Identifying Holes in the Greater Sage-Grouse Conservation Umbrella

Abstract

The umbrella species concept, wherein multiple species are indirectly protected under the umbrella of a reserve created for one, is intended to enhance conservation efficiency. Although appealing in theory and common in practice, empirical tests of the concept have been scarce. We used a real-world, semi-protected reserve established to protect a high-profile umbrella species (greater sage-grouse [Centrocercus urophasianus]) to investigate 2 potential mechanisms underlying the concept’s successful application: reserve size and species similarity. We estimated how much habitat protection the established reserve provided to 52 species of conservation concern associated with vegetation communities where greater sage-grouse occur. To illustrate the importance of reserve size, we compared the effectiveness of the established reserve to alternative greater sage-grouse reserves of various sizes and to simulated reserves of equal size but sited with no regard for greater sage-grouse. We further assessed whether key species’ traits were associated with different levels of protection under the umbrella reserve. The established umbrella reserve protected 82% of the state’s greater sage-grouse population and 0–63% of the habitat of the background species examined. The reserve outperformed equally sized, simulated reserves for only 12 of 52 background species. As expected, larger alternative reserves served as better umbrellas, but regardless of reserve size, not all species received equal protection. The established reserve was most effective at protecting the habitat of species that were most similar to the umbrella species (i.e., avian species, those highly associated with sagebrush plant communities, and those with widespread habitat). In contrast, the habitat of species with restricted distributions, particularly when combined with vegetation associations not closely matching the umbrella species, was not protected as well by the umbrella reserve. Such species require additional, targeted attention to achieve conservation objectives. Successful application of the umbrella species concept requires careful consideration of the characteristics of the umbrella species, the reserve delineated on its behalf, and the similarity of the umbrella species to its purported background species.

Article Summary:

A large conservation area protected for greater sage-grouse had limited utility in overlapping the habitat of 52 other wildlife species of conservation concern associated with sagebrush plant communities in Wyoming. The conservation area was most effective at protecting the habitat of species that were most similar to greater sage-grouse (i.e., avian species, those highly associated with sagebrush plant communities, and those with widespread habitat) but was typically less effective than hypothetical conservation areas sited randomly.

Conservation is one of the principal aims of wildlife management (Caughley 1977) and entails the protection and perpetuation of entire biological communities and ecosystems (Soulé 1985). Wildlife management has traditionally taken a single-species approach (Krausman 2002); however, surrogate species conservation strategies, where one species is used to represent other species or aspects of the environment to achieve conservation objectives, hold promise as a simplified means of effective conservation practice (Caro 2010). The umbrella species concept is one type of surrogate species strategy wherein a species with large area requirements (the umbrella species), if provided sufficient protected habitat, facilitates the protection of many other species (Wilcox 1984, Noss 1990). The main advantage of an umbrella species conservation strategy is the potential to conserve numerous species without extensive, individual consideration of each (Caro 2003). We follow the terminology of Caro (2003, 2010) to refer to the species that live in the same geographical area as an umbrella species (and therefore the species expected to benefit from the conservation of the umbrella species) as background species. Despite some criticism (Andelman and Fagan 2000, Roberge and Angelstam 2004, AMEC Environment & Infrastructure 2014), umbrella and other surrogate species conservation strategies are becoming increasingly popular in conservation practice (Caro 2003, 2010, U.S. Fish and Wildlife Service [USFWS] 2015a). Indeed, such surrogate-based strategies, including umbrella species, are suggested to play an increased role in strategic, nationwide efforts to manage wildlife and conserve ecosystems in the United States (USFWS 2015a).

Possessing a large area requirement is a well-established criterion for a suitable umbrella species (Noss 1990, Seddon and Leech 2008), so reserves created for umbrella species are, by definition, relatively large. One of the seminal concepts in biogeography, moreover, is that larger areas harbor greater numbers of species (Cain 1938, MacArthur and Wilson 1967). Whether conservation strategies centered on umbrella species are effective at conserving background species because of the characteristics of the selected umbrella species, or simply because these strategies inherently involve protecting large areas, however, remains an open question. More informally, is it the umbrella species or the space that matters? Moreover, much research on the umbrella species concept has sought to identify the traits of umbrella species that lead to effective surrogacy, with little regard for the traits of background species (Roberge and Angelstam 2004, Seddon and Leech 2008, Caro 2010). Understanding which traits predispose background species to coverage under an umbrella strategy will be important to the overall success of conservation based on the umbrella species concept.

