Modeling the Potential Habitat Gained by Planting Sagebrush in Burned Landscapes

Abstract

Many revegetation projects are intended to benefit wildlife species. Yet, there are few a priori evaluations that assess the potential efficiency of restoration actions in recovering wildlife habitats. We developed a spatial vegetation–habitat recovery model to gauge the degree to which field planting strategies could be expected to recover multi-factor habitat conditions for wildlife following wildfires. We simulated a wildfire footprint, multiple sagebrush (Artemisia spp.) planting scenarios, and tracked projected vegetation growth for 15 years post-fire. We used a vegetation transition framework to track and estimate the degree to which revegetation could accelerate habitat restoration for a Greater sage-grouse (Centrocercus) population within the Great Basin, western United States. We assessed the amount of habitat 15 years post-fire to estimate the degree to which revegetation could be expected to accelerate habitat restoration. Our results highlight a potential disconnect between the expansive areas required by wide-ranging wildlife such as sage-grouse and the relatively small areas that planting treatments have created. Habitat restorations and planting strategies that are intended to benefit sage-grouse may only speed up localized habitat restoration. This study provides an example of how linked revegetation–habitat modeling approaches can scope the expected return on restoration investment for habitat improvements and support the strategic use of limited restoration resources.

1. Introduction

Many revegetation projects are intended to benefit focal wildlife species; however, the degree to which wildlife directly benefit from specific vegetation restoration projects is often unquantified or poorly understood [1]. Although post-revegetation monitoring is often conducted to track plant establishment and growth, fewer long-term studies document habitat regeneration and wildlife responses to restoration actions. If post-restoration habitat assessments are conducted, they are often limited to evaluating a small subset of habitat factors, rather than the full environmental context as perceived by animals. In the absence of long-term studies, the conservation benefits of restoration actions can be assumed by scientists and practitioners, but not realized by wildlife. A range of habitat restoration projects could benefit from pre-assessments of habitat creation, before actions are fully designed and implemented, to quantitatively gauge the expected return on restoration investment. Predictive and projection-based spatial modeling can link planned actions with expected vegetation growth, habitat recovery, and population responses to gauge the short- and long-term benefits to wildlife populations. Such data-driven approaches can help assess the ability of different revegetation efforts or strategies to re-create the many specific environmental factors that constitute habitat. In this paper, we demonstrate the use of a spatial projection approach to evaluate the ability of past post-fire restoration practices to re-create habitat for a Greater sage-grouse (Centrocercus urophasianus (sensu Bonapart 1827, hereafter sage-grouse) population.

The sage-grouse is a widely distributed species that responds to several environmental factors at multiple spatial scales [2] and requires large areas to fulfill their seasonally varying resource needs. Sage-grouse depend on sagebrush (Artemisia species) and associated vegetation communities that are declining in western North America. Wildfires are increasing in number and severity in the sagebrush ecosystem, which has cumulative impacts on wildlife habitats and populations [3]. Wildfires can consume an entire seasonal range for a sage-grouse population [4]. The species’ strong fidelity to seasonal home ranges can limit emigration from burned areas [5], and sage-grouse often continue to use suboptimal habitats in burned landscapes [6,7,8]. Post-fire population declines are often attributed to reductions in survival or reproduction rather than emigration as a response to fire-induced habitat loss [6,9,10]. Entwined with threats from widespread wildfire, land-use change [11,12] and invasive grasses like cheatgrass (Bromus tectorum L.) are expanding across the Great Basin [13,14,15], fragmenting and degrading extant sage-grouse habitats. In addition, invasive grasses often replace less flammable vegetation (e.g., where <20% perennial grass cover) [16]) and reinforce the cheatgrass–wildfire cycle [17,18,19].

