Accuracy, accessibility, and institutional capacity shape the utility of habitat models for managing and conserving rare plants on western public lands

Public lands are often managed for multiple uses ranging from energy development to rare plant conservation. Habitat models can help land managers assess and mitigate potential effects of projects on rare plants, but it is unclear how models are currently being used. Our goal was to better understand how staff in the Bureau of Land Management currently use habitat models to inform their decisions, and perceived challenges and benefits associated with that use. We first examined litigation documents to determine whether the agency has been challenged on its use of data for rare plants and found no relevant legal challenges. Second, we analyzed model use in National Environmental Policy Act (NEPA) documents and found no clear citations of habitat models. Finally, we conducted interviews with agency staff who analyze potential effects of proposed actions on rare plants in NEPA documents. The primary challenges interviewees faced in using models related to data organization and access, model quality and accuracy, and institutional capacity. Interviewees believed models could be used more to inform decisions and actions to conserve rare plants and rare plant habitat on public lands and recommended improving staff access to models, creating models for additional species, and addressing staffing limitations.

Rates of biodiversity loss are accelerating globally (Cardinale et al., 2012;Pimm et al., 2014;Tilman et al., 2017).In the United States alone, approximately one-third of native vascular plants are considered vulnerable, imperiled, or extinct (Havens et al., 2014).Habitat degradation and fragmentation are leading causes of biodiversity loss, and plant species are no exception to this trend (Giam et al., 2010;Honnay & Jacquemyn, 2007).Widespread energy development and livestock grazing on federally managed public lands paired with climate change, severe wildfires, and the spread of exotic species can degrade habitat quality and lead to population declines for native plants (Carrell et al., 2022;Copeland et al., 2017;Jones et al., 2015).
Managing rare plant species (see Appendix S1 for definitions) is challenging.Despite the fact that there are more federally listed plant taxa than animal taxa, threatened and endangered plants do not receive the same attention, legal protection, or funding that wildlife receive (Balding & Williams, 2016;Corlett, 2023;Havens et al., 2014).Plant blindness and conservation bias, or a tendency to value charismatic wildlife species over plants, may predispose land managers to focus on protecting wildlife (Balding & Williams, 2016).Taxonomic complexities can also complicate plant conservation efforts (McGlaughlin et al., 2023).Additionally, legal incentives for land managers to emphasize wildlife conservation may be greater because conservation advocacy groups (e.g., Defenders of Wildlife, Center for Biological Diversity, Western Watersheds Project) champion more animal species than plant species.
Rare plant management and conservation efforts are largely governed by the Endangered Species Act (ESA).Federal agencies are required to conserve threatened and endangered species (ESA, 16 U.S.C. § 1531) and consult with the U.S. Fish and Wildlife Service (USFWS) to avoid jeopardizing federally listed species or critical habitat (ESA Section 7, 16 U.S.C § 1536).Section 9 of the ESA outlines a variety of prohibited acts and makes it illegal to "remove and reduce to possession" or "maliciously damage or destroy" endangered plant species on federal lands (ESA Section 9, 16 U.S.C. § 1538(a)).Protections for endangered plants are limited.Prohibitions on removing or harming endangered plant species do not apply to nonfederal lands unless these actions knowingly violate state laws.Additionally, prohibitions on removal and harm only apply to persons collecting plants for possession or sale, and do not extend to persons who displace or destroy endangered plants during land management activities (Easley et al., 2001).
Rare plants that occur on federal lands are subject to additional federal laws and agency guidance that require land managers to consider potential effects to rare plants and avoid negative effects when possible.Before major federal actions occur on public lands, potential environmental effects of the proposed action must be assessed under the National Environmental Policy Act (NEPA, 42 U.S.C. § 4332).Federal agencies also have agency-specific guidance on rare plant management.For example, the Bureau of Land Management (BLM) Special Status Species Management Manual 6840 (BLM, 2008) provides guidance for conserving BLM special status plants, which include federally listed species under the ESA and BLM sensitive species.This guidance directs agency staff to avoid, then minimize, and lastly mitigate potential adverse effects to special status plants (see Appendix S1 for definitions; hereafter rare plants).Many states also have laws protecting rare plants, and state agencies, Native American Tribes, Natural Heritage Programs, nongovernmental organizations, universities, and other stakeholders collectively conduct many activities to conserve rare plants.
Habitat models can assist land managers with evaluations of potential effects posed to rare plant species under NEPA and ESA.Habitat models relate known occurrence records with environmental variables (e.g., climate, soil, geology) to predict conditions suitable for the occurrence of a species (Franklin, 2010;Pearson, 2007;Sofaer et al., 2019).They can map the potential distribution of rare plants based on environmental conditions, shed light on habitat characteristics, and provide geospatial data to help locate previously unknown populations of rare species (Williams et al., 2009).Thus, habitat models can help target and support effective conservation and management of rare plants and their habitats and could also reduce potential land use conflicts by helping developers identify and avoid rare plants at an earlier planning stage (Reese et al., 2019).Habitat models are also often the foundation of conservation planning efforts to identify areas that are important for supporting biodiversity (Franklin et al., 2014).For example, Wu and Smeins (2000) used a multi-scale habitat modeling approach for eight rare plant species in Texas to inform conservation and development planning on the landscape scale, predict potential habitat on the regional scale, and assess habitat in specific study areas at the site scale.
Despite the potential benefits of using habitat models in the context of conservation decisions, the use of habitat models for rare plants may be constrained by a variety of factors.Habitat models may not be available or easily searchable for a wide range of rare plant species.A recent study combined habitat suitability models for 2216 total rare species to examine opportunities for protecting rare species.The authors developed models themselves for most species as they only procured preexisting, vetted, range-wide models for 63 species (Hamilton et al., 2022), illustrating the lack of available habitat models.Additionally, habitat models are not always produced at an appropriate scale and resolution for conservation (Guisan & Thuiller, 2005).For example, models developed for use in broad (e.g., national) conservation efforts may not contain the detailed environmental predictors or fine-grained resolution needed to inform local land management decisions.Since habitat models are developed by many different agencies and organizations (e.g., in the U.S. context: Nat-ureServe, Natural Heritage Programs, universities, U.S. Geological Survey [USGS]), they can be challenging to find and access.A lack of communication between researchers and practitioners as models are being developed can also limit the applicability of resulting products (Addison et al., 2013;Carter et al., 2020;Laurance et al., 2012;Sofaer et al., 2019), potentially limiting the use of models in land management actions and decisions that may affect rare plants.Finally, modelers may rely on readily available environmental predictors as model inputs rather than working with land managers to identify and use the most relevant inputs for rare plant habitat, thus affecting the quality and utility of the resulting model.
As more habitat models are being developed, it is important to understand how federal land managers are using models to help support conservation and management of rare plants.This research arose from challenges expressed by land managers related to a disconnect between available habitat models and habitat models that fit their needs.To build foundational knowledge to address this disconnect, we partnered with land managers to explore a broad set of research questions aimed at understanding how and to what extent habitat models are currently being used to manage and conserve rare plants on public lands.We specifically aimed to (1) determine whether the BLM's use of habitat models has been challenged in recent litigation; (2) understand how habitat models for rare plants are currently being used and cited in NEPA documents; and (3) speak directly with BLM resource specialists who work with rare plants to learn about their current use of habitat models, the challenges and benefits associated with that use, and what actions could make it easier for them to use habitat models for rare plants.

