A habitat‐centered framework for wildlife climate change vulnerability assessments: Application to Gunnison sage‐grouse

Abstract

The persistence of threatened wildlife species depends on successful conservation and restoration of habitats, but climate change and other stressors make these tasks increasingly challenging. Applying climate change vulnerability analyses to contemporary wildlife management can be difficult because most analyses predict direct effects of future climate on wildlife species at broad geographic scales, rather than assessing their habitats at local scales (<1 km) that correspond to site‐specific habitat management actions. We present a framework that synthesizes vegetation‐focused vulnerability assessments to assess multiple effects on wildlife species' diverse habitat needs, providing a scenario‐driven climate vulnerability assessment that maps differences in vulnerability of populations within a species' range. Our flexible habitat‐centered synthesis approach leverages available spatial datasets describing projected exposure to vegetation changes due to climate change and other potentially synergistic stressors, reclassifies and weights them based on available estimates of species' sensitivity to these changes, and recombines them to create <1‐km resolution maps of overall species vulnerability and threat‐specific habitat vulnerability. To demonstrate its potential to guide decision‐making, we applied this approach to the Gunnison sage‐grouse (Centrocercus minimus), a federally threatened habitat specialist that depends on sagebrush and mesic habitats that are imperiled by climate change. We mapped six threats forecasted out to the year 2070: direct effects of climate on (1) sagebrush cover loss and (2) mesic habitat drying, indirect changes in invasion risk from (3) pinyon–juniper conifers and (4) annual grasses, and potentially synergistic risk of (5) development and (6) wildfire. We then assessed species vulnerability for each of the eight extant populations under three climate scenarios: Optimistic, Continuation, and Pessimistic. We found that the extent of cumulative species vulnerability due to multiple habitat changes was far greater than the extent of any single habitat vulnerability. Over 75% of critical habitats were at risk under the Pessimistic scenario, and nearly two thirds of habitats were at high risk for three or more threats. Invasive species were the most widespread threat, highlighting the importance of indirect effects of climate change. We illustrate how our approach can be applied to the existing management planning strategies to better prioritize conservation of habitats for the persistence of threatened species.

INTRODUCTION

The persistence of wildlife species depends on successful conservation and restoration of their habitats (Sih et al., 2000; Taylor et al., 2005), but climate change and other landscape stressors make these tasks increasingly challenging (Millikin et al., 2020). Climate change is projected to make half of currently occupied habitats for North American bird species unsuitable by the end of the century, but current approaches to prioritizing management of these habitats generally fail to account for these future changes (Langham et al., 2015). Conservation plans that ignore the potential for future habitat change may unknowingly invest in areas that become unsuitable, possibly reducing or even nullifying returns on conservation investments. Understanding how these processes affect habitats across a species' range, as well as within local populations, is therefore key for forward-looking conservation planning—particularly for habitat specialists and range-restricted species that may experience more severe consequences from habitat loss (Foden et al., 2019; Kinzig & Harte, 2000; Malcolm et al., 2006; Pacifici et al., 2015). There is a need to develop integrated projections of vulnerability and identify areas where conservation investments are most likely to be effective in the long-term (Dunford et al., 2015; Langham et al., 2015; Millikin et al., 2020).

There are several traditional approaches to conducting climate change vulnerability assessments (CCVAs), which estimate the ability of species to withstand or adapt to climate change. Vulnerability is usually defined as the result of species' exposure to climate change, sensitivity to that change, and degree of adaptive capacity (although the latter is often excluded in wildlife CCVAs; Glick et al., 2011; Ofori et al., 2017). Index- or trait-based approaches assess relative vulnerability of different species by pairing their trait rankings that indicate sensitivity with exposure measures, niche approaches project distribution shift based on correlations with current climate associations, and mechanistic approaches employ more complex process-driven models (e.g., based on physiology and life histories; Glick et al., 2011; Pacifici et al., 2015).

While useful, the broad spatial scale of traditional CCVA approaches can weaken their utility for planning management of wildlife habitats in the face of climate change. Most approaches focus on geographic-scale range shifts or identifying which species are most vulnerable to climate change (LEDee et al., 2021; Miller et al., 2022), but most management actions occur at the “site-scale” (<1 km) where changes in habitats and biotic interactions are expected to be the dominant climate effects (Pacifici et al., 2015; Rowland et al., 2011). Recommendations based on geographic-scale CCVA approaches may omit important spatial variation in effects at this scale (Mantyka-Pringle et al., 2014; Reece et al., 2018). While there is significant uncertainty in predicting future habitat changes, there are also risks in planning without accounting for potential future changes, and natural resource agencies are mandated to use best-available science (Langham et al., 2015). CCVA designed with the goal of helping managers create local, evidence-based climate adaptation solutions for threatened wildlife species therefore remain a major gap in the literature (LEDee et al., 2021; Miller et al., 2022).

To be effective, wildlife CCVA approaches need to forecast the complex effects of climate change across the full suite of a species' habitat needs (e.g., for different life stages; LEDee et al., 2021; Michalak et al., 2022). CCVAs for vegetation have been conducted, but they generally model direct effects of climate on only a single vegetation group (Michalak et al., 2022; Small-Lorenz et al., 2013). In contrast, estimating vulnerability at the local scale of populations requires assessing climate effects on the full set of vegetation groups a species requires to persist, but this remains a key omission from many CCVAs (Dawson et al., 2011; Pacifici et al., 2015; Rowland et al., 2011). These effects include not only “direct effects” of climate change on primary habitat needs but also changing biotic interactions that may degrade habitats (frequently called “indirect effects”), such as the spread of invasive species with competitive advantages in novel climates. There is also a need to include other drivers of habitat loss in CCVAs (e.g., land-use change), which are often referred to as “synergistic” threats because of potential nonlinear cumulative effects with climate change (Mantyka-Pringle et al., 2012; Pacifici et al., 2015). Few past CCVAs have included direct effects, indirect effects, and land-use change, in part due to limited availability of spatial data (Dawson et al., 2011; Foden et al., 2019; Mantyka-Pringle et al., 2014; Rowland et al., 2011); but see Triviño et al. (2013, 2018). These limitations are especially problematic because degradation of habitats is likely the most important climate change impact for most wildlife species, and managing critical habitats is the primary focus of many regulatory agencies (Araújo & Luoto, 2007; Cahill et al., 2013; Foden et al., 2019; Rowland et al., 2011).

We developed a framework for integrating outputs from diverse vegetation forecasts to create a CCVA that maps site-scale differences in vulnerability for wildlife populations due to changes in multiple habitat needs. Our approach falls under the umbrella of “synthesis assessments,” which leverage multiple exposure datasets and merge them with existing knowledge about species. Synthesis assessments are particularly suitable for regional planning because they can more comprehensively assess multiple effects across the range of key natural resources, using multiple lines of evidence to identify areas of consensus for habitat shifts (Glick et al., 2011; Michalak et al., 2022; Pacifici et al., 2015). We leveraged published spatially explicit projections of vegetation changes or risk of loss, and weighted them based on species–habitat relationships to represent a wildlife species' differing sensitivities to these changes. The outputs provide multifaceted data on where different types of habitat degradation are likely to occur, creating a spatial atlas that can be used by managers to prioritize monitoring of actual changes, take early action, and invest in areas more likely to be stable under climate change.