We used the sagebrush-steppe ecosystem of western North America as a model system for addressing these gaps in understanding and application of the umbrella species concept. The sagebrush steppe has faced extensive degradation resulting from human land-use practices (Knick et al. 2003, Davies et al. 2011), is poorly represented within current conservation reserves (Scott et al. 2001), and is vulnerable to future anthropogenic development and climate change (Pocewicz et al. 2014). The sagebrush-steppe ecosystem is home to nearly 630 plant and animal species of conservation concern (Rich et al. 2005), many of which are endemic to sagebrush steppe and have received marginal conservation attention (Knick et al. 2003, Rich et al. 2005).

The greater sage-grouse (Centrocercus urophasianus; sage-grouse) has been the focus of much conservation attention in western North America in recent decades and has the potential to serve as an umbrella species for the conservation of other components of the sagebrush-steppe ecosystem, especially sagebrush-associated wildlife species (Rich and Altman 2001, Rowland et al. 2006, Hanser and Knick 2011, Gamo et al. 2013, Copeland et al. 2014). Grater sage-grouse are listed as endangered in Canada under the federal Species at Risk Act and have been petitioned ≥7 times for listing under the United States Endangered Species Act (ESA; Stiver 2011), though the USFWS recently determined that ESA listing was not warranted (USFWS 2015b). Each of the 11 states and 2 Canadian provinces where sage-grouse occur has a strategic plan to manage the species (Stiver 2011), many of which focus on government-established reserves created around priority sage-grouse habitats (Crist et al. 2015), such as the reserve in Wyoming (State of Wyoming 2011) evaluated herein. Additionally, in the United States, these state-designated reserves (i.e., core areas) have been largely carried over into federal-level efforts to conserve sage-grouse (USFWS 2013).

However, questions remain regarding the primary factors that affect the efficacy of the umbrella species concept in practice, in part because many investigations of the umbrella species concept have been based on hypothetical reserves (Roberge and Angelstam 2004). We used a large, real-world reserve established for sage-grouse to address 4 objectives: determine how much protection the reserve offered background species, describe the tradeoff between reserve size and the protection the reserve provided to background species, test whether the established umbrella reserve offered more protection to background species than a simulated reserve of equal size but sited with no regard for the umbrella species, and examine the relationship between the amount of taxonomic and ecological similarity between umbrella and background species and the effectiveness of the reserve.

STUDY AREA

Our study took place within the state of Wyoming, USA (~253,596 km²). Wyoming is a diverse landscape at the intersection of 3 broad ecological regions: North American deserts, Great Plains, and northwestern forested mountains (Commission for Environmental Cooperation [CEC] Level I; Wiken et al. 2011). Because elevation in Wyoming ranges from 939–4,207 m (Knight 1994), the climate varies dramatically along gradients of precipitation and temperature (PRISM Climate Group 2012, Chambers et al. 2016). Dominant vegetation types included sagebrush shrublands, desert shrublands, and mixed-grass prairie in the lower-elevation areas, and coniferous forests and alpine plant communities at higher elevations (Knight 1994). Primary land uses within the state included agriculture, mining, and livestock grazing; however, much of the state has remained relatively undisturbed by anthropogenic development (Knight 1994). Wyoming was an ideal area within which to assess the utility of sage-grouse as an umbrella species for several reasons. Wyoming was the first state to adopt a comprehensive, statewide strategy for sage-grouse conservation (USFWS 2015c), a model subsequently adopted by other state and federal agencies across the bird’s range (USFWS 2013, Crist et al. 2015). These conservation strategies have focused on delineating priority areas for conservation such as the semi-protected reserve we evaluated. Wyoming is home to 37% of the remaining sage-grouse population and supports nearly 25% of the bird’s habitat across its range (USFWS 2015c). Wyoming also has a diverse group of background species that could potentially benefit under the umbrella of sage-grouse conservation, with >50 animal species associated with vegetation communities where greater sage-grouse occur having been designated as species of concern (Wyoming Game and Fish Department [WGFD] 2010).

