies where sage-grouse are found (California, Idaho, Montana, Nevada, North Dakota, Oregon, South Dakota, Utah, Washington, and Wyoming). Only leks averaging one or more males counted within the 10-yr period examined were included in the analysis (n = 6 272). Lek counts that averaged less than one male (n = 3101) were omitted from analyses. All leks included in the study were assumed to be active through the period examined (2004–2014).

Because sage-grouse abundance is used to prioritize conservation areas (USFWS, 2015), leks were segmented into high and low abundance classes and applied as categorical factors to examine coincidence of songbird and sage-grouse distributions. We considered leks clustered within the known dispersal radius of nesting females (6.4 km; Colorado Division of Wildlife, 2008) to be a single breeding group, similar to Doherty et al. (2010b). Leks occurring within common 6.4-km radii were assigned a weighted value equal to the sum of their male counts using a kernel function applied in program R (R Core Team, 2015). Weighted values for leks not occurring within 6.4 km of another were generated by assigning a value equal to their non weighted survey count. Starting with the highest weighted lek values, we then summed the number of counted males (non weighted) until 50% of the total count was reached. This resulted in a spatial segmentation of high and low abundance populations (i.e., leks; Doherty et al., 2011; Fig. 2).

We evaluated the association of BRSP, SAGS, and SATH abundance to sage-grouse distributions using the Berman test implemented in program Programita (Wiegand and Maloney, 2004;Wiegand and Maloney, 2014). By first developing a null (e.g., random) model, the Berman test provides a goodness-of-fit measure between a univariate point pattern (i.e., leks) and a continuous spatial covariate (i.e., predicted abundance of songbirds; Berman, 1986). The resulting test statistic (Z1), allows for a statistical comparison between observed and null models. For our practice, the Berman test provides a framework to test the hypothesis that sage-grouse lek locations can be better described by the predicted abundance of sagebrush-obligate songbirds than random.

The original Berman test assumes a homogeneous point process model that does not consider spatially clustered patterns, such as those found in sage-grouse lek distributions (Knick and Hanser, 2011), and tends to overpredict the significance of spatial associations (Wiegand and Maloney, 2004).We accounted for this by using stochastic null models (n = 999), which retained the spatial clustering evident in sage-grouse lek locations (Wiegand and Maloney, 2014)^{. M}odel simulations maintained the observed clustering structure in lek locations by approximating model fit (α b 0.05) to multiple-point pattern summary statistics, as outlined in Wiegand et al. (2013). Models were developed and simulated separately for high-abundance leks. Procedures were replicated independently for population segments defined by sage-grouse management zones (Stiver et al., 2006; see Fig. 1) to account for variation in songbird abundance and sage-grouse distribution patterns throughout their range.

Figure 2.

Predicted (BRSP) Brewer's sparrow, (SAGS) sagebrush sparrow, and (SATH) sage thrasher abundance shown within Western Association of Fish and Wildlife Agency Sage-Grouse Management Zones (MZ). Songbird abundance is displayed by high, moderate, low, and sparse. Class membership was determined by ranking predicted songbird abundance from high to low and summarizing results into four bins each containing 25% of the total. High abundance sage-grouse leks are representative high bird densities containing 50% of the known breeding population.

Songbird Coincidence with Sage-grouse Landscape Protection

Coincidence of sage-grouse landscape protection with songbird abundance was assessed using Wyoming's sage-grouse core area strategy (State of Wyoming Executive Department, 2011) and sage-grouse FIAT priority areas (BLM, 2014; see Fig. 3). These actions are stratified geographically between the more productive eastern Rocky Mountain sagebrush environments (i.e., Wyoming) and dryer sage-steppe ecosystems of the Great Basin (i.e., FIAT). The efforts are mutually exclusive and provide an opportunity to examine potential songbird benefits under scenarios addressing landscape stressors unique to each region. Wyoming's approach primarily focuses on human-induced fragmentation (e.g., rural subdivision and energy development) that has resulted in sharp reductions in sagebrush-obligate populations (Bayne et al., 2008; Braun et al., 2002; Doherty et al., 2011; Gilbert and Chalfoun, 2011; Ingelfinger and Anderson, 2004; Mutter et al., 2015; Stiver et al., 2006; USFWS, 2013; Walker et al., 2007). FIAT in the Great Basin attempts to address habitat loss and fragmentation due to wildfire and invasive plants that are widely recognized as a significant threat to sage-grouse (Miller et al., 2011; Stiver et al., 2006; USFWS, 2013) and songbirds (Knick et al., 2005).

Figure 3.