| Study area and project team
Staff from the USGS and the BLM coproduced this study, sharing decision-making and communicating regularly throughout the project.We jointly defined the project scope, planned the approach and interviews, and interpreted and wrote up findings (Beier et al., 2017;Meadow et al., 2015).
The BLM manages the largest extent of federal lands in the United States and a large number of rare plant species (Stein et al., 2008).Our study area varied by data source (Figure 1).Our analysis of litigation records covered Colorado, New Mexico, Utah, and Wyoming because the original study we drew upon focused on the U.S. Department of the Interior (DOI) Upper Colorado Basin.We conducted the analysis of Environmental Assessments (EAs) and Environmental Impact Statements (EISs) solely in Colorado because the frequency and distribution of proposed actions (e.g., fluid minerals development) in the state are similar to patterns exhibited on the national scale (Carter et al., 2023).Based on the locations and jurisdictions of our BLM project team members, our study area for the semi-structured interviews consisted of the states of Arizona, California, Colorado, and New Mexico.
Across the study area for the semi-structured interviews, there are currently 226 federally endangered or threatened vascular plant taxa (ECOS USFWS search engine, as of 7/11/23).These plant taxa include flowering plants, ferns and allies, and conifers and cycads.Over 50% of these taxa occur in California, a large state with unique levels of endemism (Baldwin et al., 2017).These numbers include listed species that occur on non-BLM lands.The BLM manages approximately 15 percent (197,891 km 2 ) of the surface area across the four states (Arizona, California, Colorado, and New Mexico), covering an area roughly the size of South Dakota.Federal agencies, such as the U.S. Forest Service and the BLM, also maintain lists of sensitive species or species of conservation concern that include additional species that require special management considerations to reduce the likelihood of future listing under the ESA (National Forest System Land Management Planning, 36 CFR § 219; BLM, 2008).

| Document analysis
We used document analysis to analyze two data sources: litigation case documents and NEPA documents, namely EAs and EISs, finalized between 2015 and 2019.Document analysis is a qualitative research method for evaluating documents by interpreting and categorizing related content to condense and describe key information pertaining to specific research questions (Bowen, 2009;Elo & Kyngäs, 2008).

| Analyzing litigation records
Legal challenges can provide insight into agency use of data products.In June 2021, we analyzed results from a prior study that assessed legal challenges to the BLM's use of data and science (Foster et al., 2023) to identify challenges to the BLM's use of habitat models for rare plants.
The prior study (Foster et al., 2023) used public litigation records to identify where BLM is commonly challenged on its use of science information (see Appendix S1 for definitions).The authors searched case documents finalized between 2015 and 2019 from federal courts and the DOI Office of Hearings and Appeals.They obtained documents from federal courts via Westlaw, an online legal research database, and from the DOI Office of Hearings and Appeals via the office's web-based search function (https://www.oha.doi.gov:8080/index.html).In both databases, the authors first used the search terms, "BLM" and "data" or "scien*" to identify relevant documents.They then retained documents that involved the BLM and contained a challenge to the BLM's use of science information about a specific resource (n = 48 case documents).Further detail on methods can be found in Foster et al. (2023).
Here, we revisited the 48 case documents analyzed in Foster et al. (2023) to assess potential challenges to the BLM's use of habitat models for rare plants.First, we separated out the challenges from that study that specifically alleged issues regarding data (n = 26 case documents).Then, within this subset of case documents, we searched for challenges that focused on vegetation or plants to determine if any challenges were specific to rare plants.We also performed a keyword search across the 48 originally sampled case documents to search for specific challenges related to habitat models, by searching the terms "spatial," "geospatial," "model," "species distribution," and "SDM."

