Grassland intactness outcompetes species as a more efficient surrogate in conservation design

Mapped representations of species−habitat relationships often underlie approaches to prioritize area‐based conservation strategies to meet conservation goals for biodiversity. Generally a single surrogate species is used to inform conservation design, with the assumption that conservation actions for an appropriately selected species will confer benefits to a broader community of organisms. Emerging conservation frameworks across western North America are now relying on derived measures of intactness from remotely sensed vegetation data, wholly independent from species data. Understanding the efficacy of species‐agnostic planning approaches is a critical step to ensuring the robustness of emerging conservation designs. We developed an approach to quantify ‘strength of surrogacy’, by applying prioritization algorithms to previously developed species models, and measuring their coverage provided to a broader wildlife community. We used this inference to test the relative surrogacy among a suite of species models used for conservation targeting in the endangered grasslands of the Northern Sagebrush Steppe, where careful planning can help stem the loss of private grazing lands to cultivation. In this test, we also derived a simpler surrogate of intact rangelands without species data for conservation targeting, along with a measure of combined migration representative of key areas for connectivity. Our measure of intactness vastly outperformed any species model as a surrogate for conservation, followed by that of combined migration, highlighting the efficacy of strategies that target large and intact rangeland cores for wildlife conservation and restoration efforts.

Conserving the world's biodiversity is a global endeavor relying on science-based strategies that identify and protect the connected landscapes necessary to account for biodiversity through area-based conservation strategies (Berger, 2004;Kullberg et al., 2019;Maxwell et al., 2020;Saura et al., 2018).Selecting or prioritizing areas to focus conservation efforts rests on efficiently allocating limited resources towards maintaining ecosystem functions, such as biodiversity, in perpetuity.Over 20 million km2 of protected areas exist currently, encompassing $15% of the global terrestrial land surface, but still fall short of the now historical Aichi 17% targets (UNEP-WCMC, IUCN, and NGS, 2018).Expanding the global estate of lands designated as protected, or increasing protection of existing lands, is the traditional approach towards reaching goals.While area-based strategies are but one tool for achieving biodiversity targets, land acquisition can be costly, and privately-owned working lands whose conservation can be incentivized through publicÀprivate partnerships often harbor most of the biodiversity (Baier, 2020;Doremus, 2003), underscoring the role of private working lands as part of the portfolio to meet conservation targets.For example, the U.S. government is relying on community-led and nationally-scaled conservation of biodiversity-rich private lands in their national strategy (US Department of Interior, 2021).Ultimately the fate of conservation rests upon a matrix of multipleuse public and privately-owned lands where biodiversity can be sustained through targeted conservation and restoration actions, conservation leases, and perpetual easements (Jones, Downey, et al., 2019;Naugle et al., 2019).
Area-based conservation planning strategies for wildlife have historically been driven using surrogate species approaches for a variety of reasons (Rodrigues & Brooks, 2007).Namely, efficient use of limited resources guides the collection and analysis of data on a few selected species, for which in ideal cases, much effort has gone into judiciously selecting effective surrogates (Caro & O'Doherty, 1999;Fleishman et al., 2000).In many other cases, species driving conservation strategies aren't selected, but rather adopted by statute given their protected status, governmental trust responsibilities, or a consumptive-use funding strategy.For example, pronghorn (Antilocapra americana) is one of the priority species identified in a U.S. Department of Interior Secretarial Order (Stemler, 2020) that looks to not only maintain seasonal habitat but also to connect these areas through the identification and management of migratory corridors.Critical tests of surrogate species approaches have demonstrated their equivocal effectiveness in practice (Lindenmayer & Westgate, 2020), with breakdowns typically occurring among cross-taxon complementarity in conservation actions (e.g., Micheletti et al., 2023).Yet surrogacy has continued to dominate large-scale conservation plans in many biomes.