The Wyoming state government created a semi-protected sage-grouse reserve (i.e., Sage-Grouse Core Areas) to conserve sage-grouse and avert the need to list the species under the ESA (State of Wyoming 2008, 2010, 2011). The reserve was originally established in 2008 (State of Wyoming 2008), and we evaluated the reserve as revised in 2011 (State of Wyoming 2011). The reserve was 61,777 km² (24.4% of the state), comprised of 31 separate units (Fig. 1), and contained approximately 82% of the state’s known sage-grouse population (Gamo et al. 2013, Spence et al. 2017). The design of the reserve focused on high-abundance breeding areas of sage-grouse, with the assumption that the reserve would also encompass necessary non-breeding sage-grouse habitat (State of Wyoming 2011, Smith et al. 2016). Management actions implemented within the reserve were intended to maintain or increase sage-grouse habitat and populations, primarily through regulation of activities that threaten sage-grouse and the integrity of their habitat (e.g., energy development, habitat conversion) and promotion of beneficial activities (e.g., habitat reclamation, development of conservation agreements). Two key aspects of the policy defining reserve protections were stipulations that the accumulated surface disturbance within any proposed impact zone was not to exceed an average of 5% per 2.6 km² of the surface area, and that surface disturbance within 966 m of occupied sage-grouse leks was prohibited (State of Wyoming 2011). The state’s policy did not guarantee the existence of the reserve into perpetuity, nor did it completely arrest human activity and disturbance within the reserve. The primary qualification for nature reserves, however, is not that they are pristine and excluded from human activity. Instead, nature reserves are places where the conservation of biodiversity, ecological integrity, wilderness, or similar values takes precedence over other values and uses (Noss et al. 1999), a standard met by these semi-protected areas. Because all units of the reserve were managed under a policy that prioritized sage-grouse and sagebrush ecosystem conservation over other interests, we defined the umbrella reserve as all 31 units in aggregate and refer to this as the established reserve.

Figure 1.

The umbrella reserve created for greater sage-grouse (State of Wyoming 2011) and the range of big sagebrush (Little 1976) in Wyoming, USA, 2011.

METHODS

Background Species

We focused on 52 background species (or subspecies) of wildlife that were listed in the State Wildlife Action Plan (WGFD 2010) as Species of Greatest Conservation Need that were associated with vegetation communities where greater sage-grouse occur (Table S1, available online in Supporting Information). We used the overlap between the semi-protected sage-grouse reserve and a species’ habitat as a measure of the coverage that species received under the sage-grouse umbrella. To identify the habitat of background species, we used publicly available species distribution models (SDMs) that were a spatially explicit representation of habitat for each species within Wyoming at 30-m spatial resolution (Fig. S1). Defining habitat for each species using an SDM aligns with the definition of Hall et al. (1997), where habitat is organism-specific and relates the presence of a species to an area’s physical and biological characteristics. The SDMs we used (Keinath et al. 2010) were created separately for each background species by modeling presence-only data using maximum entropy methods shown to be robust under the given data structure (Elith et al. 2011) and using pseudo-absence data (sometimes called background data) selected from the sample set to avoid biases often associated with presence-only data (Phillips et al. 2009). Presence data came from observations made between 1960–2009, with most observations made after 1980 (Keinath et al. 2010). These SDMs are used within the state to conduct species conservation planning under the auspices of the State Wildlife Action Plan (WGFD 2010). Statistical methods and model evaluation are detailed in Keinath et al. (2010); to summarize, model performance was generally strong, with area under the receiver operating characteristic curve values of 0.80 ± 0.20 (SD) based on 10-fold cross validation using holdout test data (Hanley and McNeil 1982). Binary classification of SDM output was achieved by specifying a threshold that divided the continuous SDM output into areas where the species was predicted to be present (i.e., habitat; Hall et al. 1997) versus absent (not habitat; Hall et al. 1997). The binary-classification threshold was selected for each species to maximize the sum of training sensitivity (true positive rate) and training specificity (true negative rate), which is designed to maximize the discriminant ability of the binary output (Keinath et al. 2010). To compare the coverage of species of differing conservation priority, we used the priority ranking of each background species in the State Wildlife Action Plan (WGFD 2010), with tier-1 species being the highest priority group and tier-3 species being the lowest.