Wildfire kills most big sagebrush plants (A. tridentata (Nutt.) W. A. Weber) and postburn recovery (i.e., return to its pre-fire growth trajectory) is slow [20]. The rate of growth to reach mature pre-disturbance sagebrush height and cover conditions depends on the Artemisia species and climate, but generally is estimated at 35–100 years for mountain big sagebrush (A.t. vaseyana (Rydb.) Beetle) and 50–120 years for Wyoming big sagebrush [1] A.t. wyomingensis [12,21]). Limited natural seed dispersal further constrains the geographic extent of sagebrush recovery [22,23,24] Without actions to speed sagebrush re-establishment and suppress fires, sagebrush vegetation communities, and the wildlife that depend on them, may have difficulty keeping pace with fire-induced loss of habitat [25]. The limited availability of resources (e.g., seeds, plants, money, people) and sagebrush revegetation strategies (e.g., seeding, transplanting) typically restricts the extent to which sage-grouse habitat restoration can be undertaken, often to a fraction of the area affected by a single burn [26].

Revegetation projects undertaken after wildfire in former sage-grouse habitats are initiated to expedite habitat restoration and expected to enhance sage-grouse population recovery [27]. However, the degree to which post-fire seeding and planting actions will translate into sage-grouse habitats is unclear. Sagebrush seeding has partial success in re-creating habitats due to poor seed establishment, cheatgrass invasion and competition, and insufficient sagebrush height achievement to support sage-grouse needs, even after 20 years [1]. Additionally, not locating restoration projects where they have the best chance of benefiting sage-grouse has been an issue with current strategies [1]. The planting of sagebrush transplants (i.e., planting greenhouse-grown sagebrush plugs; hereafter, planting) is thought to be well suited to meeting wildlife-focused habitat restoration objectives because local site selection, patch size, and planting densities can be controlled and optimized to create specific cover conditions. However, the use of plantings to create sage-grouse habitats is in field testing [28] and it is unclear if plantings, in combination with post-fire vegetation changes, will result in the environmental conditions that sage-grouse are known to select (e.g., moderate sagebrush cover, perennial grass/forb cover, low tree cover, etc.). Such revegetation plans could benefit from understanding if the pace of revegetation can jump-start sage-grouse habitat recovery (i.e., within 15 years post-fire), and if restored areas can cover a sufficient area to constitute a seasonal habitat for a small population.

Our goal with this study was to develop a data-driven framework that would allow managers and conservationists the ability to forecast the expected recovery of wildlife habitats from field-based restoration strategies after large-scale disturbances like wildfire. To scope the degree to which wildlife-centric sagebrush planting strategies could restore habitats for sage-grouse, we developed a spatial habitat restoration model. First, we linked vegetation transitions (fire, vegetation loss, plantings, regrowth) with pre-fire habitat selection models and projected sage-grouse habitat recovery through time. Second, we demonstrated the framework’s utility by evaluating the planting strategies and factors that best re-established the seasonal habitat conditions selected by sage-grouse. We expected factors that increased the extent of sagebrush (i.e., the number and survival of plants) to influence habitat recovery across a burn-disturbed extent more than factors that determine the arrangement (siting) and distribution of plants (density, patch size) because sage-grouse typically select habitats based on landscape-level features before site-level factors [29]. In addition, because sage-grouse habitat needs vary by season, we expected greater improvements to seasonal habitat recovery when planting strategies addressed key seasonal habitat characteristics. For example, planting fewer but larger patches could aid the recovery of sagebrush vegetation conditions (e.g., plant height or percent cover) that cue winter habitat selection for sage-grouse (as indicated by the large spatial scales of winter habitat selection models [30]).

2. Methods

2.1. Case Study and Location

Sage-grouse are considered a sagebrush obligate based on the species’ dependence on sagebrush plant communities throughout all stages of life. Large, contiguous habitat patches are also critical for sage-grouse to meet seasonal habitat needs, which generally require greater than 10% sagebrush canopy cover, sagebrush shrub heights that remain exposed in deep snow (i.e., >30–40 cm), and sufficient grass/forb coverage (>15% canopy cover) for summer forage and invertebrate prey habitat [4]. We used the Tuscarora area of northern Nevada (Figure 1) to develop our proof-of-concept model. Tuscarora has a well-studied sage-grouse population, a history of fire disturbance, and existing seasonal (spring, summer, winter) sage-grouse habitat selection models with representative pre-fire sagebrush habitat conditions [30,31]. The predominant vegetation communities in the Tuscarora region consist of sagebrush steppe and mixed mountain shrub, characterized by big sagebrush (Wyoming and mountain species) and other species of sagebrush (e.g., low sagebrush [A. arbuscula Nutt.]) that receive an average annual precipitation of 304.8 mm; ~12 inches.