A valuable case study of conservation planning via surrogate species involves that of greater sage-grouse (Centrocercus urophasianus; hereafter sage-grouse), a species that has driven public investments in conservation among sagebrush rangelands in the western U.S. (USFWS, 2013).Large-scale assessments have demonstrated that sage-grouse have largely been a robust surrogate for sagebrush-associated species via a conservation design network aptly called Priority Areas for Conservation (PACs; Pilliod et al., 2020;Runge et al., 2019;Smith et al., 2019).Similarly, non-area based conservation actions targeted in PACs have benefited other imperiled species, such as benefits provided from the removal of expanding conifers (Donnelly et al., 2017;Holmes et al., 2017) and from more than 2900-km 2 of perpetual conservation easements that maintain usable space for wildlife through private working lands (Copeland et al., 2014;Tack et al., 2019).In addition, certain PACs in Montana and Wyoming have been identified as 'connectivity areas' specifically to conserve known migratory pathways for sage-grouse.However, research has also found that indices of sage-grouse occurrence and abundance can be equivocally (Carlisle & Chalfoun, 2020;Smith et al., 2021) or negatively-associated (Carlisle et al., 2018) with other sagebrush-obligate species of similar conservation concern, particularly at fine spatial scales, highlighting a common criticism that surrogate species approaches inherently result in gaps for the individual requirements of members across the entirety of wildlife communities (Roberge & Angelstam, 2004).
Biome-wide conservation strategies that seek to prioritize "stages" over "actors", while altogether omitting species data in their design provide an alternate approach to surrogate species for conservation design.In these cases, biotic or abiotic properties such as topographical complexity (Beier & Brost, 2010), habitat intactness (Rouget et al., 2006), geophysical diversity (Anderson & Ferree, 2010), or ecosystem properties such as carbon stocks (Carroll & Ray, 2021;Jantz et al., 2014) are assumed to give rise to resiliency in ecosystem function, and are used as primary surrogates in conservation planning.Advantages of such approaches include the ability to plan in data-deficient regions, with assumed robustness to inevitable shifts in species distributions, particularly under anticipated landscape change due to climate change (Anderson & Ferree, 2010;Beier & Brost, 2010).Yet, designs that omit species data may result in critical habitat not being included (Krosby et al., 2015), underscoring a critical need to evaluate the robustness of any design for broader species requirements.
Perhaps the greatest experiment of area-based conservation strategies in lieu of species models is currently unfolding in western North America, where multiple strategies are converging in identifying common geographies to catalyze biodiversity conservation.Across the US portion of sagebrush steppe landscapes a metric for sagebrush ecological integrity, developed from plant functional groups and human disturbance, was developed to guide strategies and investments in sagebrush conservation (Doherty et al., 2022).In the adjacent Great Plains, the Central Grasslands Roadmap used a similar approach to identify core, vulnerable, and converted grasslands for prioritizing conservation actions among many partners (https://www.grasslandsroadmap.org,accessed April 1, 2023).These approaches follow in the steps of private lands conservation frameworks developed by the US Department of Agriculture's Natural Resources Conservation Service (NRCS), which adopted biome-wide strategies to identify intact core rangelands to target investments in conservation (NRCS, 2021a(NRCS, , 2021b)).NRCS has also begun similar efforts to maintain big game migration by prioritizing private lands conservation in Wyoming to ensure connectivity among seasonal habitats (https://www.usda.gov/sites/default/files/documents/mou-vilsack-gordon.pdf).Combined, these strategies provide practitioners with targeting tools across >330 million ha of western North America.
These new approaches ultimately identify "core areas" with the goal of focusing policy makers and managers to proactively conserve common geographies among shrinking biomes, guided by a philosophy of "defend the core" (Maestas, Porter, et al., 2022).By targeting intact landscapes this approach is inherently threatbased, and assumes that managing biome-level threats (e.g., conifer expansion among rangelands) propagates benefits to species of conservation concern (Roberts et al., 2022).In practice, a defend the core strategy has been adopted by practitioners to manage the detrimental effects of invasive annual grass expansion (Allred et al., 2022;Creutzburg et al., 2022) and large wildfires (Maestas, Smith, et al., 2022), yet remains largely untested as an effective surrogate for encompassing the requirements of species underlying biome-specific strategies.
As species-agnostic conservation designs have proliferated, there have been few tests determining the efficacy of this approach with equivocal results, particularly as compared to previously adopted surrogate species methods (Rodrigues & Brooks, 2007).For example, abiotic prioritization methods were found to do a poor job at encompassing species diversity-particularly for species with restricted ranges-across Europe (Araújo et al., 2001), and despite high overlap with species occurrence did not perform much better than expected by chance in South Africa (Bonn & Gaston, 2005).However, a clustering approach based on abiotic variables was found to be a highly efficient approach for capturing diversity, including rare species, in Israel (Trakhtenbrot & Kadmon, 2005), suggesting that the performance of abiotic conservation designs may be non-stationary, scale-dependent, and sensitive to the techniques employed.