Predicted (BRSP) Brewer's sparrow, (SAGS) sagebrush sparrow, and (SATH) sage thrasher shown within Western Association of Fish and Wildlife Agency Sage-Grouse Management Zones (MZ). Songbird abundance is displayed by class; high, moderate, low, and sparse. Class membership was determined by ranking predicted songbird abundance from high to low and summarizing results into four bins each containing 25% of the total within the study area. Points representative of sage-grouse conifer removal treatments on private lands that were completed between 2010–2015 or are slated for completion by 2021. Wyoming's sage-grouse core areas and FIAT (Fire and Invasive Assessment Tool) priority areas (polygons) include private, state, and federally owned lands within Wyoming and the Great Basin region.

We assessed songbird coincidence with Wyoming's core areas strategy by summarizing BRSP, SAGS, and SATH abundance surfaces within its boundaries (6.2 million ha; see Fig. 3). Summaries were stratified by songbird abundance classes to evaluate proportion of high to moderate populations occurring within. We then evaluated summaries to determine the area of high to moderate songbird abundance encompassed by core area boundaries relative to the total high to moderate population areas estimated within the state of Wyoming. Analyses were completed using the zonal statistical function in program ArcGIS.

Songbird coincidence with sage-grouse FIAT priorities was assessed following procedures outlined previously for Wyoming's core area strategy. Proportion of high to moderate songbird abundance was evaluated within FIAT priority polygons (16.4 million ha), and summaries were compared with high to moderate totals within the entire Great Basin (i.e., management zones III–V; see Fig. 3).

Results

Sagebrush land-cover variables were consistently the most important predictors in explaining variation for songbird counts, as two of the top three variables for each species contained some measure of sagebrush (Table 2). A partial dependence plot of all sagebrush land-cover types within 120 m, the second most important variable across species, shows that the relationship between songbird counts and sagebrush is positive and nonlinear. Furthermore, there is an apparent threshold across species at a mean of 40% sagebrush lands-cover (Fig. 4). The distribution of this particular covariate across the landscape is highly skewed, with > 76% of sampled sites containing values of “all sagebrush” lower than the apparent threshold. While sagebrush was an important predictor in models, climate (drought indices and temperature) and landscape topography and elevation were influential in predicting counts across species (see Table 2).

Table 2

Variable importance. Ranked variable importance from random forest models of songbird abundance for Brewer's sparrow (BRSP), sagebrush sparrow (SAGS), and sage thrasher (SATH). Importance was determined by percent increase in mean squared error (MSE) when variable was included in bootstrap samples. Dashed entries indicate that a variable was not included in a random forest model following model selection

Figure 4.

Partial dependence plot displays the predicted relationships between the proportion of all sagebrush classes within 120 m (AllSB120m) and predicted counts of Brewer's sparrow (BRSP), sagebrush sparrow (SAGS), and sage thrasher (SATH). The background histogram represents the frequency of values for all sagebrush landscape cover (see Table 1; AllSB120m) contained within the study area.

Figure 5.

Stacked bar graph compares measured Brewer's sparrow (BRSP), sagebrush sparrow (SAGS), and sage thrasher (SATH) abundance means at lek sites (white) to predicted mean abundance (black/gray) generated using null model lek distributions. Comparisons were made among high density leks (black) and all leks (gray) by management zone (I-VI) and for all management zones combined (ALL). Management zone VI did not contain any high abundance sage-grouse leks. Right axis representative of songbird abundance classes. Class membership was determined by ranking predicted songbird abundance from high to low and summarizing results into four bins each containing 25% of the total within the study area.

Figure 6.

Stacked bar graph depicting relative coincidence of Brewer's sparrow (BRSP), sagebrush sparrow (SAGS), and sage thrasher (SATH) contained within (a) conifer removal treatments (>180,000 ha) (b) Wyoming's sage grouse core areas (6.2 million ha), and (c) FIAT (Fire and Invasive Assessment Tool) priority areas (16.8 million ha). Coincidence was estimated as a percent area encompassed by abundance classes; high, moderate, low, and sparse.

Models for BRSP, SAGS, and SATH explained 17%, 17%, and 37% of the variation in BBS count data, respectively, when comparing out-of sample data with training data within the Random Forest algorithm. All species models were good classifiers of the lowest abundance threshold for predicted and withheld data (> 76%), while misclassification rates increased with a decreasing sample size of the largest abundance bins. However, routes were more likely misclassified as adjacent abundance bins (e.g., lowest abundance class misclassified as second lowest abundance class).

Berman's test suggested that range-wide measures of songbird abundance were higher at lek sites (P b 0.0001, Z₁ = 9.6 [BRSP], 9.4 [SAGS], 9.9 [SATH]) than null model predictions, indicating that BRSP, SAGS, and SATH abundances were positively associated with lek locations (Fig. 5). High-abundance sage-grouse leks (i.e., 50% of the population containing leks with the highest male counts) averaged 15% (BRSP), 13% (SAGS), and 19% (SATH) higher songbird abundance versus all sage-grouse leks. Across their range, sage-grouse distributions were associated with higher BRSP abundance compared with SAGS and SATH (see Fig. 5).