2.2. Model Overview

Within an occupied area of known sage-grouse habitat at the study site, we simulated (1) a hypothetical fire footprint, (2) the fire-induced loss of vegetation, (3) the planting of sagebrush seedlings, and (4) the regrowth of sagebrush and other vegetation over a 15-year period. We used sagebrush growth equations derived from field data and vegetation transitions based on the prevalence of herbaceous components (e.g., age-specific changes in percent cover) to change vegetation within burned and planted areas [28]. For each year, we projected the vegetation changes into habitat recovery maps for sage-grouse using an existing habitat selection algorithm [31,32]. We calculated the annual proportion of suitable habitat changed by sagebrush planting restoration efforts and evaluated alternative planting scenarios (

Table 1. Sagebrush (Artemisia spp.) planting factors included in 48 alternative planting scenarios (factorial ANOVA design) used to simulate vegetation regrowth in a fire footprint (35,362 ha). Vegetation regrowth was linked with habitat selection models and used to project and evaluate seasonal habitat recovery for Greater sage-grouse (Centrocercus urophasianus) in the Tuscarora region of Nevada. See the Supporting Information for details on the planting factors evaluated (e.g., number of patches) and spatial targeted (e.g., transplants located in former habitat) or non-targeted simulation scenarios.

Factor	Condition	Alternative(s)
Planting Effort	Single year (350,000 plants)	Multi-year a (1.5 million plants)
Spatial Targeting	Any burned site	Nesting habitat (burned)
Transplant Density	Low (112 plants/30 m pixel)	High (235/30 m pixel)
Patch Size	Several small (9 ha; 300 × 300 m)	Few large (71 ha; 843 × 843 m)
Plant Survival	30%	70, 100%

^a Time since fire (TSF) years 2–4.

) to identify the key factors influencing habitat restoration outcomes.

2.3. Sagebrush Planting Scenarios

We approximated a fire footprint (35,362 ha) in an arbitrary area of the study area that contained habitats for the spring (breeding), summer (brood-rearing), and winter seasons, which removed mapped vegetation within the burn perimeter. To match ongoing field trials planting 1-year-old sagebrush seedlings grown from seeds in a nursery, we digitally planted patches of seedlings within the burn perimeter in the second modeled year (time since fire, TSF = 2) and projected growth for 15 years. For each year post-fire, we incremented the planted seedling height and area using sagebrush height and canopy area growth curves from post-fire sagebrush planting sites in the Great Basin ([28], (mountain and Wyoming big sagebrush combined; see Figures S2–S8, Supplemental Information for growth trajectories)). We allowed 70% or 30% of seedlings to survive the first year of planting, congruent with variations in field estimates of seedling survival [22]. To assess the degree to which replacing lost sagebrush could change habitat restoration outcomes, we also modeled a scenario with 100% survival. We projected the growth of planted seedlings to their height and canopy cover at year 15 and calculated the percent cover of planted sagebrush in pixels for each year after fire and planting treatment (see Tables S11–S15, Supplemental Information). This allowed us to integrate the projected changes in sagebrush cover resulting from planted sagebrush growth in our annual recalculations of potential habitat.