Conservation designs based on surrogacy and intactness confront practitioners with seemingly disparate philosophies on how to best allocate limited resources, behooving research to confront them with real world applications in measuring their effectiveness.A focal landscape for science and management embedded within both the northern Great Plains and sagebrush steppe conservation frameworks is a region known as the Northern Sagebrush Steppe (NSS; Jakes et al., 2018; Figure 1).Despite occurring within one of the largest intact rangelands in North America (Scholtz & Twidwell, 2022), native grasslands and shrublands are facing annual losses to cultivation, requiring further efforts to maintain remaining privately-owned grasslands (Gage et al., 2016).Existing models for conservation design in the NSS include those developed for a suite of waterfowl species, both a sagebrush-and grassland-obligate bird species, and pronghorn, a migratory ungulate.We first evaluated multiple models within a surrogate species approach by measuring individual models' ability to confer benefits to non-target species with a derived "strength of surrogacy" metric.Our test also included models of long-distance migration that occurs in this landscapeserving as a proxy for migratory connectivity (Webster et al., 2002).While migratory connectivity models are inherently species-specific, they have been demonstrated to account for ecological functionality and resiliency in the face of changing environmental and landscape conditions (Taylor et al., 2006;Tucker et al., 2018).Within our test of surrogate species effectiveness, we also compared a derived model of "habitat intactness" summarized from intact grassland and sagebrush, representing an approach not reliant on species data.

| Study area
We used previously developed spatial models predictive of selection, occurrence, or abundance for species of conservation concern within the NSS, an area covering 315,876 km 2 at the northern terminus of North America's Great Plains ecosystem.The NSS is at the confluence of sagebrush (Artemisia spp.) steppe, mixed-grass prairie, and the prairie pothole ecosystems.Native herbaceous understory with intermittent shrubland cover, colloquially known as rangelands, is the common thread throughout this landscape.Domestic livestock grazing is the dominant land-use among existing grasslands (Anderson et al., 2018) and over half (55%) of the NSS is currently cultivated (Jakes, 2015), with continued conversion of rangelands to croplands posing the biggest threat to this system.Estimates suggest >200,000 ha of rangelands are converted to croplands annually across the northern Great Plains, as tillage agriculture moves into rangelands previously determined to be marginal for crop production (Gage et al., 2016;Olimb & Lendrum, 2021).Land tenure among remaining rangelands in the NSS vary between the U.S. and Canada.Federal ownership, predominately by the Bureau of Land Management, accounts for 61% of the U.S. land base, while more productive private lands (Robinson, Allred, et al., 2019) make up 30% of the U.S. portion of the NSS.Remaining U.S. lands in the NSS are comprised of state, local municipality, and tribal lands.Conversely, the Canadian portion of the study area is largely private (74%), with a provincial (20%) and federal estate (6%) accounting for remaining lands.Conservation easements-legal agreements which limit cultivation and subdivision-along with habitat enhancement programs designed to incentivize grazing by domestic livestock over cultivation are the primary tools for conserving rangelands in the NSS (Jones, Downey, et al., 2019).

| Surrogate species for conservation
While there are myriad species that rely on the NSS for some portion of their life-history requirements, those most commonly associated with conservation planning in this region include sage-grouse, Sprague's pipit (Anthus spragueii), pronghorn, and a suite of five dabbling waterfowl species (blue-winged teal [Spatula discors], gadwall [Mareca strepera], mallard [Anas platyrhynchos], northern pintail [A.acuta], and northern shoveler [S.clypeata]), which combined demonstrate habitat requirements spanning a gradient from grassland to sagebrush to wetland landscapes.Despite varying habitat requirements, each species or guild is emblematic of the surrogate conservation strategies in this biome.Pronghorn are an iconic big game species that derive revenue for conservation through the sale of hunting opportunities (16 U.S.C. 777), and their extraordinarily large-scale migration in this landscape has been identified in warranting tailored conservation strategies to ensure its persistence (Stemler, 2020).Grassland birds are declining more severely than any other avian guild in North America (Rosenberg et al., 2019).The grasslandsensitive Sprague's pipit has been identified for broad-scale conservation planning given its threatened federal listing in Canada (Environment Canada, 2012), a warranted but precluded designation under the U.S. Endangered Species Act (ESA; USFWS, 2015), and a spatial strategy in place used by practitioners for targeting private lands conservation (Lipsey et al., 2015).Sage-grouse have been the primary surrogate for sagebrush steppe conservation, notably under the U.S. Sage Grouse Initiative that has delivered >4 million ha of private rangeland conservation, with continued investments under USDA Farm Bill (NRCS, 2021a).Lastly, international waterfowl conservation has benefitted from both the North American Wetlands Conservation Act (16 U.S.C. 4401-4412) and Migratory Bird Conservation Fund (16U.S.C. 715-715d), which combined allocated > $200 million USD in 2022 towards wetland and grassland easements.