Data Analysis

Overlap between the reserve and background species.—

We calculated the proportion of each background species’ habitat within the state that was contained within the established reserve. We refer to this as the observed overlap statistic to differentiate it from overlap statistics calculated for each species under simulated and hypothetical reserve designs discussed hereafter. We calculated the observed overlap statistic in a geographic information system (GIS) for each species as a simple fraction, where the denominator was the number of cells classified as habitat in the species’ Wyoming SDM, and the numerator was the number of habitat cells that fell within the established reserve. We used a 1-way analysis of variance (Gotelli and Ellison 2004) to test whether observed overlap was equal across species-level tiers of conservation priority. We conducted data analyses in Program R (R Core Team 2016). We used the packages sp (Pebesma and Bivand 2005), rgeos (Bivand and Rundel 2015), rgdal (Bivand et al. 2015), and raster (Hijmans 2015) for spatial analyses, and the snow package (Tierney et al. 2013) for parallel computing. We used the packages RODBC (Ripley and Lapsley 2016), reshape2 (Wickham 2007), plyr (Wickham 2011), ggplot2 (Wickham 2009), and gridExtra (Auguie 2016) for data handling and plotting.

Simulation of randomly sited reserves.—

One criticism of the umbrella species concept has been the lack of proper controls or baselines for comparison when assessing the effectiveness of umbrella reserve strategies (Caro 2003). The appropriate control for comparison would be an equally sized reserve within the same region but not delineated with the requirements of any one umbrella species in mind (Caro 2003, Dunk et al. 2006). We therefore developed a Monte Carlo test based on a spatially explicit randomization scheme to generate an expectation (or more appropriately, a null distribution) against which to test each species’ observed overlap statistic. Our method therefore tested whether the established umbrella reserve protected more or less habitat of each background species than expected by chance, given the reserve’s size. The principal mechanism of our Monte Carlo test was to create many simulated reserves sited with no regard for sage-grouse but equal in size to the reserve established for sage-grouse. Monte Carlo tests are a non-parametric, randomization-based method for testing the significance of an observed test statistic (in this case, the observed overlap) by comparing it with a random sample of test statistics generated from a user-specified randomization scheme (Manly 1997). An advantage of our approach was the ability to condition the null model (as is common in other spatial anlyses; Fortin and Dale 2005) and frame questions and hypotheses in ecologically meaningful ways. For instance, because all background species were associated with sagebrush shrublands or grasslands where sagebrush was present, we conditioned the null model of overlap by allowing simulated reserves to fall only within the range of big sagebrush (Artemisia tridentata; Little 1976; Fig. 1). Thus, we excluded areas with little biological relevance to the species at hand (e.g., forested and alpine environments).

The established reserve was based, in large part, on sage-grouse breeding density data summarized within 6.4-km radii surrounding sage-grouse leks (Doherty et al. 2010, Gamo et al. 2013). We therefore created 80 simulated reserves by randomly siting circular polygons with 6.4-km radii, to mimic the spatial structure inherent in the established umbrella reserve. Because simulated polygons were allowed to overlap, we added and aggregated polygons iteratively until the simulated reserve was the same size as the established reserve. We re-calculated the overlap statistic for each species under each simulated reserve, creating a distribution of 80 overlap statistics for each of 52 background species (52 × 80 = 4,160). We computed the expected overlap by calculating the mean of the distribution of simulated overlap statistics for each species and a 2-tailed 95% Monte Carlo confidence interval for expected overlap by isolating the 95% quantiles of the distribution of overlap statistics for each species. We then subtracted the expected value and its confidence interval limits from the overlap statistic calculated using the established umbrella reserve to compute the difference between observed and expected overlap (with 95% Monte Carlo CI). If a species’ confidence interval for the computed difference contained zero, we concluded that the established umbrella reserve performed no different than expected due to chance. If the difference between observed and expected overlap statistics was positive (and the CI did not contain zero), we concluded that the established umbrella reserve performed better than expected for that background species. If the difference was negative (and the CI did not contain zero), the established umbrella reserve performed worse than expected for that background species. We provide annotated computer code as a script file (https://github.com/jcarlis3/umbrella/blob/master/UmbrellaMethodsDemo.R) that demonstrates our methods using example data and functions provided in the umbrella package (Carlisle 2017) and the free R software environment (R Core Team 2016). Specifically, the script details the following: installing required software packages, calculating the overlap statistic, simulating reserves, performing the Monte Carlo test, and interpreting the results.