We created alternative planting scenarios that varied the spatial arrangement of a controlled number of seedlings to identify design strategies that optimized the use of plants. We varied the density (plants per unit area), location (targeted or non-targeted to pre-fire habitat), patch size, survival rates, and number of seedlings (Table 1), based on studies documenting previous sagebrush revegetation outcomes. For targeted scenarios, we located sagebrush transplants in previous sage-grouse spring (breeding) habitat areas. For non-targeted planting scenarios, transplants were placed in accessible areas that contained sagebrush, did not occur in riparian corridors, and were identified as habitats prior to fire (i.e., defined as suitable in habitat selection models). We contrasted scenarios representing the ambitious efforts of a single-year seedling plantation (approximately 350,000 plants; sensu [22]) to those of a multi-year planting effort (1,500,000 plants over TSF 2–4). Sagebrush seedlings were planted in several small patches (9 ha each) or a few large (71 ha each) patches. All scenarios used plant densities that were likely to result in the minimum percentage of sagebrush cover selected by sage-grouse [28]. We also examined whether planting a greater density of the same number of seedlings (235 plants/30 m²) near burned leks (<6 km) and nesting areas could improve habitat recovery for sage-grouse compared to a lower density (112 plants/30 m²) within each patch, which was randomly located. We randomized the locations of planting sites in the non-strategic placement scenarios but separated patches by 300–1200 m and constrained locations to areas with >5% pre-fire sagebrush (indicating suitable sites for plant growth), and considered suitable topography (i.e., elevation, slope, landscape position indices, terrain ruggedness, distance to water bodies). These planting scenarios used the same number of plants but resulted in different spatial distributions and total planted acreage among scenarios (135–1129 ha; see Tables S11–S13, Supplemental Information).

2.4. Vegetation Regrowth

We initiated vegetation transitions within the burn perimeter to return fast-growing herbaceous vegetation the year after the hypothetical fire event (Figure 2; TSF = 2). Using spatial data that included mapped inputs for sage-grouse habitat selection models (

Table 2. Model-averaged resource selection function (RSF) covariate coefficient estimates (β), scale (moving window, distance to), and coefficient 95% confidence intervals used to derive Greater sage-grouse (Centrocercus urophasianus) spring (mid-March to June), summer (July to mid-October), and winter (November to early March) seasonal habitat selection maps in Tuscarora, Nevada ^a. Habitat selection was linked to post-fire revegetation maps of simulated fire-induced loss, sagebrush planting, survival and regrowth, and vegetation transitions to evaluate the projected potential for habitat recovery benefits in restoration.

	Spring			Summer			Winter
Covariate	Scale	Model-Averaged β	95% CI	Scale	Model-Averaged β	95% CI	Scale	Model-Averaged β	95% CI
Annual grass	661.4 ha	−21.8	−25.15, −18.45	661.4 ha	−83.2	−90.05, −76.36	661.4 ha	−106.86	−115.55, −98.17
Bare ground	661.4 ha	−6.68	−7.00, −6.36	661.4 ha	−14.93	−15.31, −14.54	-	-	-
Cropland	-	-	-	661.4 ha	6.4	6.05, 6.74	-	-	-
Forest	61.5 ha	−13.32	−14.92, −11.71	61.5 ha	−22.23	−23.61, −20.85	661.4 ha	−17.7	−19.96, −15.44
Herbaceous	8.7 ha	7.77	7.37, 8.17	661.4 ha	−4.21	−4.80, −3.62	661.4 ha	15.62	14.69, 16.55
Non-sagebrush shrub	8.7 ha	−25.41	−26.44, −24.38	-	-	-	61.5 ha	−35.58	−37.16, −34.00
Sagebrush *	61.5 ha	71.81	68.70, 74.93	661.4 ha	119.19	113.65, 124.72	661.4 ha	68.86	63.08, 74.63
Sagebrush height	-	-	-	-	-	-	661.4 ha	13.2	12.54, 13.86
Riparian	8.7 ha	−3.18	−3.68, −2.69	661.4 ha	−3.34	−4.35, −2.34	61.5 ha	−10.51	−11.71, −9.31
Land cover b	8.7 ha	−0.17	−0.20, −0.15	61.5 ha	−0.04	−0.06, −0.01	61.5 ha	−0.46	−0.49, −0.43
Dist. to cropland	Exp. decay	0.57	0.45, 0.69	-	-	-	Linear	−0.26	−0.28, −0.23
Dist. to nearest stream	Linear	3.55	3.39, 3.73	-	-	-	Linear	4.12	3.89, 4.35
Dist. to spring	Linear	0.08	0.05, 0.10	Linear	0.08	0.05, 0.10	Linear	−0.51	−0.55, −0.47
Dist. to water body	Exp. decay	−2.90	−3.08, −2.73	Exp. decay	−0.87	−1.03, −0.71	Exp. decay	−3.25	−3.49, −3.01
Dist. to wet meadow	Linear	−0.19	−0.20, −0.18	Linear	−0.184	−0.19, −0.18	Linear	−0.15	−0.16, −0.14
Elevation	Linear	−6.67	−6.95, −6.38	-	-	-	Linear	−3.9	−4.19, −3.61
Roughness index c	1 ha	−12.60	−13.12, −12.09	1 ha	−13.39	−13.88, −12.91	1 ha	−18.73	−19.50, −17.95
TPI d	510 m	0.01	0.01, 0.02	-	-	-	2010 m	−0.003	−0.004, −0.002