Importantly, predictive maps of resource selection, occurrence, or relative abundance have been previously developed to prioritize conservation actions across these species (Table 1).Studies used in our analyses were conducted over a similar time frame (2002-2013) with predictions spanning the international NSS boundary, with the exception of breeding densities of sage-grouse for which we applied nearby values to encompass their complete occupied range (Appendix S1).In addition to using waterfowl counts for analyses, we created a measure of fledged young-termed productivity-to account for the potential of maladaptive settling patterns by waterfowl among highly cultivated landscapes (Buderman et al., 2020).Our model of productivity was based on the application of a previously developed matrix population model for mallards within the prairie pothole region (Hoekman et al., 2002) and an estimated relationship between nest survival and grassland cover for dabbling ducks (Reynolds et al., 2001); both models allowed for a simple derived measure of productivity that was specific to our sampling frame and available spatial data (Appendix S1).
All models were resampled, using bilinear interpolation, to a common projection and resolution (193 m) of pronghorn products, given it contained the most derived layers of any species data.

| Analytical approach
We applied methods commonly used in systematic conservation planning (SCP) approaches to test the efficacy of individual species as surrogates for conservation.These approaches are broadly defined by an algorithm applied to a suite of planning units with the aim of maximizing a user-defined objective linked to spatially-referenced attributes, such as a species relative abundance.For our planning units, we generated a hexagonal grid across the sampling frame composed of 438.3 ha hexagons, $1.3 km edges, which corresponded to the mean size of a U.S. conservation easement (http://svc.mt.gov/msl/ mtcadastral, accessed October 17, 2019), resulting in 71,875 total planning units.
An effective surrogate species is assumed to confer conservation to a broader wildlife community through the lens of a surrogate species strategy (Caro, 2010).To test the efficacy for each species to act as a surrogate, we applied maximum utility problems (Di Minin & Moilanen, 2012) to each species model using the "add_max_utility_objective" function in prioritizR (Hanson et al., 2018).Using this approach, a fixed proportion of planning units (e.g., 10%) are identified that maximize the value to underlying conservation features.Simply put, maximum utility problems applied here is analogous to identifying the top xth percent of predicted values within planning units when using single models.We generated maximum utility problems across a range of the proportion of planning units implicated from 5% to 50% of the sampling frame at intervals of 5% for each species layer, as including solutions beyond 50% would implicate largely cultivated habitats.In essence, planning units selected for inclusion would be those prioritized for conservation, with solutions identifying a smaller proportion of landscape being of the highest value to the species.
We also developed two non-single species layers to test as surrogates.Firstly, we created a 'combined migration' layer by calculating the mean value from migratory predictive grids across pronghorn and sage grouse, representing landscapes critical for large-scale movement, which are disproportionately sensitive to habitat loss and fragmentation (Tack et al., 2019;Tucker et al., 2018).Secondly, we sought to test a layer that was wholly independent from species data, by deriving a continuous measure of landscape intactness (hereafter intactness) from a landcover data source that was developed in this region across the international border, using the best available data from the approximate timeframe of when species models were developed (Jakes, 2015).This landcover layer was comprised of 12 cover types and was derived by combining the 2000 Land Cover from the Agriculture and Agri-Food Canada land use time series, and the 2010 Montana Natural Heritage Program's Montana Land Cover Framework (Appendix 1.1 in Jakes, 2015).To calculate intactness, we first considered intact rangeland areas as those classified as either sagebrush, grassland, or wetland; though not open water; and measured the amount of these cover types within a 10.4 km 2 circular moving window-the largest spatial scale of natural cover types to be predictive among species models (Lipsey et al., 2015).Using a measure of intactness based on remotely-sensed data provided a common currency for prioritizing conservation in any geography that could be useful in lieu of information on species distributions and abundance.