| Analyzing environmental assessments and environmental impact statements
We collected data from BLM EAs and EISs in Colorado to understand how habitat models are used and cited in rare plant analysis sections (see Appendix S1 for definitions).We analyzed recent, publicly available NEPA documents, including 27 BLM EAs and 9 EISs completed between 2015 and 2019 in Colorado that specifically analyzed potential effects of proposed actions on rare plants (see list of documents analyzed in Appendix S2).
Both EAs and EISs include an Affected Environment section, which provides baseline information about resources of concern, and an Environmental Effects (a.k.a., Environmental Consequences) section, which details the nature of potential effects of the proposed action on the resource of concern.We read both sections in EAs and EISs and recorded in-text data citations related to rare plants.We defined in-text citations as clear citations in a standard format.We defined data citations as citations of datasets or documents containing baseline information on the presence or condition of rare plants.We categorized citations based on whether they clearly referenced habitat models.

| Semi-structured interviews
We complemented our document analysis with interviews to more fully understand how BLM staff use habitat models to inform their decisions, and perceived challenges and benefits associated with that use.We conducted interviews with BLM field office staff from four states (Figure 1).First, BLM members of our project team provided a list of BLM employees who currently write rare plant analysis sections in EAs and EISs.Second, we used job titles to categorize each suggested employee as a botanist, wildlife biologist, or "other" (i.e., natural resource specialist, ecologist, biological science technician).Third, we stratified potential participants by state and title and randomly selected three employees from each of the three categories of job titles and from each of the four states for a total of 12 interviewees.Due to the lack of field office botanists, we only interviewed employees with the "botanist" title in two states.In the other two states, we randomly selected a third staff member from the "wildlife biologist" or "other" category.
We developed 19 interview questions based on our research objectives and iterative discussions with our project team (see Appendix S3 for interview questions).Prior to reaching out to potential interviewees, we piloted our interview questions with BLM staff outside our study area to refine the wording of these questions.We contacted potential interviewees through email to invite them to participate in our study and confirm that they had written rare plant analysis sections in EAs and EISs.We scheduled semi-structured interviews lasting 30-60 min from July to September 2022.
At the start of each interview, we defined the term habitat model as a model that includes known occurrence records and considers environmental variables (such as climate, landform, slope, soil, geology) that indicate an area is suitable for the species in question (Pearson, 2007).We asked all interviewees 11 core interview questions (see Appendix S3).We asked interviewees who indicated that they used habitat models 6 additional questions (17 total) to learn more about how they accessed, used, and cited models.We asked interviewees who indicated that they did not use habitat models 2 additional questions (13 total) to learn about barriers to using habitat models and the types of data they used instead.With interviewees' permission, all interviews were recorded digitally.We transcribed recordings verbatim.Immediately following interviews, we reflected and noted our impressions and perspectives to preserve nuanced elements of the interviews in a process called memowriting (Birks et al., 2008).
Four of the project team members used applied thematic analysis to code interview responses using NVivo version 1.6.1 (QSR International 2023; Braun & Clarke, 2006;Guest et al., 2011).We analyzed responses concurrently with the interview process and kept a record of new codes to determine when thematic saturation was reached (Guest et al., 2006).Our initial interview sample included 12 interviewees, and we also coincidentally reached thematic saturation at 12 interviews, meaning that we were not hearing new ideas or creating new codes.We tracked new ideas and codes that were generated from each interview and identified the point at which responses fully fit within existing codes, which determined our final sample size (n = 12 participants).
We first read the interview responses as a group, discussed thematic ideas, used negotiated agreement to develop codes based on grounded theory, and compiled codes into a codebook (Garrison et al., 2006;Glaser & Strauss, 2017;Tweed & Charmaz, 2011).Second, we reviewed responses from each interview to ensure that codes developed later in the interview and analysis process were applied consistently to earlier interviews, as needed.Throughout our analysis process, we worked as a group to refine our coding until we achieved intercoder agreement, meaning that we reconciled all nuanced inductive coding discrepancies through discussion (Campbell et al., 2013).We generally divided codes based on the interview questions, but also coded overarching ideas mentioned throughout interviews.This was because interviewees sometimes addressed topics, like benefits and challenges, throughout the interview regardless of the question being asked.We presented codes in tables and summarized the number of interviewees who mentioned each code (mentions) to display which ideas were addressed by more participants (Sandelowski, 2001).We provided numbers of mentions in our tables rather than percentages to be clear that our study population represented a random sample of BLM resource managers who write rare plants sections of NEPA documents within the study area rather than a broader group.Part of our quality control process included member checking, which involves contacting interviewees to receive feedback and validation on our interpretation and analysis of their responses (Guest et al., 2011).
The DOI is not a signatory of the Common Rule.However, we complied with USGS Fundamental Science Practices, and we followed ethical guidelines based on the Common Rule (45 CFR § 46.101) and the review process typically required by an Institutional Review Board (IRB, 21 CFR § 56).This process included completing human subjects research training, planning for secure data storage, and informing interviewees of risks associated with their participation.All interviewees provided their informed consent and were able to remove themselves from the study at any time.