^a Reproduced from [30]. ^b A categorical variable representing the variety of land cover types measured as the number of unique land cover types per scale area. ^c Calculated as 1 − |r|/n where |r| = sqrt[(SUM x)^2 + (SUM y)^2 + (SUM z)^2], z = 1× cos(slope), xy = 1× sin(slope), x = xy × sin(aspect), y = xy × cos(aspect). * Variable used to represent sagebrush plantings was sagebrush (other). ^d Topographic position index calculated for each cell in a raster as elevation minus average neighborhood elevation determined by annulus moving window analysis.

; including [32]), we returned the percent cover of herbaceous vegetation (forbs, perennial grass, and annual grass) to pre-fire conditions (in each pixel), while adjusting the proportion of bare ground that remained. Although herbaceous vegetation may take longer than a year to return to pre-fire conditions, we simplified this aspect in our model to focus on evaluations of sagebrush planting. Where cheatgrass was present before the simulated fire, we allowed percent cover to increase post-fire (TSF years 2–4) within the burned extent [16]. In pixels with perennial grass cover <20% [32], annual grass cover increased while bare ground and perennial grass cover commensurately decreased. Vegetation transition rules differed among planted and non-planted areas within the burn perimeter to accommodate variability in sagebrush growth within planting sites. Our focus for this study was gauging the influences of sagebrush planting on the potential for sagebrush habitat recovery; therefore, we did not transition vegetation outside of the burn footprint or attempt to model regrowth relative to seeding. Additional details on vegetation initialization (Tables S16–S17), transitions, and growth rates (Figures S2–S8) can be found in the Supplemental Information.

2.5. Projecting Seasonal Habitat Recovery

We used existing seasonal habitat models (resource selection function, RSF) for the Tuscarora region to describe pre-fire sage-grouse habitat. We also used the relationships from these models, coupled with our projected post-fire vegetation transitions, to update the habitat maps to reflect plantings and vegetation changes that influence sage-grouse resource selection. The Tuscarora habitat models used a typical logistic regression RSF approach that contrasted the habitat characteristics (e.g., environmental covariates) at sage-grouse locations collected during spring (mid-March to June; breeding and nesting period), summer (July to mid-October; brood-rearing period), and winter (November to early March) with population-level availability for each season (see [30,31] for additional details on RSF models). The environmental covariates included vegetation and bare ground cover, sagebrush height, distance to water features, topographic features, and number of land cover types. These covariates varied in their spatial scales (8.7 ha, 61.5 ha, 661.4 ha) and strength of influence (Table 2). A detailed model description can be found in the Supplementary Information (Figure S1, Tables S1–S10). The percent cover of vegetation (all sagebrush [psb], annual grass [pag], perennial grass [ppg], non-sagebrush shrubs [pnns], forests and other conifer pinyon juniper [pj], and bare ground [pbar]) were converted to discrete states based on dominant cover and precedence each year of the simulated period (TSF = 1–15) and used to project habitat suitability relative to existing seasonal resource selection models for sage-grouse.