We apply this framework to create a CCVA for the Gunnison sage-grouse (Centrocercus minimus), a threatened and rapidly declining species in urgent need of management interventions (U.S. Fish and Wildlife Service [USFWS], 2019). This species is a range-restricted habitat specialist reliant on sagebrush and mesic habitats that are imperiled by climate change (USFWS, 2019). The only current vulnerability assessment for Gunnison sage-grouse is the USFWS’ Species Status Assessment (SSA; USFWS, 2019), which used an aspatial approach based on expert opinion. We built on this effort by creating a spatially explicit, quantitative assessment based on forecasts of six detrimental types of habitat change under three divergent scenarios of climate change and development. We mapped overall vulnerability (defined as average vulnerability across all six changes) of the species' eight extant populations and specific vulnerability of habitats to each of the six types of changes. We conclude by illustrating how this approach can be used in management planning frameworks to meet Bureau of Land Management (BLM) goals for habitat restoration to support population recovery (USFWS, 2020).

CASE STUDY

We assessed the full range of the Gunnison sage-grouse, a species that is range-restricted to southwestern Colorado and southeastern Utah, federally listed as threatened, and at high risk of extinction by 2050 (Braun et al., 2014; Davis, 2012; Schroeder et al., 2004; USFWS, 2019). Sage-grouse are sagebrush-obligates, depending on it for both cover and food during all life stages (Connelly et al., 2000). During brood-rearing periods, hens and chicks rely on higher elevation mesic areas, such as wet meadows, for food (Young et al., 2015). Habitat degradation has already extirpated the species from roughly 90% of its range (Schroeder et al., 2004). Both habitat conservation and restoration will be necessary to maintain and recover the populations (USFWS, 2020). Even with aggressive habitat management actions, there is substantial risk of further habitat loss due to ongoing development and climate change shifting the climatic niches of vegetation within the sagebrush ecosystem (Millikin et al., 2020). These factors are likely to influence the long-term viability of Gunnison sage-grouse populations and jeopardize the intended benefits of habitat management efforts (USFWS, 2019, 2020).

METHODS

Habitat-centered synthesis vulnerability framework

We employed a spatially explicit, habitat-centered synthesis framework for wildlife CCVA that used only simple raster reclassification in a four-step process (Figure 1). Major steps to our approach were identifying climate-driven detrimental vegetation cover changes or risk of critical habitat loss (hereafter “threats”; Step 1), collecting available forecasts of exposure to these threats (Step 2), and aligning and grouping the forecasts into divergent scenarios (Step 3). Species–habitat models are then used to represent sensitivity to these changes, producing spatial recommendations for management of wildlife habitats for climate adaptation (Step 4). In the following sections, we first provide an overview of the conceptual approach of each step, and then illustrate its application for the Gunnison sage-grouse.

[IMAGE OMITTED. SEE PDF]

Step 1: Defining management targets and climate scenarios

Conceptual approach

First, we identified (1) the extent of assessed range and any relevant management units (e.g., population boundaries), (2) focal threats to habitats for the species, and (3) multiple divergent “narrative scenarios” that describe future changes (e.g., “hot and dry scenario: temperature increases exceed 5°F and precipitation declines”), which were used in Step 3 to group projections according to similar climate pathways and assess uncertainty over a range of possible outcomes (Michalak et al., 2022). Land managers participated in this step to ensure the CCVA can inform their management (Foden et al., 2019; Miller et al., 2022; Rowland et al., 2011).

Application to sage-grouse

We assessed a 67,370 km² bounding box encompassing the extant range of Gunnison sage-grouse (Figure 2; Saher et al., 2022). We masked our study area to only current (2016) rangeland and agricultural land (Appendix S1: Section S1.1). There are eight extant Gunnison sage-grouse populations (west to east): Dove Creek/Monticello (hereafter referred to as a single population, “Dove Creek”), San Miguel, Piñon Mesa, Cimmaron/Cerro/Sims Mesa (CCS), Crawford, Gunnison, and Poncha Pass (Figure 2). Gunnison has the largest population—containing almost 85% of individuals—and the other “satellite” populations are perilously small, generally declining, and in urgent need of habitat improvements (USFWS, 2019). The populations have substantial differences in environmental characteristics, land ownership, and genetic adaptation, highlighting the need for population-specific management (USFWS, 2019; Zimmerman et al., 2022). USFWS has designated 7775 km² of critical habitat for Gunnison sage-grouse (Figure 2, black outlines; USFWS, 2014). Recent studies of habitat use (Aldridge et al., 2012; Saher et al., 2022) have also identified high-quality crucial habitats (Figure 2, red shading) that represent estimated currently utilized breeding habitats (described in Appendix S1: Section S1.1). Delineated habitat polygons sometimes included nonhabitat land-covers (e.g., forest), so we report only the proportions of actual habitats within our study area mask (rangelands and agricultural land).

[IMAGE OMITTED. SEE PDF]

We worked with BLM habitat managers and reviewed the USFWS' SSA (USFWS, 2019) to identify key threats to the persistence of Gunnison sage-grouse. Habitat degradation has been primarily due to not only loss of sagebrush from human development and disturbance (Braun et al., 2014; Oyler-McCance et al., 2001) but also encroachment by native pinyon–juniper conifers that provide predator perches and invasive annual grasses (primarily cheatgrass, Bromus tectorum) that alter fire cycles (Aldridge et al., 2012; Apa et al., 2021; Saher et al., 2022). Many mesic habitats are already disappearing due to drought (Maestas et al., 2018), which could be exacerbated by climate change, resulting in the reduction of a key specialized food resource needed for brood rearing (USFWS, 2019). We therefore assessed vulnerability of habitats to four threats under climate change: (1) declining sagebrush cover, (2) drying of mesic resources, (3) increases in pinyon–juniper encroachment, and (4) invasion by annual grasses. We also assessed two synergistic threats that could cause habitat loss: (5) development and (6) wildfire.

We defined three future narrative scenarios: Optimistic, Continuation, and Pessimistic (described more in Step 3). These were based on the SSA (USFWS, 2019) to ensure our assessment was comparable with ongoing management planning.

Step 2: Literature review to identify relevant spatial datasets

Conceptual approach

We reviewed the literature and data repositories to find the best-available spatial datasets for each focal habitat or threat, of both (1) current habitat status and (2) projected future changes (ideally with multiple scenarios). Diverse data types were included, including projected percent vegetation cover, probability of change, and presence/absence for different threats. Transformations were used to increase applicability to the species (e.g., reclassifying percent cover based on minimum requirements for the species). Current habitat conditions were used to mask out areas (to not applicable [NA]) if they are irrelevant for a given threat (e.g., risk of habitat loss where that habitat is already absent).

Application to sage-grouse

We found spatial datasets representing the current and possible future status of all six threats that ranged in resolution from 30- to 800-m resolution (Table 1). We used 2016 as our “current” year to align with this study. We used 2070s as the approximate future projection decade, although some datasets differed in their exact start and end year (Appendix S1).

TABLE 1 Datasets used to construct three scenarios of habitat vulnerability for Gunnison sage-grouse (Centrocercus minimus), along with thresholds for each used to consider habitats as currently partially or fully degraded or at risk of future degradation from each of six threats.