| Legal challenges related to the BLM's use of data
While we identified legal challenges to the BLM's use of data for vegetation and invasive plants (Foster et al., 2023), we found no legal challenges specific to the BLM's use of data for rare plants.Additionally, our keyword search for terms related to habitat models did not result in any matches.Past litigation around the use of habitat models could discourage land managers from using such models in their NEPA analyses, so this finding provided useful context going into our analysis of rare plant data citations in recent EAs and EISs.

| Data citations in environmental assessments and environmental impact statements
In our sample of 27 Colorado EAs, we found no citations that clearly referenced habitat models for rare plants.We found that 41% of rare plant analysis sections included at least one clear citation to data of any type.Examples of references to data included USFWS Species Lists, BLM Sensitive Species Lists, BLM Land Health Assessments, BLM Resource Management Plans, and contractor survey reports.Though these document types may not immediately appear to be data sources, for the purposes of this study, we defined data citations as citations of datasets or documents containing baseline information on the presence or condition of rare plants.It is likely that some of these citations contain a geospatial component, but that was not evident from looking at the citation or title of the referenced product alone.
The nine Colorado EISs we analyzed contained more detailed information in rare plant analysis sections, more citations, and a wider variety of data types.We found geospatial data citations, but we did not find any explicit references to rare plant habitat models in EISs.All rare plant analysis sections in our sample of EISs included at least one citation to data, but most (91%) were not clearly geospatial.Examples of references to data included Federal Register listing decisions, Species Recovery Plans, Species Status Assessments, 5-year Status Reviews, and annual monitoring reports from the BLM Colorado State Office.Similar to EAs, some of these information sources likely contain geospatial data, but that was only clear in a small selection of references: 9% of data citations in EISs clearly referenced geospatial data.Those references included element occurrence polygons, a USFWS geospatial data product delineating Critical Habitat for a federally listed species, and a proposed rule to designate Critical Habitat for three federally listed species.

| Semi-structured interviews
We conducted 12 interviews that averaged 32 min in length, for a total of 6 h and 23 min of interview data.Below, we first present information about the study population based on our sampling approach and demographic questions.We then report codes that relate to the use of habitat models for rare plants by BLM field office staff who write rare plant analyses for EAs and EISs.We include quotes that exemplify these codes and support our interpretation of interview responses.We present thematic results by four main topics: (1) current use of habitat models, (2) benefits of using habitat models, (3) challenges of using habitat models, and (4) actions that could make it easier to use habitat models in the future (Tables 2-5).

| Study population
We asked three multipart questions to understand the background and experience of interviewees.Interviewees had worked with the BLM for 2 to over 20 years (Table 1).It is common for staff in the BLM to change positions frequently, so we also asked how long interviewees had been in their current position.Half of the interviewees had been in their current position for ≤5 years (Table 1).We spoke with two botanists, five wildlife biologists, and five staff members with other titles (i.e., ecologist, natural resource specialist, biological science technician).Interviewees typically self-identified as natural resource specialists or ecologists.However, only one study participant said that they worked primarily with plants.
Most interviewees (9 of 12) were involved in the NEPA process as authors responsible for developing content for rare plant analysis sections.Interviewees also identified as project leads (5 of 12), reviewers (5 of 12), interdisciplinary team members (4 of 12), and specialists (3 of 12; Table 1).Interviewees were evenly split in terms of whether they were responsible for all the rare plant work in their field office or they shared the responsibility with other employees.Interviewees described working with rare plants by collecting data (7 of 12), analyzing potential effects for NEPA (7 of 12), coordinating and consulting with USFWS (4 of 12), and writing reports (2 of 12; Table 1).

| Current use of habitat models
Two-thirds of interviewees had used habitat models in their work to analyze potential effects of proposed actions on rare plants.Those who did not use models (n = 4) used species occurrence records (3 of 4), field visits (2 of 4), professional judgment (1 of 4), and other available geospatial environmental data (1 of 4; Table 2; Appendix S4).The main reason interviewees did not use habitat models was related to the lack of awareness of available models, as illustrated by this quote: "If there are any [models] out there, I don't know about them for our field office.I don't think they've been developed for the plants we have."-P02 Even though we did not find any clear references to habitat models across our document analysis of Colorado EAs and EISs, interview responses confirmed that staff are using models to inform the early stages of their work.For those who said they used models (n = 8), we asked how they accessed and used models and what challenges and benefits they faced in using models.Most interviewees (6 of 8) accessed models on internal BLM servers, and interviewees also accessed models through public websites (3 of 8) or through colleagues and partners (2 of 8; Table 2; All 6

Shares responsibilities 6
Note: A summary of response categories from semi-structured interviews with BLM resource specialists.We present a shortened version of the interview question (column 1), followed by the response category (column 2), and the number of interviewees who mentioned each category (column 3).Interview questions are immediately followed by the number of interviewees who responded to the question.Abbreviations: BLM, Bureau of Land Management; USFWS, U.S. Fish and Wildlife Service; NEPA, National Environmental Policy Act.
Appendix S4).Most interviewees (7 of 8) used models to establish baseline understanding of the presence and condition of rare plant populations and habitat, and 75% (6 of 8) of interviewees also used them to inform where to survey for rare plants prior to beginning development.Interviewees also used models to analyze potential effects of proposed actions on rare plants (3 of 8; e.g., quantify habitat loss) and inform mitigation measures (2 of 8; e.g., inform where to avoid or transplant rare plants).This quote illustrates an example of models being used to inform where to survey for rare plants and to establish baseline understanding of the presence or condition of rare plants and their habitat: "My preferred method of using models in the NEPA process is [using them] in those planning processes rather than just the analysis because we want to make sure we're identifying survey areas that are appropriate and not missing out on areas that may have potential habitat."-P06 Half of the interviewees who said they used habitat models also said it was typical to cite models.Of those cited models (n = 4), 75% said they included citations to increase the quality, clarity, and objectivity of their analyses.Interviewees said they did not cite models