For each year of the simulation (time since fire; TSF 1–15), we updated the vegetation maps to include the influences of fire-induced loss, sagebrush planting, survival and regrowth, and vegetation transitions (Figure 2). After incorporating these changes into vegetation, we recalculated the seasonal habitat equation to project a new habitat map linked to each planting scenario.

2.6. Assessing Habitat Recovery

We assessed post-fire projected habitat recovery by calculating the amount of habitat in each habitat suitability category (High, Moderate, Low, Non-suitable) for spring and summer and summed all habitat categories for winter. We compared the amount of pre-fire habitat to 1- and 15-year post-fire habitat. To evaluate the habitat contributions of alternative planting scenarios, we used analysis of variance (ANOVA) to quantify the relative influence of planting factors (Table 1; n = 48) on total habitat recovery (e.g., percent of the burned landscape exceeding the non-habitat threshold of the habitat models; see Section S1 of the Supporting Information). We used generalized linear models (GLMs) to assess the influence of planting factors on sage-grouse habitat recovery by ranking the resulting standardized effect sizes of variables in each seasonal habitat evaluation. ANOVA and GLM analyses were performed with JMP statistical analysis software [33] following standard procedures [34].

3. Results

One year after the loss of all vegetation within the burn perimeter, fast-returning herbaceous vegetation supported the partial recovery of habitat (TSF 2; Figure 3). On average, fast-returning vegetation (e.g., herbaceous) allowed the recovery of a small portion of spring (13.5%) and summer (0.63%) habitat, and comparatively more (43.6%) winter habitat.

By the end of the simulated study (year 15), much of the landscape recovered to minimum habitat conditions. However, restoration to high suitability conditions was limited (Figure 4; yellow high suitability). In the summer brood-rearing season, much of the former habitat was restored (74–84%) to its former coverage in the minimally suitable habitat category, but only half (40–56%) of the highly selected habitat was restored. Minimally suitable winter habitat was fully restored (115–123%; total suitable) to its pre-fire coverage, but highly suitable winter habitat was not generally restored.

Scenarios with more planted seedlings resulted in greater habitat recovery (e.g., spring models;

Table 3. Statistical (ANOVA) ranking of sagebrush (Artemisia spp.) planting factors influencing the simulated recovery of Greater sage-grouse (Centrocercus urophasianus) spring (mid-March to June), summer (July to mid-October), and winter (November to early March) seasonal habitats in Tuscarora, Nevada. Recovery was measured by the percentage of total suitable habitat 15 years after a hypothetical fire (35,362 ha). Non-targeted refers to simulated sagebrush planting scenarios that did not locate seedlings in a former spring habitat. Survival describes seedling mortality rates for the planted sagebrush. Total plants were calculated as the sum number of seedlings planted. Patch size describes the area (ha) of each patch of planted seedlings. Plant density expresses the number of plants per unit area. Generalized linear models (GLMs) were used to produce standardized estimates of planting factors and rank the effect sizes in each seasonal evaluation. A p-value α ≤ 0.05 was used to determine the evidence of the effect.

	Estimate	Std. Error	T Ratio	Pr (>|t|)	Standardized Estimate	Rank
Spring a
Intercept	70.749	0.126	560.80	<0.001	71.100
Non-targeted	0.153	0.029	5.34	<0.001	0.152	1
Survival	0.003	<0.001	3.05	0.004	0.106	2
Total plants	1.42 × 10−7	<0.001	2.81	0.008	0.082	3
Summer b
Intercept	48.558	0.518	93.67	<0.001	51.560
Total plants	2.24 × 10−6	<0.001	10.73	<0.001	1.289	1
Survival	0.029	0.004	6.84	<0.001	1.003	2
Non-targeted	0.545	0.120	4.53	<0.001	0.545	3
Plant density	−0.004	0.002	−2.05	0.047	−0.246	4
Winter c
Intercept	45.997	0.342	134.49	<0.001	47.142
Total plants	6.16 × 10−7	<0.001	4.71	<0.001	0.354	1
Patch area	0.011	0.002	4.63	<0.001	0.334	2
Survival	0.007	0.003	2.60	0.0129	0.233	3

^a Plant density and patch area were non-significant; n = 48, df = 41, R² = 0.56. ^b Patch area was non-significant; n = 48, df = 41, R² = 0.82. ^c Non-targeted and plant density were non-significant; n = 48, df = 41, R² = 0.54.