Reserve size.—

To assess the importance of reserve size to the utility of an umbrella species conservation strategy, we compared the observed overlap statistics from the established reserve to those generated using 5 alternative reserves of various sizes. We used spatially explicit rankings of sage-grouse breeding abundance (Doherty et al. 2010) to delineate 5 plausible alternative reserves within Wyoming (Fig. S2). Rankings were based on count data from sage-grouse leks and enclosed high-abundance population centers of the species across its range (Doherty et al. 2010). The smallest alternative reserve included only the sage-grouse habitats with the highest breeding density; specifically, the smallest area that contained 25% of the known breeding population of sage-grouse across their entire range (Doherty et al. 2010), and was 13,060 km² (21% of the size of the established umbrella reserve; Fig. S2). The largest alternative reserve included the current distribution of sage-grouse (Schroeder et al. 2004; including additions of Doherty et al. 2010) and was 177,827 km² (2.88 times larger than the size of the established umbrella reserve; Fig. S2). We calculated 312 overlap statistics, one for each of 52 background species within each of 6 reserves (i.e., the established reserve and the 5 hypothetical, alternative reserves). We then classified each species as meeting 1 of 5 hypothetical conservation objectives for each of the 6 reserves: >0, ≥25, ≥50, ≥75, or 100% protection of habitat.

Species traits.—

To explore relationships between key species traits and the performance of the established reserve, we first combined background species into 2 performance groups based on the results of our Monte Carlo test: species for which the established reserve performed better than expected due to chance (and given the reserve’s size) and species for which the established reserve performed as expected or worse than expected. We then compared the values of 3 species-specific traits (taxonomic class, sagebrush association, and prevalence of habitat) between the 2 performance groups. We summarized taxonomic class within each group as the proportion of species that were of the same taxonomic class as the umbrella species (i.e., Aves). We calculated an index of sagebrush association as the proportion of each species’ habitat within Wyoming that was within the range of big sagebrush (Little 1976). We calculated prevalence of habitat as the area predicted to be habitat for each species within Wyoming. We used a non-parametric randomization test with 1,000 resampling iterations to test for differences between the 2 groups and generated 2-tailed P-values, using α = 0.05 as the significance threshold.

RESULTS

Larger reserves generally overlapped a higher proportion of the habitat of background species (Fig. 2A). The number of species for which a given conservation objective was met (e.g., ≥25% overlap), increased with reserve size for all conservation objectives (Fig. 2B). The mean amount of background species’ habitat protected by the established umbrella reserve was 20.9% but varied widely among the 52 background species (range = 0–63.2 ± 16.4%; Fig. 3A). Forty-eight of the 52 background species had habitat within the established reserve, 19 species had overlap statistics ≥25%, 1 species had an overlap statistic ≥50%, and no species had ≥75% of its habitat protected (Fig. 3A). Across all background species, the established umbrella reserve overlapped less habitat than expected by chance given the reserve’s size (mean difference in overlap statistic [i.e., observed overlap with the established reserve − overlap expected by chance] = −1.49 ± 10.5%, range = −19.2–46.9%; Fig. 3B). The established umbrella reserve performed better, no different, and worse than expected for 12, 12, and 28 species, respectively (Fig. 3B).

Figure 2.