Field visits 2
Other geospatial environmental data 1

Professional judgment 1
Note: Emergent codes that we interpreted from semi-structured interviews with BLM resource specialists summarizing how staff members work with rare plants and use habitat models to inform their work.We present the shortened interview question (column 1), the category or code that we interpreted from responses (column 2), and the number of interviewees who mentioned each category or code (column 3).Interview questions are immediately followed by the number of interviewees who responded to the question.See Appendix S4 for additional details.Abbreviations: BLM, Bureau of Land Management; EA, Environmental Assessment; EIS, Environmental Impact Statement; NEPA, National Environmental Policy Act.
when there were issues with model accuracy or quality, models were used before the NEPA process officially began, models were too complex, or models were unpublished (Table 2; Appendix S4).The quotes below are examples of not citing models due to issues with model quality or accuracy and due to models being used early in project planning: "[The model] ends up not being formally cited […] because none of the formal models do I regard as accurate enough or good enough to actually define the habitat of the species."-P10 "Basically, we'll mention the model and refer to it in the prescreening process and surveying process."-P07

| Benefits of using habitat models
We asked if habitat models were useful for resource specialists, whether rare plants and their habitats benefited from model use, and whether the BLM as a whole benefited from the use of habitat models.We also captured responses throughout interviews that related to general benefits, regardless of the interview question.
When we heard about general benefits throughout interviews (see Table 3; Appendix S5), all interviewees said that habitat models improved their understanding of species habitat or distribution.Two-thirds (8 of 12) of interviewees indicated that models helped them assess potential effects of proposed actions on rare plants and their habitat, and models helped them target restoration or conservation efforts.Interviewees often (7 of 12) said models increased objectivity in decision-making and reduced guesswork (Table 3; Appendix S5).This quote is an example of models being used to assess potential effects and increase objectivity: "And [models] would give me a little bit more confidence when I say this project has the potential to impact this species when I don't have the data up front."-P04 When we asked if habitat models were useful for resource specialists who used models (n = 8), most T A B L E 3 Perceived benefits of using habitat models.

Shortened interview question or general topic Assigned response category or code Mentions
If you use habitat models, have habitat models been useful for you?(n = 8) Improve understanding of species habitat or distribution 5 Increase defensibility of decisions 5

Fill information gaps 3
Reduce workload 2 Help prioritize funding 1 (Continues) interviewees (7 of 8) said models were useful for analyzing and describing potential effects to rare plants in EAs and EISs.One interviewee said rare plant habitat models were not useful because there was a lack of institutional support for using models.In addition to the general benefits described above, interviewees said models were useful because they provide information in the early project stages about where to focus survey efforts (2 of 8) and how to avoid rare plant habitat (1 of 8; Table 3; Appendix S5).This quote is an example of models improving understanding of species habitat or distribution and informing where to survey for rare plants: "I find [models] super helpful because they let me know where I need to focus survey efforts on the ground.
[…] Before I can do any kind of analysis on any kind of project, I need to know if the species occurs there."-P12 When we asked whether rare plants and their habitats benefited from the use of habitat models, interviewees said rare plants benefited because models improve understanding of species habitat or distribution (11 of 12), help assess potential effects (5 of 12), and inform conservation efforts (4 of 12).They also said rare plants benefitted from habitat models because models can help resource specialists avoid or minimize potential effects to rare plants and their habitat (3 of 12; Table 3; Appendix S5).This quote displays an example of models improving understanding of species habitat or distribution and informing conservation efforts: "I feel like annual plants hugely benefit from modeling because [these plants] don't always show up every year.Sometimes people will argue with you if you don't find that rare plant out there… 'why are we protecting this habitat if you didn't find the plant?'Well, habitat models show that it did or should occur here."-P02 Finally, when we asked whether the broader agency benefited from the use of habitat models, interviewees said that models can benefit decision-making in the BLM across agency levels (Table 3; Appendix S5).Half (6 of 12) of the interviewees said habitat models support evidence-based decision-making, and 5 of 12 respondents said models increase the defensibility of decisions.Interviewees also mentioned that habitat models can help fill general information gaps (3 of 12; e.g., improve knowledge about inaccessible places), reduce workload (2 of 12), and prioritize funding (1 of 12).This quote displays an example of how habitat models can reduce workload: "I don't know that a lot of BLM offices have a botanist on their payroll anymore.I think we're all wearing multiple hats, so having these tools at hand can be super help-ful…" -P01 Increase objectivity and reduce guesswork 7 Increase defensibility of decisions 5 Inform where to survey for rare plants 5 Support BLM information exchange 5 Help avoid or minimize potential effects 4 Support decision-making 4 Inform landscape-scale management 3 Reduce workload 2 Help inform ESA listing decisions 2 Note: Emergent codes that we interpreted from semi-structured interviews with BLM resource specialists summarizing perceived benefits of using habitat models for rare plants in land management activities.We present the shortened interview question or general topic (column 1), the category or code that we interpreted from responses (column 2), and the number of interviewees who mentioned each category or code (column 3).Interview questions are immediately followed by the number of interviewees who responded to the question.See Appendix S5 for additional details.Abbreviations: BLM, Bureau of Land Management; ESA, Endangered Species Act.