Threat	Current landscape (ca. 2016)	Future landscape (ca. 2070)
	Dataset(s)	Fully degraded	Partially degraded	Dataset(s)	At risk
Sagebrush	2016 National Land Cover Dataset RFC (Rigge et al., 2020)	0% cover	1%–5% cover	Projected change in RFC (Rigge et al., 2021)	≤5% cover
Mesic resources	Regional map of mesic resources (Saher et al., 2022)	Absent	Not modeled	Change in aridity index (Rangwala & Dewes, 2018)	≥15% drier aridity index
Pinyon–juniper	Mapped tree canopy cover (Falkowski et al., 2017)	>4% cover	1%–4% cover	Pinyon–juniper bioclimate envelope models (CNHP, 2018)	≥1% cover
Invasive annual grasses	2016 National Land Cover Dataset RFC (Rigge et al., 2020)	>10% cover	1%–10% cover	Projected change in RFC (Rigge et al., 2021)	≥1% cover
Development	2016 National Land Cover Dataset (Jin et al., 2019)	>10% of landscape in a 1.5-km buffer	2.5%–10% of landscape in a 1.5-km buffer	FORE-SCE land-use change projections (Sohl et al., 2018)	≥2.5% of landscape in a 1.5-km buffer
Wildfire	Historic burn perimeter databases (GeoMAC, 2019; Picotte et al., 2020)	Burned 1984–2016	Not modeled	Annual burn probability maps (CSFS, 2017; Short et al., 2020)	≥33% chance of fire by 2070

Table 1 lists source datasets and Appendix S1 describes reprocessing steps used to make each dataset relevant to Gunnison sage-grouse habitat use and vulnerability. In overview, shifts in vegetation classes—sagebrush, annual grasses, and pinyon–juniper (Pinus edulis, Juniperus osteosperma, and J. monosperma)—were mapped based on projected 2035–2070 climate conditions, as described in Table 2. We estimated drying of mesic habitats (mapped by Saher et al., 2022) by overlaying percent change in aridity index (the ratio of potential evaporation to precipitation). Aridity index projections regionally downscaled to 800 m were obtained from Rangwala and Dewes (2018). We obtained maps of projected new development (domestic/industrial, cropland, and hay/pasture) from 2005 to 2070 under three climate scenarios from the FORE-SCE land-use/land-cover change model (Sohl et al., 2018). This model projects changes based on demand from future socioeconomic projections as well as biophysical factors including suitability under future climates (Sohl et al., 2014). Development not only directly causes habitat loss but disturbance from it also strongly negatively effects habitat use within 1.5 km (Aldridge et al., 2012; Saher et al., 2022). We therefore summarized current and new development (from 2005 to 2070) as the proportion of the landscape within a 1.5-km buffer. Annual burn probability maps were from two datasets, the Colorado State Forest Service's (2017) 30-m map for Colorado and Short et al.'s (2020) 270-m map for Utah, which were mosaicked together (Appendix S1: Table S1).

TABLE 2 Characteristics of three narrative scenarios' climate forcings and land-use change projections used as forcings in other different models (superscripts), which subsequently generated our input datasets of future landscape changes; climate change values were taken for the four-corners region of the southwest from Rangwala and Dewes (2018), while land-use change was calculated for the Gunnison sage-grouse (Centrocercus minimus) study area as the percent change from historic ([2070–2005]/2005) from the projections of Sohl et al. (2018).

Narrative scenario	Climate change forcings	Development forcing
Climate model	RCP	Δ°C annual temp.	%Δ in annual precip.	IPCC pathway	%Δ (km2) in hay–pasture	%Δ (km2) in cropland	%Δ (km2) in domestic–industrial
Pessimistic (hot and dry, high dev.)	HADGEM2-ESm,p	8.5	3.7	−8.2%	A2	−6.4% (−125.4)	+39.0% (+474.0)	+74.3% (+284.8)
AdaptWestr	8.5	3.3	−0.5%
Continuation (moderately hot, moderate dev.)	CESM1-BGCp	8.5	2.6	+4.6%	A1B	−1.1% (−21.3)	+51.9% (+631.7)	+45.5% (+174.6)
CCSM4m	8.5	2.4	+2.2%
AdaptWestr	4.5	2.4	−0.2%
Optimistic (warm and wet, low dev.)	CNRM-CM5m,p	4.5	2.0	+5.2%	B1	−7.4% (−145.5)	+22.0% (+267.9)	31.1% (+119.3)
AdaptWestr	4.5	2.4	−0.2%

To avoid “zero-inflating” future risk estimates with many areas that were irrelevant to management planning for that threat, we used current habitats to mask each threat layer to only where it was relevant based on their current landscape. Since we defined our study area as rangeland or agricultural land, all other areas were “NA” in all six threat layers. Our development and wildfire layers did not have additional areas masked out. For beneficial habitats, we set nonhabitats to NA: non-sagebrush areas for projected loss of sagebrush cover, and non-mesic areas for risk of mesic drying. For invasive annual grasses, we further masked already-invaded areas to NA because management is focused on preventing spread rather than eradicating current invasions (Miller et al., 2011); we did not do so for pinyon–juniper conifers because management of already-invaded areas is feasible for them. Agricultural land was set to NA for both, because crops and hay are unlikely to be replaced by encroaching vegetation.

Maps of the resulting forecasts are in Appendix S2: Figures S1–S14.

Step 3: Alignment of climate scenarios with spatial datasets

Conceptual approach

Individual threat forecasts were then aligned into the narrative scenarios defined in Step 1. Forecasts are generally the result of an emissions scenario (e.g., Representative Concentration Pathway [RCP] 8.5) run through a Global Climate Model (GCM; e.g., HADGEM2-ES) to simulate a future climate. Ideally, future climates can be aligned precisely by matching the same emissions scenario–GCM combination. Because our approach synthesizes preexisting datasets, an exact emissions–GCM match was not always possible, in which case forecasts were assigned to the narrative scenario with the closest matching change in temperature and precipitation. Narrative scenarios therefore represent “fuzzy” climate futures defined by qualitative similarity (e.g., “hot and dry” vs. “warm and wet”), which are used to explore a range of possible outcomes rather than predict a single precise future climate (Miller et al., 2022).

Application to sage-grouse

The SSA used single GCM models that were representative of three divergent climate forecasts (USFWS, 2019). The Optimistic scenario had a “warm and wet” (model CNR-CM5 RCP 4.5) climate, categorized by mild warming and increasing precipitation. The Continuation climate scenario was “moderately hot” (CESM1-BGC RCP 8.5) and had little change in precipitation and a moderate degree of warming. The Pessimistic climate scenario was “hot and dry” (HADGEM2-ES RCP 8.5), with strong increases in mean annual temperature and declining precipitation. We assigned each threat forecast to the narrative scenario whose projected change in temperature and precipitation was closest to its own (Table 2).

Our development projections (Sohl et al., 2014, 2018) were based on development pathways (A2, A1B, and B1; Nakicenovic et al., 2000) that were predecessors to the RCPs, but which aligned fairly well with them by 2070 (Glick et al., 2011). These were matched accordingly, with high, moderate, and low development corresponding to the Pessimistic, Continuation, and Optimistic scenarios (Table 2; Appendix S1: Section S1.1.2).

Wildfire risk maps were based on the current landscape and described annual probability of burning, which we reclassified into cumulative probability of burning by 2070 (Colorado State Forest Service, 2017; Short et al., 2020). They were the same in all scenarios because future wildfire projections were not available.

Step 4: Integrate sensitivity information to create vulnerability maps

Conceptual approach

Two complementary sets of products were produced: (1) scenario-specific maps of species vulnerability based on cumulative threats to multiple habitats for use in population prioritization, and (2) threat-specific maps of habitat vulnerability to inform habitat management and monitoring for each individual threat to habitats.

For the former, an overall species vulnerability map was created for each scenario by taking a weighted mean across the different forecasted threats. Rather than represent sensitivity as a product of expert-ranked species traits, we modeled sensitivity by deriving these weights from the coefficients of previously published habitat selection models. We reviewed published models of species–habitat relationships, and for each model, identified whether each covariate was related to one of the focal threats (i.e., a detrimental or beneficial land-cover). If a covariate (1) is in the final model, (2) has its sign in the expected direction (e.g., a negative relationship to a detrimental cover), and (3) is statistically significant, it receives a vote of 1. If it meets only the first two criteria, it receives a vote of 0.5. Otherwise, it receives a vote of 0. The mean of all votes across models is the estimated relative sensitivity of the species to changes in that habitat characteristic and was then used to weight calculation of the mean species vulnerability. Where current conditions meant a given threat is not relevant (i.e., masked out as described above), the NA values exclude these from the weighted mean. The resulting scenario-specific maps identify hotspots of species vulnerability to future changes that integrate multiple habitat needs and threats (equivalent to an ensemble model, which increases confidence; Gardali et al., 2012; Michalak et al., 2022; Triviño et al., 2018) to identify the most imperiled populations in need of management and/or triage.

