Matching existing and future native plant materials to disturbance‐driven restoration needs

Assessing the appropriateness of existing native plant materials can both determine which seed source to utilize for restoration projects, and identify locations for which new seed sources need to be developed. Here, we demonstrate an approach to meet these needs. This method identifies areas of high restoration need based on disturbance patterns, assesses the regional suitability of existing native plant materials based on climate similarity, and highlights geographic (and climatic) gaps where existing materials are likely unsuitable and where plant material development projects can be prioritized. We examined 12 high priority restoration species across the Colorado Plateau, a 38‐million‐ha region of the Intermountain West, United States to test our methodological pipeline. Fifty‐four percent of the Colorado Plateau is disturbed by livestock grazing, wildfires that have burned in the past 20 years, or energy production from oil and gas wells, natural gas pipelines, and coal mines. Of the 28 commercially available plant materials for six of the focal species, only 3 have climate similarity that encompass more than 50% of the species modeled habitat on the Colorado Plateau. Across all commercial materials, most species (10 of 12) do not have any suitable plant material for 70% or more of their geographic range on the Colorado Plateau. Of those areas identified as not having any suitable plant materials, 47–56% are also disturbed. Our method provides usable, flexible protocols and spatially referenced data sources for optimizing the planning of new native plant materials in any region where restoration is needed and spatial data are available.


Introduction
Effectively rehabilitating and restoring land is a challenge that requires prompt solutions to achieve sustainability goals related to climate change and biodiversity conservation, among others (UN 2021).Yet, successful restoration of broad areas requires an abundance of native plant materials (i.e.any wild-collected or propagated germplasms including seeds, stem cuttings, pollen, seedlings, etc.; Basey et al. 2015) that are genetically and ecologically appropriate for a restoration site.Supply issues and challenges in identifying appropriate materials have been among the major issues facing practitioners today (Camhi et al. 2019;McCormick et al. 2021).Perhaps nowhere is this more evident than in drylands where systems have warmed 20-40% more than humid lands (Huang et al. 2017) and land use has made natural regeneration of plant species improbable (Schlaepfer et al. 2015;Stevens-Rumann et al. 2018).Unsurprisingly, efforts to Author contributions: DEW, SS shared co-first author; all authors conceived the ideas and designed the methodology; SS, DEW collected and analyzed the data; DEW, SS led the writing of the manuscript; all authors contributed critically to the drafts and gave final approval for publication.restore native plant species in drylands around the globe have struggled (Shackelford et al. 2021).
Native plant materials development strategies (e.g.natural track or multi-source pooling; Aubry et al. 2005;Broadhurst et al. 2008;Bucharova et al. 2019) have received more attention in recent decades as resource managers and conservationists seek to create resilient, productive, and diverse communities that sustain ecological function while providing ecosystem services (Tischew et al. 2011;Plant Conservation Alliance 2015).Although awareness and demand have rapidly increased, supplies of locally adapted, genetically appropriate plant materials have been unable to meet needs (Jalonen et al. 2017;National Academies 2023).To increase supplies, land managers can increase wild-collected seed agronomically (Smith 2017;McCormick et al. 2021), though many roadblocks remain including slow propagation rates, species-level knowledge gaps, germplasm collection and storage challenges, as well as potentially prohibitive costs (e.g.Le on-Lobos et al. 2020).
Successful utilization of wild-collected seed sources requires managers and practitioners to select genetically appropriate seed not only to supply the raw material for developing native plant materials, but also to identify the most appropriate materials for current and future restoration projects.Ensuring that new plant materials are suitable for restoration needs and complement existing materials is a key step to increase chances of success, especially in seed-based restoration (e.g.Padilla et al. 2009;Shackelford et al. 2021).Yet, selecting species and seed sources for restoration projects remains a limiting step in restoration globally because it typically requires knowledge of species regeneration requirements, abiotic and biotic interactions, short-and long-term successional trajectories, and ecosystem requirements for species persistence (Oldfield et al. 2019;Shaw et al. 2020;McCormick et al. 2021).Since this information is frequently limited, restoration practitioners are left to choose from commercially available plant materials while relying on trial-and-error approaches (Betz et al. 1996;Howe & Martínez-Garza 2014;Lamers et al. 2015) and local knowledge to make species and seed source selections, which has its own benefits and drawbacks (e.g.Su arez et al. 2012;Meli et al. 2014;Reyes-García et al. 2019;Fremout et al. 2021).
Existing methodologies that seek to guide native plant materials development are rare but those that exist utilize everything from simple occurrence or categorical data (e.g.NatiVeg; www.quailcount.org/NatiVeg) to software that predicts seedling survival based on user-inputs of desired functional traits (e.g. the Restoration Plant Species Selection Platform; Wang et al. 2021).Here, we developed a methodology to identify (1) disturbed areas in need of restoration, (2) existing native plant materials likely suitable for a given restoration site based on climate similarity, and, perhaps more importantly, (3) geographic gaps where existing materials are unsuitable and native plant material development projects should prioritize.Our study relies on the expectation that plant materials sourced from sites that are climatically similar to the restoration site may have the best restoration outcomes when genetic data are unavailable, regardless of the plant material development strategy (e.g.Gustafson et al. 2005;Baughman et al. 2019;Pedrini & Dixon 2020).The assumption guiding plant materials transfer based on minimizing climate distance (and associated ecological distance) is implicit in seed transfer zone concepts, including provisional (e.g.not species-specific; Bower et al. 2014) and empirical approaches (e.g.species-specific; St. Clair et al. 2013).We examine 12 high priority restoration species across a 38-million-ha (383 k km 2 ) region of the Intermountain West, United States to test our methodological pipeline, describe applications and user-defined flexibility included in the pipeline, and explore limitations and potential next steps for improving this strategic process for optimizing the identification of new native plant materials.By providing methodology to evaluate the suitability of existing native plant materials (e.g.Baker et al. 2020) and identify gaps where novel seed sources are likely suitable for native plant materials development (Breed et al. 2018;Bucharova et al. 2019), we provide a critical tool for managers and practitioners to address restoration challenges under current and future climates.