The outputs from maximum utility problems were surfaces with planning units that were either selected (in) or omitted (out) for inclusion in final solutions.We extracted the mean state variable value from each model across planning units to test the efficacy of each species model to act as a conservation surrogate.In this test, we assumed each model would act as a surrogate by using its output from maximum utility models, and then measure the coverage afforded to all remaining non-surrogate "passive" models.We fit a series of logistic regression models, using whether or not a planning unit was selected as the response variable (i.e., 1 selected for inclusion and 0 otherwise), with the single independent variable being the mean value for each passive species model within a planning unit.We interpreted coefficient estimates from logistic regression models as a 'strength of surrogacy' metric, with positive, larger estimates reflecting greater agreement between a surrogate and passive species pair, while negative values reflecting discordance between pairs.We scaled all values from species models Þsuch that coefficient estimates from each model represented comparable measures between species models.We only fit models of surrogate and passive pairs among different species; for example, we did not allow multiple seasonal migration models from pronghorn to be a surrogate for one another; and we did not consider combined migration or intact grassland layers to be passive because they were derived measures not currently used for surrogate species conservation.Lastly, we calculated the mean coefficient across strength of surrogacy estimates, from 5% to 50% coverage, for each species model to estimate a mean strength of surrogacy for each model.
Species models used in our approach spanned predictions of various state variables across taxa, though models viewed in total were not disparate as selection, occurrence, and abundance are inherently hierarchically linked population-level parameters (Lele et al., 2013;Royle & Nichols, 2003).Though ultimately using an approach that relied on relative targets of conservation features and a fixed portion of the landscape (maximum utility) obviated the need for a standardized metric among species models, because solutions were based on relative values rather than absolute measures of population parameters.We used integer linear programming to solve conservation planning problems (Beyer et al., 2016) implemented using the prioritizR package ( HansonF I G U R E 2 Strength of surrogacy as measured by coefficient estimates from logistic regression models for each surrogate species considered (panels) conferred to passive species (data points).Models were fit using maximum utility solutions implicating 5%À50% of the study area, to mean values extracted from candidate surrogate species models.Because covariate values were scaled, estimates are comparable between species in direction and magnitude.Positive values indicate surrogate species confer surrogate value, with increasing values providing concomitant increases in surrogacy to passive species, while negative values incongruent distributions between surrogate and passive pairs.et al., 2018) in program R with Gurobi optimization solver and a 0.05 gap to optimality (Gurobi Optimization, Inc, 2016).

| RESULTS
Strength of surrogacy as measured by the coefficient estimates from logistic regression models revealed heterogeneity in the value of species as surrogates to the passive community of species, with non-monotonic changes at varying levels of protection (Figure 2).Overall, each species model provided positive surrogacy for most species across varying levels of simulated landscape conservation, with Sprague's pipit breeding density providing the highest levels.Among species, waterfowl counts, though not waterfowl productivity, was the only model that produced solely negative surrogacy metrics, and consequently was not estimated to have positive strength of surrogacy metrics from other species layers (Figure 2).
The top species models for strength of surrogacy averaged across all passive models were Sprague's pipit, followed by pronghorn autumn and spring migration, sage-grouse breeding density, pronghorn summer habitat, and lastly waterfowl productivity and counts (Figure 3).Derived measures for conservation planning-intactness and combined migration-performed well as surrogates compared to individual species models, with intact grasslands consistently outperforming any other measure.Furthermore, as the proportion of the landscape conserved increased, the strength of surrogacy for intactness increased concomitantly, with approximately three times the mean strength of surrogacy compared to the best performing species-specific model (Figure 3).General patterns of strength of surrogacy held true using either waterfowl productivity or counts when averaged across landscape prioritization values (Figure 4).

| DISCUSSION
Intact grassland landscapes provided the best overall predictor of underlying species requirements, particularly as the land base considered for wildlife conservation increased.Thus, in our comparison of surrogate species for conservation, the simplest measure to acquire and derive outperformed all surrogate models as a singular measure for conservation.While surrogate species or guilds remain a key currency for guiding conservation, they inherently represent idiosyncratic patterns of habitat selection that are useful, yet inevitably imperfect models for biodiversity conservation (Caro, 2010).Thus, it is unsurprising that a metric representing the key underlying requirements at sufficiently large spatial scales prevailed in providing a generalizable currency for conservation design across the diverse life history strategies for multiple species.