Second, each threat dataset was reclassified into categorical habitat vulnerability maps designed to be applied in planning frameworks that spatially prioritize habitat management actions (e.g., the Resist–Accept–Direct framework; Fisichelli et al., 2016; Schuurman et al., 2020). Thresholds from the published literature were used to bin current and future status into five hierarchical categories (fully degraded, partially degraded, always at risk, sometimes at risk, and stable). “Fully degraded” and “partially degraded” describe the current state (e.g., whether a vegetation class meets current minimum or recommended percent cover for a species). Habitats that are currently intact habitats but projected to become degraded in all future scenarios are “always at risk,” or “sometimes at risk” if degraded in only some future scenarios; this consensus mapping expresses the degree of uncertainty in future projections (Glick et al., 2011; Miller et al., 2022). Habitats that were neither degraded nor at risk were “stable.”

Application to sage-grouse

We estimated overall species vulnerability as the weighted mean of the six future forecasted vulnerability threat maps in each scenario. The mean excluded NAs and was weighted based on each threat's relevance to Gunnison sage-grouse based on Saher et al.'s (2022) habitat use models (Appendix S1: Tables S2 and S3).

To create the threat-specific habitat vulnerability maps, we first applied numeric thresholds to the current landscape datasets for considering habitats as “partially degraded” or “fully degraded” (Table 1) based on Gunnison sage-grouse habitat use studies (described in Appendix S1: Section S1.1; Aldridge et al., 2012; Apa et al., 2021; Saher et al., 2022). Layers for mesic resources and historically burned areas were binary (presence/absence) and did not have a partially degraded category. We defined remaining areas as “at risk” for each threat if they were projected to be in a partially or fully degraded state by 2070 (Table 1). For habitat vulnerability maps only, some masked NA values were overridden with these new classifications (e.g., non-sagebrush rangeland areas were set to “fully degraded” rather than NA for the sagebrush map).

Summary statistics

We summarized relative vulnerability for each threat by calculating the proportion of critical habitat in each population that was degraded or at risk, as defined in Table 1. To summarize cumulative vulnerability, we also calculated the proportion of habitats at risk to multiple threats (0–6) in each scenario, and the “maximum” vulnerability as the proportion at risk under any of the three scenarios. Appendix S1 contains complementary figures summarizing the total area of critical habitats at risk (Appendix S1: Figure S1), as well as the proportion and total area of crucial habitats (Appendix S1: Figures S2–S4) known to be occupied in the breeding season (Aldridge et al., 2012; Saher et al., 2022).

RESULTS

Overall species vulnerability to cumulative habitat threats

Gunnison sage-grouse vulnerability to future changes in habitats varied substantially across populations (Figure 3). The satellite populations are both currently more degraded than the species “stronghold” in the Gunnison basin and more vulnerable to future threats (Figure 4). Over half of critical habitats were at risk for at least one threat, in all populations and scenarios (Figure 4).

[IMAGE OMITTED. SEE PDF]

[IMAGE OMITTED. SEE PDF]

The western half of the study area had greater vulnerability. The CCS and San Miguel populations were the most imperiled, with ≥3 high-risk threats for 64.8% and 62.7% of critical habitats (Figure 4, maximum). Over half of habitat in Piñon Mesa and Crawford was projected to be imperiled by ≥3 threats, and almost all their habitat (≥98%) was at risk for at least one threat (Figure 4). Gunnison was the least imperiled, with only 16.0% of habitat at risk for ≥3 threats across scenarios (Figure 4). Poncha Pass and Dove Creek were also less threatened, with only 21.1% and 28.8% of habitat imperiled by ≥3 threats (Figure 4).

Cumulative threats increased with scenario severity, although variability among scenarios was less than variability across populations (Figure 4). Average vulnerability differed by ≤0.05 in most populations (Appendix S1: Figure S5). The proportion of habitat at risk for ≥3 threats increased by 5.9 ± 5.5% from the Optimistic to the Continuation scenario, and by 9.7 ± 5.5% from the Continuation to the Pessimistic scenario (mean ± SD across populations; Figure 4). The proportion of fully stable habitat decreased by −10.6 ± 7.6% from the Optimistic to the Pessimistic scenario, and by −4.4 ± 7.9% from the Continuation to the Pessimistic scenario (Figure 4). A notable exception was Gunnison, whose average vulnerability increased fourfold from the Optimistic to the Pessimistic scenario (Appendix S1: Figure S5). Most of Gunnison was low risk in the Optimistic scenario (Figure 3a), higher elevations became at risk in the Continuation scenario as pinyon–juniper expanded and mesic habitats dried (Figure 3b), and most of the basin was at risk in the Pessimistic scenario (Figure 3c). In the Optimistic scenario, 48.3% of Gunnison's habitat was fully stable (the largest extent observed), but this declined to only 14.3% in the Pessimistic scenario.

Patterns of vulnerability across populations and scenarios were broadly consistent across alternative weighting schemes we considered (Appendix S1: Figure S5), indicating that estimated overall species vulnerability was robust to this choice.

Direct loss of habitats under climate change

Direct loss of sagebrush cover due to shifting climatic suitability was not a major threat to most populations (Figure 5a and 6a). Region-wide sagebrush percent cover was projected to decrease by −1.8 ± 1.4% (mean ± SD) in the RCP 4.5 climate, and −2.1 ± 1.5% in the RCP 8.5 climate (a decrease of nearly one third of current average cover). Declining sagebrush cover mostly imperiled the viability of habitats in the more arid western populations—Dove Creek, San Miguel, Piñon Mesa, and CCS—which already had higher proportions of degraded (i.e., trace or absent) sagebrush, which covered 29.0%–57.0% of critical habitats there (Figure 6a). San Miguel was a notable vulnerability hotspot, with over half of the remaining intact sagebrush at risk. Because the Gunnison population encompassed 57% of all critical habitat's intact sagebrush, a significant acreage of sagebrush was projected to be lost there, despite affecting a small proportion of its total area (Figures 5a and 6a).

[IMAGE OMITTED. SEE PDF]

[IMAGE OMITTED. SEE PDF]

Mesic resources represented a small proportion of overall study area (largely within the high-elevation fringes of critical habitats; Figure 5b) but serve as key habitats for brood rearing and were all projected to be at high risk of drying (change in aridity index ≥15%) in some future climates (Figure 6b). Differences in risk across scenarios dwarfed differences among populations: no areas in the warm–wet climate experienced drying past the 15% threshold, while 100% of mesic habitats were at risk in the hot–dry climate (Appendix S1: Figure S6b). Mesic resources' average percent change in aridity index was 4.1 ± 2.7% (mean ± SD) in the warm–wet climate, 15.3 ± 3.0% in the moderately hot climate, and 25.1 ± 4.0% in the hot–dry climate, representing “insignificant,” “moderate,” and “extreme” increases in aridity, respectively (Lickley and Solomon, 2018).