The observed overlap statistic (proportion of a species’ habitat protected by a reserve) for 52 background species and 6 reserves (A). Reserves include the established umbrella reserve created for greater sage-grouse within Wyoming, USA, 2011 (dashed line), and 5 hypothetical, alternative reserves. We also present the number of background species for which hypothetical conservation objectives (e.g., × = at least 25% of habitat overlapped by the reserve) were met across the range of reserve sizes (B).

Figure 3.

The observed overlap statistic (the proportion of a species’ habitat protected) for 52 background species and the umbrella reserve created for greater sage-grouse within Wyoming, USA, 2011 (A). We also present the difference between the observed overlap statistic and a null model of expected overlap (observed − expected), given the size of the reserve, including 95% Monte Carlo confidence intervals (B). The umbrella reserve created for greater sage-grouse performed better (triangles), no different (hollow circles), and worse (squares) than expected for 12, 12, and 28 species, respectively.

The amount of overlap between the established reserve and the habitat of background species varied by the species’ rank of conservation priority (P = 0.04; Fig. S3). The species of highest conservation priority had a mean overlap of 34.9% (95% CI = 23.0–46.7%). The mean overlap was 19.6% (95% CI = 14.4–24.7%) for species of intermediate conservation priority, and 14.6% (95% CI = 3.5–25.7%) for those of lowest priority.

The background species for which the umbrella reserve overlapped more habitat of that species than expected by chance, given the reserve’s size, were more likely to be birds (compared to mammals, reptiles, and amphibians; effect size = 33.3%, P = 0.04; Fig. 4A), had higher association with sagebrush plant communities (effect size = 29.0%, P < 0.01; Fig. 4B), and had higher prevalence of habitat within the study area (effect size = 45,517.1 km², P < 0.01; Fig. 4C).

Figure 4.

Mean values of 3 species-specific traits, taxonomic class (A), association with sagebrush plant communities (B), and prevalence of habitat (C) in relation to the performance of an umbrella reserve created for greater sage-grouse within Wyoming, USA, 2011. Species in the better group (n = 12) had more habitat overlapped by the umbrella reserve than expected given the size of the reserve. Species in the worse or no different group (n = 40) had less or as much habitat overlapped by the umbrella reserve than expected given the size of the reserve. Error bars are bootstrap-generated 95% confidence intervals based on 1,000 resampling iterations.

DISCUSSION

In concordance with biogeographic theory (MacArthur and Wilson 1967, Soulé and Terborgh 1999), we found strong positive relationships between the size of the umbrella reserve designed for the greater sage-grouse and the number of background species protected and the amount of protection each species received. In addition, we demonstrated that given a particular reserve size, the relationship of umbrella species’ traits to those of the background species is important and can be used to predict which species may necessitate additional conservation action beyond the umbrella strategy.

The conservation of sage-grouse habitat has conferred substantial habitat protection to several other species of concern, including mule deer (Odocoileus hemionus; Copeland et al. 2014), shrubland-associated songbirds (Hanser and Knick 2011), and a suite of other shrubland-associated vertebrates (Rowland et al. 2006). Indeed, our findings also demonstrated considerable spatial overlap between protected sage-grouse habitat and the habitat of some sagebrush-associated wildlife species. We had a priori reason to suspect that the background species included in our analyses were plausible to consider under the sage-grouse umbrella because of their broad-scale associations with vegetation communities where sage-grouse occur in Wyoming (i.e., shrublands or grasslands where sagebrush was present). However, we caution that the amount of overlap varied widely among these background species. Conservation practitioners, therefore, should not assume that the umbrella species concept is a uniform cure for all conservation concerns in the sagebrush-steppe ecosystem. Rather, even among sagebrush-associated species, our results revealed that broad-scale protection of important sage-grouse areas has less utility in the conservation of species that are less similar to sage-grouse in taxonomy, habitat prevalence, and sagebrush association.