F I G U R E 3 Strength of surrogacy estimates averaged across all passive species for each surrogate species using either a measure of waterfowl counts, or a derived measure of productivity for waterfowl that includes heterogeneity in grassland cover.Estimates were averaged among logistic models fit to maximum utility solutions optimizing 5%À50% of the study area for each surrogate species.
Identifying easy to measure components for surrogacy of wildlife occurrence has several advantages.Firstly, metrics to derive and monitor habitat intactness are increasingly available and updated annually (Hansen et al., 2013;Jones et al., 2018), making them ideal candidates for conservation design when robust wildlife data are lacking.Conversely, monitoring wildlife populations is timeconsuming, expensive, and relies on statistical inference from sampled populations, adding challenges to develop products that are dynamic in the face of changing landscapes.In contrast, advances in remote-sensing coupled with cloud computing have made annual development and monitoring of metrics like disturbance, vegetation cover, or productivity possible (Hansen et al., 2013;Jones et al., 2018;Olimb & Lendrum, 2021;Robinson, Jones, et al., 2019).Lastly, remote-sensing metrics provide the most direct links to future scenarios underlying climate and land use projections, which can in turn be used to design efficient conservation portfolios considering impending uncertainty (Ando & Mallory, 2012;Reside et al., 2018).
Rarely are species-agnostic models considered when testing the efficacy among multiple species as surrogates for conservation (Rodrigues & Brooks, 2007), yet given their emergence in practice, it behooves researchers to test the efficacy of such designs for biodiversity conservation.Our approach in deriving a strength of surrogacy metric provides a template for future efforts in evaluating conservation designs.Past examples that have evaluated the effectiveness of surrogate species typically measure overlap among predicted distributions (Rowland et al., 2006) or relative departure from an "optimal" model (Runge et al., 2019).Our surrogacy metric provides a common currency among models with disparate state variables without the assumptions of a singular conservation design to test against.This approach can prove useful in future evaluations and may serve as an objective metric for use in adjacent conservation planning work, such as an objective metric for weighting species models in multispecies algorithms.
Findings here provide a tentative proof-of-concept to emerging biome-wide conservation strategies that are anchored in the NSS.Previous research has demonstrated the non-stationary nature of other environmental surrogates in practice (Ferrier & Watson, 1997;Trakhtenbrot & Kadmon, 2005), so it promising to find a common theme of intactness working effectively among a system implicating both sagebrush (Doherty et al., 2022;NRCS, 2021a) and grassland conservation frameworks (NRCS, 2021b).While wildlife conservation is central to conservation designs in this region, it is important to note that biomewide frameworks, opposed to surrogate-models, may receive wider adoption and support by appealing to more diverse stakeholder groups.Biodiversity is a key, yet singular property in the NSS threatened by stressors including tillage agriculture, invasive annual grasses, and woodland expansion, which similarly threaten other ecosystem services including forage production for domestic livestock (Morford et al., 2022), carbon stocks (Bradley et al., 2006), and water quality (Bhattacharyya et al., 2022).Shifting the narrative from species to large and intact rangelands provides a path for more parity and inclusion among conservation frameworks, ultimately increasing the likelihood of success, while ensuring their connectivity is critical to conserving the migratory movements necessary to sustain wildlife in this landscape.
Providing a conservation network that is robust to both seasonal habitat needs and large-scale movement poses perhaps the greatest challenge to practitioners (Berger, 2019;Joly et al., 2019;Tucker et al., 2018).Terrestrial species, encumbered to make every step of migration, are particularly vulnerable to catastrophic population consequences from severed pathways (Bolger et al., 2008).Even beyond migration and dispersal, daily movements of terrestrial species, exemplified by pronghorn, can be affected by habitat fragmentation and barriers to movement such as roads and fences (Eacker et al., 2023;Jones, Jakes, et al., 2019).Therefore, it may be worthwhile to address movement requirements by focusing on terrestrial species such as ungulates (Gaston & Fuller, 2008).Initiatives designed to embrace the challenge of conserving migratory populations are necessarily broad, cross-jurisdictional, and ambitious (Chester et al., 2012).Our measure of 'combined migration' outperformed surrogate species metrics likely because migration routes generally converged on intact grassland, sagebrush habitats, and topographic variance (Jakes et al., 2020;Newton et al., 2017;Tack et al., 2019).Although assessment of migration has utility across species distributions, it was especially of interest as many of our species are at the northern edge of their range, which required them to move in the face of stochastic weather events (Jakes et al., 2018;Jones et al., 2020;Newton et al., 2017).This underscores the potential role of using habitat models to represent migratory connectivity as a surrogate for conservation as science continues to reveal key animal migration routes (Jakes et al., 2020;Rudd & Kauffman, 2015), though estimating the relationship between migratory behavior and ecosystem measures could help pave the way for developing tools for prioritizing conservation in lieu of such migratory data (Kölzsch et al., 2015;Merkle et al., 2016;Rickbeil et al., 2019).