Indirect loss of habitats to climate-mediated invasion

Climate change has the potential to dramatically expand the climatic niche of pinyon and juniper species into sagebrush and mesic habitats (Figure 5c). Most critical habitats were threatened by pinyon–juniper encroachment under at least one scenario in every population (Figure 6c). Current encroachment was a comparatively small proportion of critical habitats (mean proportion = 14.5%; Figure 6c). Under all three scenarios, over 75% of rangeland critical habitats in Dove Creek, CCS, and Piñon Mesa were suitable for pinyon–juniper (Appendix S1: Figure S6c), with higher elevation areas at greater risk (Figure 5c). Poncha Pass' lower risk (roughly half of their habitats) was also consistent across scenarios. Piñon Mesa had 80.0% of rangeland critical habitats at risk in a warm–wet climate, but only half in the moderately hot and hot–dry climates (Appendix S1: Figure S6c). In contrast, Gunnison experienced a dramatic increase in risk from 5.7% of rangeland critical habitats at risk in a warm–wet climate to 46.1% in a hot–dry climate (Appendix S1: Figure S6c). Despite having a smaller proportion of habitats at risk, Gunnison represented nearly half of the at-risk acreage because of its size (Figure 6c). Furthermore, encroachment risk was scattered throughout central Gunnison areas, potentially fragmenting habitat (Figure 5c).

Current annual grass infestations varied among populations, with Gunnison and Poncha pass relatively uninvaded (Figure 5d), but invasions could spread to most critical habitats by 2070 (31.5%–79.6% at risk across populations; Figure 6d). Contemporary infestations were relatively established in the western populations and concentrated in lower elevations near agriculture (Figure 5d). Piñon Mesa had the greatest proportion of fully invaded areas and the greatest extent at risk for future invasion (Figure 6d). Relatively uninvaded basins were also at risk, particularly in the Pessimistic hot–dry climate where risk of invasion increased at higher elevations even as it decreased at lower elevations (Figures 5d and 6). In Gunnison, the Pessimistic scenario resulted in 18.2% more of its rangeland critical habitat projected to be invaded (Appendix S1: Figure S6d), resulting in 57.0% of its total critical habitat being at risk (Figure 6d). Poncha Pass was also largely uninvaded, but 30.7% of its habitat was at risk. Differences between the scenarios were minor in all other populations (mean absolute change = 1.6% of rangeland habitat; Appendix S1: Figure S6d).

The combined threat from both conifer and grass invasion was greater than either threat individually because high-risk areas were complementary, with lower elevation areas imperiled by invasive annual grass and higher elevation areas imperiled by pinyon–juniper conifers (Figure 5). Across populations, a mean 74.8% of critical habitats were at risk from one or both types of invasion in at least one scenario.

Potentially synergistic loss of habitats loss to disturbance

The highest development disturbance risk (proportion new development within a 1.5-km buffer) was in San Miguel, followed by CCS and Crawford (Figures 5e and 6e). Most of Dove Creek was already in a degraded state due to extensive existing development (Figures 5e and 6e), although one third to two thirds (depending on the scenario) of undeveloped areas were still projected to be in the proximity of additional development (Appendix S1: Figure S6e). At-risk populations tended to have high current disturbance by development (Appendix S1: Table S4), which was concentrated in Dove Creek, CCS, and Poncha Pass, and to a lesser degree in Crawford (Figure 6e). Projected development in the Gunnison basin covered a sizable area relative to the other subpopulations (Figure 5e; Appendix S1: Figure S1e) but a low proportion of habitats within this basin (Figure 6e), and was limited mostly to the southeastern highlands and current development in the basin's center (Figure 5e). Surprisingly, vulnerability was greatest in the Continuation scenario of moderate future development (A1B pathway) owing to widespread agricultural intensification, as compared with the Pessimistic scenario (A2 pathway) where domestic/industrial development was more prominent but further from critical habitats (Appendix S1: Table S5).

Wildfire had the smallest acreage at risk of any threat (Figure 6f), although many areas in the western Colorado Plateau ecoregion had high long-term risk (>0.333 probability by 2070). Over half of the critical habitats in Piñon Mesa and San Miguel had a >0.333 probability of burning by 2070 (Figure 6f) based on an average annual burn probability of 0.016 and 0.011, respectively (mean across critical habitats). These were also the only populations that had experienced notable wildfire between 1984 and 2016, which affected 3.6% of critical habitats in Sin Miguel Basin and 1.8% in Piñon Mesa (Figure 6f). CCS and Crawford both had an intermediate average wildfire risk of 0.004, respectively. Despite having some high-risk areas (Figure 5f), Dove Creek had a lower mean annual probability of wildfire overall (0.003) due to many zero risk areas owing to the large extent of cropland (Figure 5e). Gunnison and Poncha Pass had the lowest risk of burning (both 0.002), with no areas clearing the at-risk threshold (Figure 6f).

DISCUSSION

Integrated vulnerability assessments for effective sage-grouse conservation

Our integrated vulnerability assessment of Gunnison sage-grouse demonstrates how to advance goals for management-oriented CCVAs with a simple framework for integrating habitat vulnerability forecasts to identify the relative vulnerability of different populations and habitats under divergent climate change scenarios (Figure 1). This habitat-centered synthesis approach identified clear differences in vulnerability among spatially disjunct populations of Gunnison sage-grouse. We found the Gunnison population had the lowest overall vulnerability (in line with historical trends; Oyler-McCance et al., 2001), while four western populations were highly vulnerable with nearly 100% of critical habitats at risk (Figures 3 and 4). Our approach also elucidated how predominant threats differed across populations (Figure 5). However, novel invasion by pinyon–juniper conifers and invasive annual grasses were the most common threats facing critical habitats, driven in part by forecasted shifts in their climatic niches under climate change (Figure 6c). This highlights the importance of indirect climate effects, in agreement with predictions that climate effects on wildlife will be most strongly driven by how climate alters habitat conditions and biotic interactions (Araújo & Luoto, 2007; Cahill et al., 2013; Foden et al., 2019; Rowland et al., 2011).

The threat-specific habitat vulnerability maps we produced can be applied to target specific habitat interventions to mitigate these threats (e.g., planting sagebrush to address current or future deficits). Translating results of traditional trait-based, niche, or mechanistic CCVA approaches into management recommendations has been challenging (Langham et al., 2015), but designing CCVAs with clearer goals on how they will inform habitat management actions may help CCVAs better inform decision-making (Foden et al., 2019). Niche approaches usually project changes in wildlife ranges, while trait-based approaches rank relative vulnerability across species. While useful, neither focus on the suite of vegetation characteristics managed to provide critical habitats for wildlife (Dawson et al., 2011; Pacifici et al., 2015). A habitat-centered synthesis approach can provide spatial projections of manageable effects such as increasing risk of invasion (Figure 5c,d) that can be linked to site-specific monitoring and conservation plans, potentially accelerating the integration of CCVAs into management (Miller et al., 2022).

Application for management to conserve Gunnison sage-grouse

Our map products (downloadable as a U.S. Geological Survey Data Release; Van Schmidt et al., 2023) can serve as valuable tools for managers, if their limitations (discussed below) are properly considered. We here give brief examples of how datasets generated by our habitat-centered synthesis approach can be applied to spatially prioritize management of existing habitats to achieve Gunnison sage-grouse habitat goals while considering potential future threats (USFWS, 2020).

Relative overall vulnerability ranks (Figure 4) can be used to prioritize restoration and conservation efforts in areas likely to maximize returns on investments (Glick et al., 2011). Funding for restoration actions and translocation of birds from the core Gunnison population into the satellites is limited (USFWS, 2019), so triage and prioritization of populations may be necessary. Larger contiguous areas with high overall vulnerability (Figure 3) facing several threats (Figure 4) could be considered for triage where averting habitat degradation may not be possible. Confidence that habitats are vulnerable is greater in populations that face many cumulative threats such as CCS (Figure 4; Miller et al., 2022). Confidence is lower in populations with threats that vary widely across scenarios (i.e., a major difference in vulnerability between the Optimistic scenario and the Pessimistic scenario), such as Gunnison (Figure 4).