Spatial variability in songbird abundance patterns influenced population effect size and rates of co-occurrence among songbirds and sage-grouse populations by management zone. Lek locations were associated with higher songbird abundance (P b 0.001, = 5.6 [BRSP], 5.4 [SAGS], 5.5 [SATH]) in all but zone VI (P ≥ 0.29, Z₁ = -1 [BRSP], 0.6 [SAGS], -0.7 [SATH]; see Figs. 2 and 5). Low songbird predictions in management zones I and VI resulted in measured abundance rates 80% (BRSP, SAGS) and 86% (SATH) below management zones II–V (see Figs. 2 and 5). Sage-grouse populations were associated with highest BRSP and SAGS abundance in management zones III and V (see Figs. 2 and 5). Associated SATH abundance was highest in zone V (see Figs. 2 and 5).

Table 3

Percentage of separate high to moderate songbird populations encompassed by sage-grouse landscape protection strategies, Wyoming's core area strategy, and Fire and Invasive Assessment Tool priorities within the entire state of Wyoming and the Great Basin (see Fig. 3)

Coincidence of songbird abundance with conifer treatments, Wyoming's core area strategy, and FIAT priority areas were similar to co-occurrence trends observed with sage-grouse distributions. Greater than 85% of conifer cuts targeting sage-grouse habitat restoration occurred within high to moderate BRSP abundance areas (see Figs. 3 and 6). Conversely, only ∼30% of sage-grouse conifer treatments occurred within high to moderate abundance SAGS and SATH areas (see Figs. 3 and 6).Wyoming's core areas and FIAT priorities contained a greater proportion of high to moderate abundance areas for SAGS (49–51%) and SATH (52–58%) than observed in conifer treatments (see Figs. 3 and 6). Core area and FIAT boundaries encompassed 40% and 39% (BRSP), 48% and 51% (SAGS), and 50% and 55% (SATH), respectively, of high to moderate abundance songbird areas occurring separately within Wyoming and the Great Basin (see Fig. 3; Table 3).

Discussion

We provide the first landscape view of sagebrush-obligate songbird distribution and abundance linked to ecological gradients across the western United States. Relating bird abundance to ecological patterns explained 19–37% of variation in count data. Resulting spatially explicit surfaces offer a tremendous improvement to the resolution in predicted bird distributions over traditional BBS distribution maps. Our continuous estimates of songbird abundance are consistent with coarse distributions of breeding ranges of each species (Martin and Carlson, 1998; Reynolds et al., 1999; Rotenberry et al., 1999). Sagebrush land-cover predictors were primary determinants of songbird abundance (see Table 2). Each songbird model independently inferred a threshold of > 40% sagebrush landscape cover (i.e., percentage of sagebrush shrubland occurring within a 16.6-km² sample grid) as an ecological minimum above which songbird abundances nearly doubled (see Fig. 4); predictions were further enhanced by attributes explaining drought, temperature, topography, and elevation. Previous sage-grouse research shows 90% of active leks are set in landscapes with > 40% sagebrush land cover (Knick et al., 2013), and high probability of lek persistence is associated with > 50% land cover (Wisdom et al., 2011), further suggesting long-term viability of songbird and sage-grouse breeding habitats may be closely linked through this common landscape requisite. Not surprisingly, our findings are also consistent with past regional-scale evaluations that identify large patch size and continuous sagebrush land cover as factors predictive of sagebrush songbird distributions (Knick and Rotenberry, 1995;Wiens et al., 1987).

Our point-process analyses quantitatively support the long-held notion that sage-grouse distributions spatially predict songbird abundance in sagebrush ecosystems. Populations for each of three species aligned themselves with sage-grouse distributions (P < 0.0001, Z1 > 9.4; see Figs. 2 and 5). Range-wide, landscapes containing > 50% of males on leks also supported 13–19% higher songbird abundance compared with all leks. Patterns in sage-grouse and songbird configurations converged within landscapes of high sagebrush land cover (see Fig. 4) with concentrations largely centered within sage-grouse management zones II–V (see Fig. 2). Previous studies demonstrate the importance of landscape features in shaping songbird (Knick et al., 2008) and sage-grouse distributions (Doherty et al., 2010a; Donnelly et al., 2016) and identify common threats influencing their distribution and abundance. In particular, sage-grouse and songbirds are sensitive to anthropogenic disturbance including energy development (Gilbert and Chalfoun, 2011; Walker et al., 2007), cultivation (Knight et al., 2014; Tack, 2009), conifer encroachment (Baruch-Morodo et al., 2013; Holmes et al., 2017-this issue), and exotic annual grass invasion (Earnst and Holmes, 2012).