; R² = 0.56). Similarly, increased seedling survival or the replacement of dead seedlings enhanced habitat restoration. For instance, over half (55%) of the highly selected spring habitat returned to pre-fire conditions in the 30% survival scenario, whereas full restoration was achieved in the 100% seedling survival scenario. The overall planting effort (i.e., number of seedlings) and seedling survival were the key factors influencing the proportion of restored habitat for all seasons (spring, summer, winter).

Targeted site selection, patch size, and plant density factors had lesser effects on habitat recovery than the scale of planting effort (number of seedlings) and the survival of seedlings. Spatial targeting of planting sites marginally increased the percentage of highly selected habitat (spring, 0.29%; summer, 0.27%; 30% seedling survival scenario). When all factors were considered in statistical analyses, the non-targeted scenarios outperformed the targeted scenarios in the proportion of total habitat restored. Greater habitat recovery was generally observed in scenarios that did not spatially target plantings in former nesting habitat.

Similarly, in the spring and summer models, greater plant densities and larger patch sizes did not significantly influence the amount of recovered habitat. For example, only 1–3% improvements in the amount of habitat were achieved by changing patch sizes or plant densities (30% seedling survival scenario). In the winter model, the plantings arranged in fewer larger patches resulted in greater habitat improvements than those in more and smaller patches (Table 3; R² = 0.54). Plant density did not increase the amount of restored winter habitat. Simulated patch sizes (9–71 ha) were based on past transplant plot sizes [22]; however, many patch sizes were small in comparison to the scales at which sage-grouse respond to resources in the Tuscarora area (8.8–661 ha). Larger patches were more influential in recovering winter habitat (ranked second) than scenarios with the same number of plants distributed in smaller patches.

4. Discussion and Conclusions

We scoped the degree to which planting sagebrush seedlings might aid in restoring sage-grouse habitats after a wildfire. We used planting scenarios that were based on past, real-world planting efforts (number of seedlings, number of years of plantings), and placements designed to benefit sage-grouse. We found that planting efforts are likely to jump-start habitat restoration; however, realistic planting areas were small relative to the size of moderate–large wildfires. Much of the burned landscape was unplanted and habitat restoration was spatially incomplete across the burn. The unassisted return of herbaceous vegetation hastened the return of selected vegetation conditions across the burn and accounted for much of the early habitat restoration in both unplanted and planted areas. In unplanted areas, restoration was incomplete due to a lack of sagebrush. In planted areas, accelerated sagebrush restoration could have near-term benefits for sage-grouse that return to familiar sites after fire [6,10,35]. The earlier return of sagebrush in planted sites could provide food resources and restore canopy cover that serves as refuge from predators [36], improving localized survival and reproduction relative to other strategies such as seeding [1] and natural revegetation [20,25]. However, field and simulation experiments are needed to elucidate the importance of localized plantings in avoiding the loss or decline of resident sage-grouse.

The degree to which habitat was restored, locally and regionally, depended on the planting scenario. The most effective strategies planted the most sagebrush, with the highest seedling survival (or replacement). Intense planting strategies created more habitat than lesser efforts but still often fell short of restoring total pre-fire amounts or high-suitability areas. Additional seedlings and complementary habitat restoration efforts (e.g., seeding other vegetation components) are likely needed to restore sage-grouse habitats to highly selected vegetation and resource conditions, but may be logistically challenging. Future research could quantify and compare habitat gains associated with individual and combined planting, seeding, and natural revegetation approaches.