We found the satellite populations are both more at risk due to current habitat degradation (and small population size; USFWS, 2019) and more vulnerable to future threats than the Gunnison population (Figure 4). Preventing habitat degradation within the Gunnison population (particularly in a hot–dry climate; Figure 4c) is likely to be important in ensuring this area can continue to support the species' largest population (USFWS, 2019). Yet, sustaining the satellite populations is key to preserving representation of environmental heterogeneity and maintenance of genetic diversity and redundancy to minimize the risk of extinction due to local catastrophes (Shaffer & Stein, 2000). We found that four western populations (Piñon Mesa, Crawford, CCS, and eastern San Miguel) were most vulnerable (Figure 4). Piñon Mesa is the only satellite population believed to have a stable population, although it has depended on past translocations (USFWS, 2019), and we found its habitats were notably less at risk of future degradation (Figure 4). This population thus appears to be the subject to the least change within this unique ecoregion, along with eastern areas of San Miguel (Figure 4). Western areas of San Miguel and the CCS population were the single most vulnerable areas (Figure 3) due to possible development, drier climate with greater wildfire and invasion risk, and lower sagebrush cover (Figure 5), so these areas may not be the best place for projects that require stable environmental conditions. The Poncha Pass and Dove Creek populations have very small and/or declining populations (USFWS, 2019) but had below-average vulnerability to future habitat changes (Figure 6). Restoring habitat in Poncha Pass may yield long-term benefits if the population is able to persist. By contrast, the low vulnerability of Dove Creek (Figure 4) was partially because much of the habitat has already been developed for agriculture (Figure 5e), and the remaining rangeland was still highly at risk (Appendix S1: Figure S6), so opportunities for restoration may be limited.

The habitat vulnerability maps produced by this framework can be used for targeting specific interventions. For example, pinyon–juniper (and invasive annual grasses) removal efforts could be targeted at current early-stage encroachment (“partially degraded” in Figure 5c,d; Olsen et al., 2021). However, removal of established forest (“fully degraded” in Figure 5c) has low success at reestablishing other habitats (Miller, 2005) and is likely unnecessary because historic encroachment is not believed to be populations' limiting factor. Because the scale of pinyon–juniper threats dwarfed other stressors in the future (Figure 6), conifer removal could become increasingly important for management. Areas identified as at future risk under climate change (“at risk” in Figure 5d) could help target monitoring efforts, especially for areas that are currently viable sagebrush or mesic habitat (“stable” or “at risk” in Figure 5d). Areas with extant but trace sagebrush cover (1%–5%; light red in Figure 5a) could be prioritized for planting, if those areas are projected to be suitable for sagebrush in the future (Appendix S2: Figures S1 and S2). Because all mesic resources were at risk of drying in the hot–dry Pessimistic scenario (Figure 6), enhancing existing mesic resources by installing rock structures (Zeedykes) to increase mesic habitats' acreage and stabilize their hydrological cycles will likely be a key strategy for resilience (Maestas et al., 2018; USFWS, 2020). High-risk areas for wildfire (yellow in Figure 5f) could be prioritized for strategic fuel reduction and placement of fire breaks (Shinneman et al., 2018). Managers ranked residential development as the greatest threat (USFWS, 2019), and areas at high risk of future development (yellow in Figure 5e) could be targeted for establishing conservation easements (USFWS, 2020). Appendix S1: Section S1.2 contains a more comprehensive discussion of management applications using the Resist–Accept–Direct planning framework (Fisichelli et al., 2016; Schuurman et al., 2020).

Lastly, managers could incorporate other planning tools that provide complementary information, such as maps of where management actions would most increase Gunnison sage-grouse habitat use (Shyvers et al., in review). The USFWS' SSA also provides a complementary view of threats informed by regional on-the-ground knowledge. Appendix S1: Section S1.3 provides a more detailed comparison of our results with the conclusions and threat rankings of the SSA (Appendix S1: Table S6), highlighting areas of agreement and discussing discrepancies. We recommend this local knowledge and other spatial resources be used in conjunction with our spatial data products when making management decisions.

Limitations and uncertainty

Our analysis has important limitations to consider when applying results for conservation and management planning. First, our habitat-centered synthesis approach did not examine direct effects of climate change on Gunnison sage-grouse survival or reproduction due to worsening droughts, although this is of concern (Foden et al., 2019; USFWS, 2019). Second, most input vegetation models used niche approaches, which themselves have limitations and uncertainties. Errors can arise in niche model projections from unfilled but suitable habitats, limits on future dispersal, omitted key predictor variables (especially biotic interactions), microhabitats missed at the scale of the model, the assumptions of mathematical models, and cascading uncertainty from input climate projections (reviewed in Wiens et al., 2009). Third, while some input models were created specifically for this region (Colorado Natural Heritage Program, 2018; Colorado State Forest Service, 2017; Rangwala & Dewes, 2018), others relied on national models (Rigge et al., 2021; Short et al., 2020; Sohl et al., 2018), which may not be as accurate at local scales. Fourth, complex interrelationships between vegetation, fire, and land-use decisions will mediate vegetation shifts under climate change (Foden et al., 2019; Mantyka-Pringle et al., 2012; Michalak et al., 2022). For example, changes in wildfire frequency mediate the likelihood of encroachment into sagebrush by annual grasses and pinyon–juniper conifers (Miller et al., 2011), but our models did not account for this as most were independently developed.

Input threat datasets also had limitations that should be noted. For sagebrush and invasive annual grasses, impacts may be conservative because the climate inputs were model-averaged and therefore exhibited little change in mean precipitation (Table 2; Rigge et al., 2021). Unfilled but suitable habitats caused underestimation of percent cover estimates for invasive annual grasses, necessitating simplification of projections to an “invasion risk” index (Appendix S1: Section S1.2). Rocky microhabitats caused some current pinyon–juniper encroachment to be missed in projected pinyon–juniper suitability (Figure 5c, central Gunnison basin). Although we used a regionally downscaled model, projected changes in aridity index for mesic resources were still largely uniform (Figure 5b); restoration sites could instead be targeted based on suitable hydro-geomorphology (i.e., prioritization following methods of Nagel et al., 2014). The cited articles for each input data product (Table 1) have detailed discussions of their errors and uncertainties.

Despite the uncertainty associated with climate-related forecasts, such assessments can still provide valuable input into planning if uncertainty is properly considered (Miller et al., 2022; Rowland et al., 2011). Methods for quantifying uncertainty in synthesis CCVA are limited, but there are guidelines for identifying areas of greater or lesser confidence (Glick et al., 2011). Agreement among scenarios can help gauge where we may be more confident in changes. Combining projections based on different methodologies can identify areas stable under multiple models (i.e., our map of overall weighted mean vulnerability), further increasing confidence (Michalak et al., 2022; Pacifici et al., 2015). Nevertheless, true uncertainty in our projections is greater than this modeled range because uncertainty arises not only from scenarios and model design but also from input parameters, data collection, modeling errors, and system variability (Dunford et al., 2015; Evans, 2012; Glick et al., 2011; Miller et al., 2022). Uncertainty of climate and spatial vulnerability predictions generally increases at finer spatial scales (Glick et al., 2011), so our projections for individual raster cells likely have problematically high uncertainty. We chose to retain the native resolution of the finest datasets to preserve spatial heterogeneity representing local environmental conditions and provide managers with as much information as possible. However, we recommend managers focus on relative comparisons of larger contiguous tracts of high- or low-risk habitats or populations and not interpret maps at the pixel level.