In addition to quantifying range-wide alignment, our spatial analyses also identified regional variation in songbird abundance that is best explained by common threats facing grouse and passerines. Most apparent are differences between desert sage-steppe habitats of the Great Basin (management zones III–V) and the more eastern and productive Rocky Mountain sagebrush environments (management zones I, II; see Figs. 2 and 5). In the Great Basin, sage-grouse and songbirds are caught within the “big squeeze” of expanding conifer impacts at mid to upper elevations and catastrophic wildfire and cheatgrass invasion in dryer low-lying habitats (Chambers et al., 2014b; Davies et al., 2011). Targeted tree removal in the mid to upper elevations has expanded availability of sage-grouse nesting habitat (Severson et al., 2016) and increased BRSP abundance by + 55% (Holmes et al., 2017-this issue). Our co-occurrence analyses found 85% of restorative conifer cuts, all conducted in a manner similar to Holmes et al., 2016, coincide with areas of high to moderate BRSP abundance, suggesting comparable benefits may be accruing over roughly 180 000 ha of the Great Basin (see Figs. 3 and 6). Outcome-based evaluations following tree removal (Holmes et al., 2016) corroborate previous studies that suggested BRSP as a primary beneficiary of sagebrush restoration occurring through conifer removal (Rowland et al., 2006; Hanser and Knick, 2011).

Findings suggest that reducing wildfire and cheatgrass threats will benefit songbird composition and abundance different from that of conifer removal. Greater than 50% of FIAT priorities were made up of areas containing high to moderate SAGS and SATH abundance (see Fig. 6). These areas also represented 51% and 55% of the total estimated SAGS and SATH abundance within the Great Basin (see Table 3). This compares to conifer restorations that occurred within high to moderate abundance areas of these same species only ∼30% of the time (see Fig. 6) largely because SAGS and SATH occupy dryer and lower elevation sites (Martin and Carlson, 1998; Reynolds et al., 1999) less affected by conifer encroachment (Miller et al., 2008; Davies et al., 2011). High rates of co-occurrence in FIAT landscapes are where sage-grouse conservation actions are most likely to translate into community-level benefits for SAGS and SATH (Hanser and Knick, 2011).

Threats facing sage-grouse and songbirds in the Rocky Mountain states (management zones I, II; see Fig. 2) are primarily anthropogenic including energy development (Harju et al., 2010; Gilbert and Chalfoun, 2011), subdivision, and cultivation of sagebrush rangelands. As 1 of 11 sage-grouse states, Wyoming contains 37% of the species-wide distribution. Wyoming's core area strategy focuses on minimizing landscape stressors within a quarter of the state's land base containing 80% of the state's grouse population (see Fig. 3). Since 2010, $147 million have been invested in conservation easements to reduce subdivision impacts in core areas (Copeland et al., 2013a; NRCS, 2015c). Co-occurrences show that restrictive energy policy and easements are also helping reduce fragmentation across ∼40–50% of the areas containing high to moderate songbird abundance in the state (see Table 3). Habitat diversity captured within Wyoming's landscape protection strategies likely explain elevated coincidence of sage-grouse protections with both SAGS and SATH (Wiens et al., 1987). Low co-occurrence in other Rocky Mountain states (i.e., management zone I) is more an artifact of geography than of habitat quality or threats (Miller et al., 2011); states such as eastern Montana and the Dakotas lay within a transition zone from sagebrush to prairie ecoregions. Intact sagebrush landscapes in these geographies may support high abundance of songbird species beyond the scope of our evaluation (e.g., lark bunting [Calamospiza melanocorys] and western meadowlark [Sturnella neglecta]; Lipsey, 2015).

Implications

To date, sagebrush songbird conservation has garnered comparatively less attention than sage-grouse. Conservation partners predict that by 2020 habitat restoration and protection actions targeting sage-grouse will increase by > 35 million ha (NRCS, 2015b; USFWS, 2015). The future of sagebrush songbirds may hinge in part on our ability to identify mutually beneficial outcomes generated through sage-grouse conservation, as suggested by the 55–85% increase in passerine abundance in conifer cuts (Holmes et al., 2016) designed to expand the availability of sage-grouse nesting habitat (Severson et al., 2016). We offer our mapping products as additional decision support tools to further accelerate the targeting and integration of community-level benefits resulting from sage-grouse conservation. Newly available spatial datasets mapping conifer extent and density (Falkowski et al., 2017-this issue) can be combined with our predictive songbird maps to further enhance practitioners' ability to regionally target ongoing sagebrush habitat restorations.