Optimizing patch size, plant density, and spatial targeting was less effective at restoring the habitat at the landscape level, relative to scenarios that increased the number of plants. Dispersed planting locations, rather than locations targeting leks and spring habitat, better matched the larger spatial scales at which sage-grouse select habitats in this region [30,31], resulting in greater habitat improvements. However, if these random planted sites are infrequently found and used by sage-grouse, a dispersed planting strategy may not encourage use by wildlife. The distribution of seedlings among several small or fewer large patches yielded mixed results. During winter, when sage-grouse select sagebrush at larger spatial scales, our simulations found that larger patches better augmented habitats. Hence, arranging plants in larger patches could facilitate the recovery of winter habitats. The creation of several small high-quality patches may also benefit individually nesting birds by providing resources within a frequently used seasonal location, as sage-grouse are known to have high fidelity to previous breeding and nesting areas [37], even after fires [9]. Without restored nesting habitats after a burn, grouse often return to breed in burned areas and experience reduced reproductive output compared to non-burned areas [6,9]. This ecological trap can cause population decline [9,38,39]. We did not evaluate the influence of plantings on local or microsite habitat recovery. However, we postulate that high-density plantings in small, targeted sites could be effective in reconstructing preferred local or microsite conditions that support sage-grouse reproduction. We tested plant densities that were likely to result in sagebrush cover preferred by sage-grouse and tighter spacing did not increase habitat recovery at the scale at which this population selects habitat. However, plant density is an important factor in establishing adequate cover for sagebrush obligate species.

A strength of using habitat selection models to project habitat restoration is the ability to quantitatively consider how multiple environmental factors and changes could influence habitat restoration. We found this approach to be insightful in gauging how much habitat restoration may be needed to achieve habitat restoration goals, and the degree to which restoration can be attributed to planting, relative to other vegetation changes. However, if the objective is to predict the specific outcomes of site-level actions, additional information (e.g., finer scaled habitat maps) could be added to increase the specificity of the results. If possible, future studies may consider relaxing assumptions including the consistency of habitat selection across pre- and post-fire landscapes. Sage-grouse often maintain site fidelity and make few adjustments to optimize fitness in post-fire landscapes [6,9,35,40]; however, other species may differ in this respect. If selection behavior temporarily or permanently changes after disturbance, pre-disturbance habitat associations may not capture post-disturbance habitat needs [41]. It may also be important to consider the possible negative consequences of encouraging animals to re-inhabit recovering habitats, particularly if restoration sites become ecological traps [42].

As intensifying and recurring fires contribute to the decline of sagebrush and greater sage-grouse [9,25], revegetation and wildlife recovery plans in the Great Basin can benefit from sagebrush planting strategies with direct and quantitative links to wildlife habitat needs. Generally, our results indicate that sagebrush plantings can be used to accelerate the restoration (within 15 years) of localized sage-grouse habitats following fire, but this strategy is difficult to implement at the spatial extents used by wide-ranging species such as sage-grouse. Plantings may accelerate the recovery of sagebrush [28], but they may not be of sufficient size or quality to support a local population or persistent populations. For example, an estimated tens to a few hundred sage-grouse might benefit from the revegetation actions outlined in our case study scenarios, depending on the planting intensity, seedling survival, and placement. More rigorous estimates of population benefits should consider wildlife responses to fire disturbance (e.g., emigration, changes in space use and density), and vegetation restoration including use of restored sites and associated fitness outcomes. These data are seldom collected in conjunction with restoration and revegetation projects [43,44], but could help design more impactful vegetation restoration strategies.

By linking revegetation scenarios with habitat selection relationships, we found a mismatch in the scale of disturbance and revegetation efforts, indicating that wildlife habitat restoration may require multiple strategies across multiple spatial scales to create the vegetation cover sufficient to support post-disturbance habitat selection. Restoration strategies driven by assumptions that revegetation actions will benefit wildlife habitat, without considering the multiple factors and scales that influence habitat selection, may risk creating poor quality habitats or unrealistic expectations for habitat outcomes and population improvements. Explicit modeling of the range of expected improvements to wildlife habitats could indicate the return on restoration investment and help guide the selection of wildlife-centric revegetation actions.