Future directions

The habitat-centered synthesis approach to wildlife CCVA we developed can produce spatially explicit projections of habitat changes that allow for intrapopulation comparisons at a spatial scale matching the typical size of habitat management actions (<1 km²; Rowland et al., 2011; Shyvers et al., 2022). While conceptually simple, wildlife CCVA have rarely implemented synthesis approaches in a spatially explicit manner due to limited availability of climate-driven forecasts of vegetation and related threats. Triviño et al. (2013) assessed vulnerability based on multiple habitat shifts, although at a coarser 10-km scale that is less applicable to site-scale habitat management (Dawson et al., 2011). As high-resolution forecasts of vegetation changes and other threats are becoming more widely available, such synthesis approaches may become more viable and common, although challenges remain in assembling and integrating disparate datasets with different spatial extents, resolutions, and climate forcings. Our approach can serve as a template for other assessments, as our general framework can be applied to additional species and ecosystems where vegetation forecasts and species–habitat relationships are available.

CONCLUSIONS

We explored the use of a habitat-centered synthesis framework to assess climate vulnerability of Gunnison sage-grouse. Our approach integrated existing datasets on multiple habitats changes to rank the relative vulnerability of spatially disparate populations and provide information on specific habitat vulnerabilities at the local scale at which management decisions are made for this species (Dunford et al., 2015; Langham et al., 2015; Millikin et al., 2020). Varied changes in different locations and scenarios created a shifting mosaic of risk to critical habitats of Gunnison sage-grouse. Uncertainty in climate and habitat forecasts is high, yet there are risks in proceeding with management without anticipating such changes (Langham et al., 2015). Our framework does not predict the future but gauges where different kinds of change could be more likely to happen, which can be useful for planning and monitoring the actual degree of changes to take early action. Despite the limitations of spatial projections of future habitat changes, such analyses can provide an early view of plausible changes. Our methods were conceptually simple, requiring only knowledge of the species' biology to reclassify published datasets. If replicated for other species, our framework could help prioritize local siting of habitat actions as regional climate and land-use/land-cover projections proliferate, empowering managers to better plan how and where to resist, accept, or direct changes to landscapes under climate change.

AUTHOR CONTRIBUTIONS

Methodology, software, validation, formal analysis, investigation, data curation, writing—original draft, visualization: Nathan D. Van Schmidt. Conceptualization, methodology, writing—review and editing: Jessica E. Shyvers. Conceptualization, methodology, writing—review and editing, supervision, funding acquisition: Julie A. Heinrichs. Methodology formal analysis, formal analysis, writing—review and editing: D. Joanne Saher. Conceptualization, methodology, resources, writing—review and editing, supervision, project administration, funding acquisition: Cameron L. Aldridge.

ACKNOWLEDGMENTS

We thank the U.S. Geological Survey, the U.S. Bureau of Land Management (BLM) National Office, and the U.S. BLM Colorado State Office for funding this research. We thank BLM staff Christopher T. Domschke, Robin A. Sell, and Leah R. Waldner for helping to guide the design of this study. This work was possible because of the researchers who provided their forecast datasets and helped us work with them: Karin L. Decker & Michelle M. Fink (Colorado State University; pinyon–juniper), Lauren N. Hargis (CSFS; wildfire), Imtiaz Rangwala (CIRES; aridity index), and Matthew Rigge (USGS; sagebrush and annual herbaceous). Ashley L. Whipple, Michael S. O'Donnell, and Shawna J. Zimmerman assisted in data processing and interpretation. Thanks to the anonymous reviewers for providing their time and feedback to improve this manuscript. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the US Government.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

Data (Van Schmidt et al., 2023) are available from the USGS ScienceBase Catalog: .

References

Aldridge, C. L., D. J. Saher, T. M. Childers, K. E. Stahlnecker, and Z. H. Bowen. 2012. “Crucial Nesting Habitat for Gunnison Sage‐Grouse: A Spatially Explicit Hierarchical Approach.” Journal of Wildlife Management 76(2): 391–406.

Apa, A. D., K. Aagaard, M. B. Rice, E. Phillips, D. J. Neubaum, N. Seward, J. R. Stiver, and S. Wait. 2021. “Seasonal Habitat Suitability Models for a Threatened Species: The Gunnison Sage‐Grouse.” Wildlife Research 48(7): 609–624.

Araújo, M. B., and M. Luoto. 2007. “The Importance of Biotic Interactions for Modelling Species Distributions under Climate Change.” Global Ecology and Biogeography 16(6): 743–753.

Braun, C. E., S. J. Oyler‐McCance, J. A. Nehring, M. L. Commons, J. R. Young, and K. M. Potter. 2014. “The Historical Distribution of Gunnison Sage‐Grouse in Colorado.” Wilson Journal of Ornithology 126(2): 207–217.

Cahill, A. E., M. E. Aiello‐Lammens, M. C. Fisher‐Reid, X. Hua, C. J. Karanewsky, H. Yeong Ryu, G. C. Sbeglia, et al. 2013. “How Does Climate Change Cause Extinction?” Proceedings of the Royal Society B: Biological Sciences 280(1750): [eLocator: 20121890].

Colorado Natural Heritage Program. 2018. “Modeling Ecological Response to Support Adaptation Strategies.” Colorado Bureau of Land Management. https://cnhp.colostate.edu/wp‐content/uploads/download/documents/misc/FactSheet_EcosystemModeling_Final.pdf.

Colorado State Forest Service. 2017. “CO‐WRA Burn Probability.” co-pub.coloradoforestatlas.org.

Connelly, J. W., K. P. Reese, R. A. Fischer, and W. L. Wakkinen. 2000. “Response of a Sage Grouse Breeding Population to Fire in Southeastern Idaho.” Wildlife Society Bulletin 28(1): 90–96.

Davis, A. J. 2012. “Gunnison Sage‐Grouse Demography and Conservation.” PhD diss., Colorado State University.

Dawson, T. P., S. T. Jackson, J. I. House, I. C. Prentice, and G. M. Mace. 2011. “Beyond Predictions: Biodiversity Conservation in a Changing Climate.” Science 332(6025): 53–58.

Dunford, R., P. A. Harrison, J. Jäger, M. D. A. Rounsevell, and R. Tinch. 2015. “Exploring Climate Change Vulnerability across Sectors and Scenarios Using Indicators of Impacts and Coping Capacity.” Climatic Change 128(3): 339–354.

Esri. 2012. “‘Topographic’ [Basemap]. Scale Not Given. ‘World Topographic Map’.” http://www.arcgis.com/home/item.html?id=30e5fe3149c34df1ba922e6f5bbf808f.

Evans, A. 2012. “Uncertainty and Error.” In Agent‐Based Models of Geographical Systems, edited by A. J. Heppenstall, A. T. Crooks, L. M. See, and M. Batty, 309–346. Dordrecht: Springer Netherlands.

Falkowski, M. J., J. S. Evans, D. E. Naugle, C. A. Hagen, S. A. Carleton, J. D. Maestas, A. H. Khalyani, A. J. Poznanovic, and A. J. Lawrence. 2017. “Mapping Tree Canopy Cover in Support of Proactive Prairie Grouse Conservation in Western North America.” Rangeland Ecology & Management 70(1): 15–24.

Fisichelli, N. A., G. W. Schuurman, and C. H. Hoffman. 2016. “Is ‘Resilience’ Maladaptive? Towards an Accurate Lexicon for Climate Change Adaptation.” Environmental Management 57(4): 753–758.

Foden, W. B., B. E. Young, H. R. Akçakaya, R. A. Garcia, A. A. Hoffmann, B. A. Stein, C. D. Thomas, et al. 2019. “Climate Change Vulnerability Assessment of Species.” WIREs Climate Change 10(1): [eLocator: e551].

Gardali, T., N. E. Seavy, R. T. DiGaudio, and L. A. Comrack. 2012. “A Climate Change Vulnerability Assessment of California's At‐Risk Birds.” PLoS One 7(3): [eLocator: e29507].