| Challenges of using habitat models
Interviewees noted a variety of challenges that affected their use of habitat models (Table 4; Appendix S6).Over half (7 of 12) of the interviewees mentioned issues related to data organization or access, as illustrated by this quote: "If there are rare plant habitat models, I'm just not sure where that information would be.It's not all centrally accessible.And there's so much information […] that it's kind of overwhelming."-P01 Interviewees also regularly (5 of 12) described issues with model quality or accuracy.For example, one respondent mentioned that they used a habitat suitability model to identify areas to survey for a species but when they surveyed the area; they realized it was very unlikely that the species could actually occur there.Another respondent explained that rare plants have very nuanced habitat conditions and affinities for specific soil types, geology, or hydrology, and models do not adequately capture those nuances.Issues with model quality or accuracy were often described in relation to issues with model scale, which onethird (4 of 12) of respondents mentioned, since resource specialists often have to work with data products that are too coarse for local, fine-scale projects.Two interviewees linked issues related to model quality or accuracy with a lack of trust in models.A quarter (3 of 12) of respondents described general issues related to trust and indicated that they needed convincing tools, but existing models were unreliable.This quote describes issues with model quality or accuracy: "There are some limited [models], but I'd also say a lot of those models are so vague that they don't do a very good job in terms of defining the very specific niche of some species."-P10 Interviewees also described a lack of institutional capacity (5 of 12), specifically related to staffing limitations that affect the application of existing models or development of new models (Table 4; Appendix S6).They also talked about a lack of models or a lack of awareness of available models (4 of 12), a lack of institutional support for using models in decision-making (2 of 12), and issues understanding or explaining models (2 of 12; Table 4; Appendix S6).This quote notes issues regarding a lack of institutional capacity: "I guess the only other issue would be the lack of botanists in our office.[…] It would be nice to have someone who could fully put this model to use.I feel I'm going to fall short with my workload…" -P07 We also asked interviewees who indicated that they used habitat models whether they experienced challenges with accessing models.Some interviewees spoke about general challenges associated with using habitat models rather than specifically addressing access issues.Half (4 of 8) of the interviewees who used habitat models said they did not experience challenges associated with accessing models, but 25% (2 of 8) of respondents described challenges related to access (Table 4; Appendix S6), as displayed by this quote: "Generally, the [model] I have is at my fingertips at all times, but the other models that our partners have developed, it's been a T A B L E 4 Perceived challenges of using habitat models.

Assigned response code Mentions
General

No challenges 4
Issues finding or accessing models 2 Note: Emergent codes that we interpreted from semi-structured interviews with BLM resource specialists about perceived challenges of using habitat models for rare plants in land management activities.We present the shortened interview question or general topic (column 1), the code that we interpreted from responses (column 2), and the number of interviewees who mentioned each code (column 3).Interview questions are immediately followed by the number of interviewees who responded to the question.See Appendix S6 for additional details.Abbreviation: BLM, Bureau of Land Management.
challenge to get them.When they have provided them, access is great, but it's been somewhat challenging since it seems like they're constantly tweaking them or they're not quite ready…" -P05 3.3.5 | What actions could make it easier to use habitat models One interviewee told us that models were already easy to use, but most interviewees voiced ideas for improving their ability to use rare plant habitat models in their decisions (Table 5; Appendix S7).Parallel to issues interviewees faced surrounding data organization and access, half (6 of 12) of the interviewees suggested improving access to models and aggregating data in a central repository.Parallel to issues surrounding model availability, 25% (3 of 12) of interviewees suggested creating models for additional species.This quote suggests improving model access and the quality and accuracy of models: "If they're not produced by the bureau, definitely making it easy to access them is essential.If they're difficult to use, people will give up on them.And then I would say ditto to the scale stuff.The more accurate, the more comfortable we feel with them, the more likely you'll get buy-in from other specialists to use them with a high degree of confidence."-P05 Other suggestions included building partnerships with modelers (2 of 12), improving model quality and accuracy (2 of 12), increasing the transparency of model inputs and metadata (2 of 12), adding more staff to reduce individual workloads (2 of 12), obtaining funding to allocate to model development (2 of 12), and developing guidance for using models to provide regulatory incentives for resource specialists to apply habitat models (1 of 12; Table 5; Appendix S6).The first quote suggests creating models for additional species of management concern and the second quote suggests addressing staffing limitations, building partnerships, and allocating funding for model development: "To have experts in the field create [models], at least for those species that are most likely to be affected by land-use planning."-P09 "It all kind of relates back to capacity and I don't think the solution is [learning] how to become a modeler because that just doesn't seem practical.But I think just connecting with people who have [modeling] as their specialty and allocating funding to develop [models]."-P03

| DISCUSSION
We paired document analysis with semi-structured interviews to better understand how U.S. land management agency employees use habitat models to inform their analysis of potential effects to rare plants associated with activities on public lands.We also sought to understand the challenges and benefits agency staff perceive with using rare plant habitat models.In our document analysis, we found that BLM is not being litigated on their use of habitat models, but that they are also not citing habitat models in their NEPA analyses.Contradictory to our finding that BLM staff are not citing habitat models based T A B L E 5 Perceived actions that could make it easier to use models.Note: Emergent codes that we interpreted from semi-structured interviews with Bureau of Land Management (BLM) resource specialists about perceived actions that could make it easier to use habitat models for rare plants in land management activities.We present the shortened interview question (column 1), the code that we interpreted from responses (column 2), and the number of interviewees who mentioned each code (column 3).Interview questions are immediately followed by the number of interviewees who responded to the question.See Appendix S7 for additional details.