Reserve Size

We emphasize that reserve size was likely the key mechanism underlying the performance of the umbrella species reserve in protecting background species in our study system. The longstanding assumption has been that space or size is the critical feature of successful umbrella reserves (Caro 2003, Seddon and Leech 2008). Our findings explicitly demonstrate the relationship between reserve size and the conservation utility of an umbrella-species conservation strategy centered on sage-grouse. Implementing an umbrella-species conservation strategy involves both the selection of an umbrella species and the delineation of a reserve for that species (Caro 2010). Our assessment of 5 alternative reserve configurations shows that while holding the umbrella species constant, the characteristics (i.e., size) of the umbrella reserve designed on its behalf played a critical role in determining the efficacy of that reserve as a conservation umbrella for other taxa. Additionally, we found that simulated reserves (i.e., reserves of equal size to the established umbrella reserve and sited with no regard for sage-grouse), performed as well as, or better than, the established umbrella reserve for most background species (40 of 52 species; 76.9%). Therefore, a reserve as large as that designated for sage-grouse but delineated for another sagebrush-associated species, or even delineated at random within sagebrush steppe, would likely be as effective, or more effective, than the sage-grouse umbrella at protecting the 52 sagebrush-associated wildlife species we examined. These results provide further evidence that reserve size is a key consideration when using the umbrella species concept in practice, at least in our study system. However, we acknowledge that broad-scale protection schemes alternative to the sage-grouse protections we examined would likely not be feasible from a sociopolitical perspective, and that unique motivations and opportunities exist for focusing conservation efforts on sage-grouse (Knick and Connelly 2011). As such, using sage-grouse as an umbrella species has conservation utility, but managers should recognize where a sage-grouse-centered strategy might fall short in conserving background species, and how to expediently identify such holes in the sage-grouse umbrella.

Examining the observed overlap between established sage-grouse reserves and the habitat of background species offers a way to identify holes in the sage-grouse umbrella, but a criticism of such approaches (and the umbrella species concept in general) has been that umbrellas are not compared to a proper control or baseline (Caro 2003). Our Monte Carlo simulation comparing the sage-grouse reserve to equally sized, simulated reserves demonstrated that simulation can provide meaningful insights when real-world comparison reserves are non-existent or impractical to create. For example, the established umbrella reserve overlapped 18.7% of the habitat of the state-endemic Wyoming pocket gopher (Thomomys clusius), which was less than half of the 37.9% overlap we would expect based solely on the size of the umbrella reserve. Although covering nearly 20% of Wyoming pocket gopher habitat under the sage-grouse umbrella is certainly not trivial, it is arguably too small a proportion to be the sole conservation mechanism for a species faced with considerable threats. Indeed, Wyoming pocket gophers are relatively rare, geographically restricted, not strongly associated with sagebrush at finer scales (Keinath et al. 2014), and have a potentially high exposure to oil and gas development despite the sage-grouse reserve (Pocewicz et al. 2014). Wyoming pocket gophers are therefore an example of a species likely necessitating a more-targeted conservation strategy beyond protections offered by the sage-grouse umbrella. The umbrella strategy, however, is typically not implemented as a method to augment conservation plans for background species but rather as a shortcut to replace a multitude of species-specific conservation plans with one that encompasses the needs of multiple species under the umbrella of one (Wilcox 1984, Noss 1990). Our results emphasize that the small amounts of protection afforded some species by the established umbrella (e.g., 30 of 52 species had 20% protection or less; Table S1) may not be adequate as the only conservation measure for those species. The analyses we presented here provide a way to identify and potentially predict species that would be poorly covered under an umbrella species strategy, and clarify where additional resources could be directed to cover holes in the conservation umbrella.

Species’ Traits

The background species considered in this study were all associated with the same general vegetation communities as sage-grouse and therefore expected to benefit from the sage-grouse umbrella. Nonetheless, our results suggest that background species with traits that did not closely match the umbrella species (e.g., non-birds with a lower sagebrush association and lower prevalence of habitat, such as swift fox [Vulpes velox]) were relatively poorly covered by the established umbrella reserve (Fig. S4), and fared better under random reserve siting than the reserve established for sage-grouse (Fig. 4).