Implementing one overarching conservation approach to maintain the majority of species' breeding ranges and/or migratory movements is an ambitious goal for biodiversity conservation and likely unattainable in any landscape.In the NSS system, a matrix of private/publicly owned working rangelands that connect intact grasslands across the international boundary is likely more achievable and effective at maintaining ecosystem processes than large gains in protected areas (Tack et al., 2019).An alternative to protected areas that encompass intact grasslands would be to conserve working rangelands and reward private landowners for the ecosystem goods and services (Jones, Downey, et al., 2019).In addition, the allocation of funds to expand this land base via restoration practices (Downey et al., 2013) will be required, as this landscape is home to many declining species for which their currently available habitats fail to produce the templates necessary for recovery.Finer-scale products could help guide practitioners to areas for implementing projects demonstrated to alleviate known threats to species (e.g., MacDonald et al., 2022).Ultimately stakeholders need to be equipped by strategically identifying key landscapes for protection, and 'chipping away' with limited funds and opportunities on incentive-based conservation over long-term investments.
Spatial prioritizations as presented here must always be interpreted in the context of the quantity and quality of data used to train models (Kujala, Lahoz-Monfort, et al., 2018).Our prioritization approach benefited from using only models that were underpinned by peerreviewed science, with unbiased estimates of population parameters.However, models and resulting maps suffered from being inherently static, trained on a snapshot of conditions over several recent years.The NSS is not unique in that large-scale changes from climate and land use interact to shape this system such that wildli-feÀhabitat relationships are likely non-stationary (Runge et al., 2016).For example, native perennial bunchgrasses and bare ground are increasingly replaced by non-native invaders (e.g., Agropyron cristatum), altering the selection and demographic performance of associated species (Pulliam et al., 2020).Conservation designs that grow and adapt through time will be increasingly important to maintain resiliency among threatened biomes.
Though a single measure of intactness proved a superior measure for conservation design in this landscape, its adoption over species-centric approaches in practice faces challenges.A common thread among current conservation delivery mechanisms is an expectation, in some cases by statute, to wholly focus on surrogate-species strategies.State and provincial fish and game agencies, for example, leverage the sale of harvest opportunities for ungulates to fulfill their public trust of game species duties, and in Montana a land protection fund dedicated to sage-grouse identifies core habitats to implement projects (https://sagegrouse.mt.gov).Most notably, the North American Wetlands Conservation Act funded by the sale of federal "duck stamps" has led to >1.8 billion USD investment across >12 million ha to conserve wetland and upland habitats with waterfowl as the primary driver of conservation.While settling patterns of waterfowl alone demonstrated to be an inefficient strategy for a more holistic conservation approach in this landscape, the inclusion of measures of productivity have promise for more representative conservation of upland habitats and effective strategy for waterfowl.Emerging models that take advantage of more recent data and system dynamics offer one path forward for more efficient conservation design for waterfowl (Kemink et al., 2021(Kemink et al., , 2023)).Ultimately, partnerships that leverage funding across sources provide a roadmap for prioritizing future conservation actions.Non-governmental organizations, and biome-level initiatives that are not beholden to species centric objectives may be ideal brokers for conservation in these landscapes, and can work closely with agencies and private landowners to best facilitate species and associated habitats (Jones, Downey, et al., 2019).

F
I G U R E 1 The Northern Sagebrush Steppe (NSS) landscape is embedded at the confluence of the northern sagebrush biome, northern Great Plains, and western prairie pothole regions across northern Montana, and southern Alberta and Saskatchewan.Source: Cropland is depicted from 2020 Copernicus Global Land Cover dataset.

F
I G U R E 4 Strength of surrogacy estimates averaged across both all passive species and percent of landscape prioritized solutions for each surrogate species using either a measure of waterfowl counts (Counts) or a derived measure of productivity (Productivity) for waterfowl.Values demonstrate the relative efficacy of a model to encompass other surrogates.
Attributes and sources of species models used in analyses.
T A B L E 1