GeoMAC. 2019. “GeoMAC Historic Wildland Fire Perimeters—2000‐2018.” Geospatial Multi‐Agency Coordination Group. https://data‐nifc.opendata.arcgis.com/datasets/nifc::historic‐perimeters‐combined‐2000‐2018‐geomac/explore.

Glick, P., B. A. Stein, and N. A. Edelson. 2011. Scanning the Conservation Horizon: A Guide to Climate Change Vulnerability Assessment. Washington, DC: National Wildlife Federation 168 pp.

Jin, S., C. Homer, L. Yang, P. Danielson, J. Dewitz, C. Li, Z. Zhu, G. Xian, and D. Howard. 2019. “Overall Methodology Design for the United States National Land Cover Database 2016 Products.” Remote Sensing 11(24): 2971.

Kinzig, A. P., and J. Harte. 2000. “Implications of Endemics‐Area Relationshiops for Estimates of Species Extinctions.” Ecology 81(12): 3305–3311.

Langham, G. M., J. G. Schuetz, T. Distler, C. U. Soykan, and C. Wilsey. 2015. “Conservation Status of North American Birds in the Face of Future Climate Change.” PLoS One 10(9): [eLocator: e0135350].

LEDee, O. E., S. D. Handler, C. L. Hoving, C. W. Swanston, and B. Zuckerberg. 2021. “Preparing Wildlife for Climate Change: How Far Have we Come?” The Journal of Wildlife Management 85(1): 7–16.

Lickley, M., and S. Solomon. 2018. “Drivers, Timing and Some Impacts of Global Aridity Change.” Environmental Research Letters 13(10): 104010.

Maestas, J., S. Conner, B. Zeedyk, B. Neeley, R. Rondeau, N. Seward, T. Chapman, and R. Murph. 2018. Hand‐Built Structures for Restoring Degraded Meadows in Sagebrush Rangelands: Examples and Lessons Learned from the Upper Gunnison River Basin, Colorado. Fort Collins, CO: Colorado State University.

Malcolm, J. R., C. Liu, R. P. Neilson, L. Hansen, and L. Hannah. 2006. “Global Warming and Extinctions of Endemic Species from Biodiversity Hotspots.” Conservation Biology 20(2): 538–548.

Mantyka‐Pringle, C. S., T. G. Martin, D. B. Moffatt, S. Linke, and J. R. Rhodes. 2014. “Understanding and Predicting the Combined Effects of Climate Change and Land‐Use Change on Freshwater Macroinvertebrates and Fish.” Journal of Applied Ecology 51(3): 572–581.

Mantyka‐Pringle, C. S., T. G. Martin, and J. R. Rhodes. 2012. “Interactions between Climate and Habitat Loss Effects on Biodiversity: A Systematic Review and Meta‐Analysis.” Global Change Biology 180(4): 1239–1252.

Michalak, J. L., J. J. Lawler, J. E. Gross, M. C. Agne, R. L. Emmet, H.‐W. Hsu, and V. Griffey. 2022. “Climate‐Change Vulnerability Assessments of Natural Resources in U.S. National Parks.” Conservation Science and Practice 4(7): [eLocator: e12703].

Miller, B. W., G. W. Schuurman, A. J. Symstad, A. N. Runyon, and B. C. Robb. 2022. “Conservation under Uncertainty: Innovations in Participatory Climate Change Scenario Planning from U.S. National Parks.” Conservation Science and Practice 4(3): [eLocator: e12633].

Miller, R. F. 2005. Biology, Ecology, and Management of Western Juniper (Juniperus occidentalis). Corvallis, OR: Oregon State University Agricultural Experiment Station.

Miller, R. F., S. T. Knick, D. A. Pyke, C. W. Meinke, S. E. Hanser, M. J. Wisdom, and A. L. Hild. 2011. “Characteristics of Sagebrush Habitats and Limitations to Long‐Term Conservation.” In Greater Sage‐Grouse: Ecology and Conservation of a Landscape Species and its Habitats, Vol. 38, edited by S. Knick and J. W. Connelly, Studies in Avian Biology 145–184. Berkeley, CA: University of California Press.

Millikin, R. L., R. Joy, J. Komaromi, M. Harrison, N. Mahony, and W. M. Vander Haegen. 2020. “Critical Habitat Identification of Peripheral Sage Thrashers under Climate Change.” Conservation Science and Practice 2(12): [eLocator: e290].

Nagel, D. E., J. M. Buffington, S. L. Parkes, S. Wenger, and J. R. Goode. 2014. A Landscape Scale Valley Confinement Algorithm: Delineating Unconfined Valley Bottoms for Geomorphic, Aquatic, and Riparian Applications. Fort Collins, CO: Rocky Mountain Research Station.

Nakicenovic, N., J. Alcamo, G. Davis, B. de Vries, J. Fenhann, S. Gaffin, K. Gregory, et al. 2000. IPCC Special Report on Emissions Scenarios. Berkeley, CA: Lawrence Berkeley National Laboratory https://escholarship.org/uc/item/9sz5p22f.

Ofori, B. Y., A. J. Stow, J. B. Baumgartner, and L. J. Beaumont. 2017. “Influence of Adaptive Capacity on the Outcome of Climate Change Vulnerability Assessment.” Scientific Reports 7(1): [eLocator: 12979].

Olsen, A. C., J. P. Severson, J. D. Maestas, D. E. Naugle, J. T. Smith, J. D. Tack, K. H. Yates, and C. A. Hagen. 2021. “Reversing Tree Expansion in Sagebrush Steppe Yields Population‐Level Benefit for Imperiled Grouse.” Ecosphere 12(6): [eLocator: e03551].

Oyler‐McCance, S. J., K. P. Burnham, and C. E. Braun. 2001. “Influence of Changes in Sagebrush on Gunnison Sage Grouse in Southwestern Colorado.” Southwestern Naturalist 46(3): 323–331.

Pacifici, M., W. B. Foden, P. Visconti, J. E. M. Watson, S. H. M. Butchart, K. M. Kovacs, B. R. Scheffers, et al. 2015. “Assessing Species Vulnerability to Climate Change.” Nature Climate Change 5(3): 215–224.

Picotte, J. J., K. Bhattarai, D. Howard, J. Lecker, J. Epting, B. Quayle, N. Benson, and K. Nelson. 2020. “Changes to the Monitoring Trends in Burn Severity Program Mapping Production Procedures and Data Products.” Fire Ecology 16(1): 16.

Rangwala, I., and C. F. Dewes. 2018. Downscaled Climate Projections at 800m Spatial Resolution for the North Central United States Based on the Multivariate Adaptive Constructed Analog (MACA) Method from Selective CMIP5 Models. Fort Collins, CO: ScienceBase. [DOI: https://dx.doi.org/10.21429/C9704J].

Wiens, J. A., D. Stralberg, D. Jongsomjit, C. A. Howell, and M. A. Snyder. 2009. “Niches, Models, and Climate Change: Assessing the Assumptions and Uncertainties.” PNAS 106: 19729–19736.

Young, J. R., C. E. Braun, S. J. Oyler‐McCance, C. L. Aldridge, P. Magee, and M. A. Schroeder. 2015. “Gunnison Sage‐Grouse Centrocercus minimus.” In Birds of North America, No 721, edited by A. Poole. New York: Cornell Lab of Ornithology.

Zimmerman, S. J., C. L. Aldridge, M. B. Hooten, and S. J. Oyler‐McCance. 2022. “Scale‐Dependent Influence of the Sagebrush Community on Genetic Connectivity of the Sagebrush Obligate Gunnison Sage‐Grouse.” Molecular Ecology 31: 267–328.