Despite the expectation that understanding which traits predispose background species to coverage under an umbrella strategy could be critical to the overall success of conservation based on umbrella species, approaches such as ours that establish the relationship between the traits of background species and their level of coverage under an umbrella-species conservation strategy have been rare (but see Roberge and Angelstam 2004). Indeed, most work on the umbrella species concept has centered on the traits of umbrella species that lead to effective conservation outcomes, with little regard for the traits of background species (Caro 2010). Our findings support previous work showing that taxonomy (Prendergast et al. 1993, Reid 1998, Caro 2003), habitat preference (Martikainen et al. 1998, Suter et al. 2002), and prevalence of habitat (Lawler et al. 2003), considered independently, can affect multi-species conservation strategies. We extend these findings to show that considering these traits in combination can further assist wildlife managers in prioritizing conservation efforts. For example, our results suggest that conservation practitioners should pay special attention to background species with limited habitat, particularly if their vegetation associations do not align closely with those of the umbrella species. Conversely, less concern may be warranted for species with more widespread habitat, regardless of how closely their vegetation associations match those of the umbrella species, because such species tend to exhibit a moderately high level of overlap with the umbrella reserve (Fig. S4C).

Species in our study that had low prevalence of habitat frequently had generally low overlap with the umbrella reserve, but these species also exhibited the widest range of overlap values (Fig. S4C). This effect is likely influenced in part by random process in combination with the size of the area of habitat, a manifestation of the so-called law of large numbers (Lindgren and McElrath 1966). For example, a species with a small area of habitat was likely to have that habitat either entirely missed or entirely overlapped by the umbrella reserve due to chance; however, the amount of overlap stabilized at a moderately high value as the prevalence of habitat increased (Fig. S4C). Regardless of the mechanism by which species with limited habitat were generally not well conserved by the umbrella, the pattern that their habitat was missed with more frequency suggests they should be given special attention when evaluating the effectiveness of umbrella conservation efforts.

Finally, sage-grouse reserves are traditionally delineated to encompass areas used by sage-grouse during the breeding season (Doherty et al. 2010), with the assumption that the bulk of non-breeding habitat is also protected (State of Wyoming 2011, Smith et al. 2016). Because habitats used by sage-grouse in other parts of the year (e.g., winter) often have different characteristics than breeding-season habitat (Connelly et al. 2000, Smith et al. 2016), efforts to protect other seasonal habitats of sage-grouse have the potential to broaden the conservation umbrella provided by sage-grouse, potentially reaching species dissimilar to sage-grouse that do not align closely with the breeding-season habitat of sage-grouse.

MANAGEMENT IMPLICATIONS

The assumption that sage-grouse management entirely replaces the need for tailored management of other sensitive species is unfounded. We stress that special consideration should be given to species of concern with scarce habitat, particularly if their taxonomy and vegetation associations do not align closely with those of sage-grouse. Our results should also be interpreted within the context of conservation objectives set by wildlife managers. For example, if the objective is that ≥25% of the habitat needed by at-risk species be within a conservation reserve, the sage-grouse reserve we evaluated met that objective for only 19 of 52 background species. If the desired objective was ≥50% habitat overlap, only 1 background species would qualify. Moreover, the boundaries and size of sage-grouse reserves are often revised over time (e.g., State of Wyoming 2008, 2010, 2011). Our work suggests that enlarging sage-grouse reserves, even by small amounts, would increase the number of species protected and the amount of protection for each. In contrast, reductions in reserve size would diminish the utility of sage-grouse as an umbrella species and likely exacerbate holes in the sage-grouse umbrella. Sage-grouse reserves traditionally focus on breeding-season habitat, so efforts to protect other sage-grouse seasonal habitats have the potential to broaden the conservation umbrella provided by sage-grouse. Finally, we caution that our results should be interpreted only at the broad spatial scale at which our methods were applied. For example, although we demonstrate that broad-scale protection of sage-grouse habitat may overlap with the habitat of some species, there is no guarantee that fine-scale management actions (e.g., vegetation treatments to improve sage-grouse habitat condition, or preferential siting of habitat disturbances in lower quality sage-grouse habitat) would benefit those same species. We suggest that studies of the fine-scale ecological requirements of species of concern relative to sage-grouse are warranted, as are investigations into the non-target effects of sage-grouse management actions other than reserve establishment.