Shortened interview question
on our analysis of citations in NEPA documents, subsequent interviews revealed that half of the resource specialists who indicated that they used habitat models said it was typical to cite them.This mismatch may be due to resource specialists citing models in internal record documents supporting the NEPA analysis process, but not in the NEPA documents themselves.Alternatively, this could reflect a difference in practices between the NEPA documents from 2015 to 2019 that we analyzed and when we interviewed BLM staff in summer 2022.
Through interviews, we found that respondents perceived models as beneficial for increasing understanding of species habitat or distribution, assessing potential effects to rare plants, and informing restoration or conservation actions targeting rare plants.Top challenges that interviewees faced in using rare plant habitat models related to data organization and access, model quality or accuracy, and institutional capacity.Interviewees suggested improving staff access to models, creating models for additional rare plant species, and addressing the lack of staff capacity to use and interpret models.
The challenges discussed by the interviewees span the processes of model creation, delivery, and use, indicating the need for an integrative approach to improve the viability of models as tools for assessing potential effects to rare plants.Certain challenges described by interviewees may be addressed specifically by model developers, such as developing additional models, improving the quality and accuracy of models, increasing the transparency of model inputs and metadata, and addressing potential scale mismatches (Gogol-Prokurat, 2011).However, many challenges surrounding the use of models in management and conservation contexts require collaborative solutions.A recent study suggested that collaborative development of hydrological models gives rise to trust in models since stakeholders can help shape the product and iteratively provide information to improve the model, and modelers can support subsequent use (Ulibarri, 2018).Another study suggested broadening the use of models by incorporating structured decision-making and using a participatory approach to promote acceptance and build trust (Addison et al., 2013).A different study brought together researchers and managers to understand why model-based tools were not being used by water managers and found that improving communication should be paired with actively improving understanding (e.g., researchers' understanding of management processes) across groups (Borowski & Hare, 2007).These studies collectively support the need for participatory processes and decision frameworks that allow for the generation and iterative adaptation (Runge, 2011) of products that are useful and useable for their intended users.
Coproduction, a highly participatory process, can provide a framework to create spatial products that are useful for end users (Beier et al., 2017).Coproduction is a model of collaboration where stakeholders engage in regular communication with researchers and share decision-making power (Bamzai-Dodson et al., 2021).
Coproduction between researchers and practitioners can increase trust and buy-in from the management community, which could incentivize greater use of habitat models in the management context (Kruk et al., 2017;Reed et al., 2018).Coproduction has the potential to be particularly beneficial in the creation of ensemble habitat models, where site-specific knowledge is essential for creating accurate, high-quality models.Ensemble habitat models combine multiple models, using alternative algorithms or data inputs for example, limiting the weaknesses of any one model (Araújo & New, 2007).These ensembles and their formats can be tailored to meet the needs of end users, such as producing binary or categorized habitat suitability maps (Sofaer et al., 2019), to guide different management actions.Identifying environmental parameters to best model habitat requires expertise and review from on-the-ground specialists, so using an iterative coproduced approach can improve and build trust in the resulting product.Resources exist to support researchers and resource managers in their efforts to engage in coproduction to develop actionable science (Selby et al., 2024).
The barriers to using models that interviewees described do not all relate to the modeling process.Respondents identified issues surrounding data organization and access, which could be addressed by creating central data repositories and providing instructional workflows to support staff in finding and using available habitat models (Ladouceur et al., 2022).Resource specialists are responsible for managing multiple resources at a breakneck pace, sometimes including resources they were not specifically trained to manage.Thus, it is important to provide the information and tools, like habitat models, they may need in a central location to streamline effects analysis (see Appendix S1 for definitions; Appendix S7).Institutional barriers, like staffing limitations, could be addressed by hiring more botanists, ensuring adequate funding to support the needs of resource specialists, and providing needed time and training for staff to comprehensively assess potential effects to rare plants (Kramer et al., 2013;Pearman & Cravens, 2022).
Participants also addressed agency-level challenges in their responses.One interviewee suggested updating agency guidance for managing rare plants, which could include implementing universal buffers to increase the distance between planned development activities and rare plant habitat and help to redirect development away from rare species early in the planning process.The BLM Special Status Species Management Manual 6840 emphasizes the importance of avoidance (BLM, 2008).Strategies to avoid potential effects to rare plants and their habitat may be better supported by policy or guidance that directs resource specialists to use habitat models in project planning when they are available.In 2022, the BLM released the Strategic Plan for Special Status Species Conservation and Recovery which commits the agency to developing and using sound science, including habitat models, to inform conservation and recovery efforts (BLM, 2022).This plan could be instrumental for increasing the use of habitat models in land management activities.
Our methodology has limitations.For our analysis of litigation records, our sample only included publicly available case documents (Foster et al., 2023), and the timeframe was restricted to documents completed between 2015 and 2019, several years prior to our semistructured interviews conducted in 2022.Similarly, our sample of EAs and EISs was restricted to Colorado and the same timeframe.Sampling more recent documents could have resulted in more instances of citations of habitat models.Additionally, we reached thematic saturation with 12 interviewees (Guest et al., 2006), but our study population only included two botanists due to staffing limitations at field offices.This accurately reflects staffing dynamics at field offices but limits our ability to represent botanist perceptions.While we sampled a representative population in our study area, our results do not comprehensively represent perspectives across the BLM due to the fact we did not include participants representing the full geographic footprint of the agency.This means that our results may not speak to broader patterns related to the use of habitat models for rare plants across the BLM in general or across additional land management agencies.Given this, future studies could include participants from a wider geographic area or across multiple land management agencies.Targeting agency employees across different administrative levels (state, regional, national) and expanding to include other organizations (e.g., State employees, Natural Heritage Program employees) could offer additional insight and capture other approaches for fostering the use of habitat models.While our results are not a comprehensive evaluation of the use of habitat models across land management agencies in the United States, they may be valuable for informing future research on this topic.
Our findings suggest that resource specialists find rare plant habitat models to be useful in the context of public land management, but that their usefulness is influenced by their accuracy, accessibility, and other factors like institutional capacity.Rare plant species are vulnerable to threats associated with land use change and climate change (Still et al., 2015), and species could face extinction without proper management on public lands.Improving the utility of habitat models and initiating collaborations that increase trust in habitat models can support proactive actions to protect rare plant species.Conservation actions are more likely to succeed if those responsible for conserving biodiversity have the necessary informational tools and support to inform their actions.Increasing connections between model developers and end users could improve the utility of habitat models to support public land management by providing opportunities for model users to help determine model inputs and collect data to help validate models, for example.The approaches described above for addressing challenges related to using habitat models could also be applied to different resources, such as invasive plant species or rare wildlife species, since models can inform management actions for many resources.
Working directly with end users to understand how habitat models are used provides important insight into how products intended to inform management actions could be better developed.Centering the approach on understanding the needs of data users shifts the narrative away from simply creating new data products toward creating the right data products.Understanding the nuances of challenges described by data users can facilitate future collaborations between resource managers and the modeling community to develop habitat models that can better support conservation and management of rare plants on public lands.
AUTHOR CONTRIBUTIONS E.M. Samuel managed the project, implemented data collection and analysis, and led the writing.S.K. Carter conceived the idea, designed the study, managed the project, and reviewed drafts of the manuscript.J.K. ACKNOWLEDGMENTS E Roberts helped conceptualize our project approach.Bureau of Land Management (BLM) staff, including C Lund, P Krening, J Davis, A Roe, P Alexander, C Domschke, and T Haby, helped develop this project.We thank the interviewees for their time and input, and the FORT Social and Economic Analysis Branch for insight on interview and analysis methods.T Rutherford, A Bamzai-Dodson, and L Romin reviewed earlier manuscript drafts.We thank Colorado State University and K

F
I G U R E 1 Map of public lands managed by the Bureau of Land Management (BLM) in the western United States, paired with sampling areas for the different data sources.We sampled public litigation records across the U.S. Department of the Interior Upper Colorado Basin Region (Colorado, New Mexico, Utah, and Wyoming), we sampled Environmental Assessments (EAs) and Environmental Impact Statements (EISs) in the state of Colorado, and we selected interviewees across four states associated with our BLM project team members (Arizona, California, Colorado, and New Mexico).
T A B L E 2 Rare plant work and use of habitat models.Shortened interview question Assigned response category or code MentionsHave you ever used rare plant habitat models in your NEPA analyses?(n = 12) habitat models, how do you use them?(n = 8) Establish baseline presence or condition of rare plants habitat models, is it typical to cite models in the EA or EIS itself?(n = 8) habitat models but do not cite them, what is your reasoning for not citing models?(n = 4) not use habitat models, what do you use instead to inform your rare plant analyses?(n = 4) habitat models, how have they been useful for you?(n = 8) Improve understanding of species habitat or distribution 4 Increase objectivity and reduce guesswork 4 Inform where to survey for rare plants 2 Help assess potential effects of proposed actions on rare plants and their habitat 1 Help avoid rare plant habitat when planning new projects 1 Regardless of whether you use habitat models, do you think rare plant populations or habitats have benefited from model use?(n = 12) you use habitat models, how do you think rare plant populations or habitats have benefited from model use?(n = 12) you use habitat models, how do you think models benefit decision-making in the BLM across agency levels?(n = 12) Support evidence-based decision-making 6 T A B L E 3 (Continued) Shortened interview question or general topic Assigned response category or code Mentions General benefits of habitat models mentioned throughout interviews (n = 12) Improve understanding of species habitat or distribution 12 Help assess potential effects of proposed actions on rare plants and their habitat 8 Inform restoration or conservation efforts 8 Meineke contributed critically to project management and led qualitative methods development with L.E.McCall.J.K. Meineke, L.E.McCall, L.B. Selby, and A.C. Foster contributed significantly to data collection and analysis and reviewed manuscript drafts.Z.M. Davidson, C.A. Dawson, and C.S. Jarnevich provided expert input to inform project development and reviewed manuscript drafts.