Study System and Target Species
We included 12 graminoid and forb species distributed throughout the Intermountain West of the United States that are commonly used in seed-based restoration projects (Achnatherum hymenoides; Astragalus lonchocarpus; Bouteloua gracilis; Cleome lutea; Cleome serrulata; Elymus elymoides; Heliomeris multiflora; Heterotheca villosa; Machaeranthera canescens; Pleuraphis jamesii; Sporobolus cryptandrus; Sphaeralcea parvifolia).Dominated by arid and semiarid drylands, much of the public land in this region directly supports rangeland production on lands managed by federal agencies including the Bureau of Land Management and U.S. Forest Service (Gentner & Tanaka 2002).These drylands are similar to other systems that together cover ca.45% of the terrestrial surface (Pr av alie 2016) and include regions of the globe where disturbance is common and successful restoration is limited (Shackelford et al. 2021).Of the 12 species examined in this study, 6 had commercially available material included in demonstrating our methodological pipeline described herein (Table 1; Fig. 1).Our pipeline uses MaxEnt, ArcMap, and SDM Toolbox but users can easily substitute similar software as needed.

Species Distribution Models
We used species occurrence data from 12,870 herbarium records of our 12 target species (285-2,807 per sp.) to build species distribution models (SDMs) using the niche modeling software MaxEnt 3.4.4(Phillips et al. 2017).All species occurrence data were sourced from the SEINet Portal Network (Gilbert et al. 2019) and truncated to exclude records prior to 1995 since georeferencing prior to this time is often estimated based on herbarium voucher labels, which may introduce unnecessary error into SDMs (Bloom et al. 2018).We used 19 BioClim climate variables (BIO1-19; Fick & Hijmans 2017) at a 30-arc-second resolution (ca. 1 km 2 ) to quantify habitat suitability across each species' range within a management boundary for the Colorado Plateau region of the Intermountain West.We clipped bioclimatic raster layers to our management boundary in ArcMap 10.6 using the SDM Toolbox (Brown 2014).We then averaged 20 model replicates per species to identify the minimum presence logistic threshold for each SDM (Phillips et al. 2006) and created a binary raster depicting suitable or unsuitable habitat.It was appropriate to sometimes use low threshold values to ensure models captured the majority of suitable habitat for these common species while excluding habitat wherein the probability of species' occurrence was low (van Proosdij et al. 2016).Note that advanced settings are available in MaxEnt to format background assumptions based on individual study systems (Fourcade et al. 2014).A detailed, usable protocol for building SDMs is provided as Supplement S1.

Disturbance Data
We compared our SDMs with disturbance layers of management areas (i.e.energy production, livestock grazing, and wildfires; Table S4) as a proxy for land area likely to be prioritized for future native plant restoration activities on the Colorado Plateau.Disturbance layers related to energy production include geospatial layers for all oil and gas wells, crude oil and natural gas pipelines included in the U.S. Department of Homeland Security's HIFLD geodatabase (https://hifld-geoplatform.opendata.arcgis.com/),and coal mines included in the U.S. Energy Information Administration's data layers (https://www.eia.gov/maps/layer_info-m.php).We also included geospatial layers of active livestock grazing allotments for lands managed by the U.S. Forest Service (https://data.fs.usda.gov/geodata/edw/edw_resources/meta/S_USA.Allotment.xml) and Bureau of Land Management (https://www.sciencebase.gov/catalog/item/5d810644e4b0c4f70d056dad).Last, we included the perimeters of all wildfires reported in the U.S. National Interagency Fire Center's geospatial database (https://data-nifc.opendata.arcgis.com/datasets/historic-geomac-perimeters-2019/) that occurred between 2000 and 2019 within our Colorado Plateau management boundary.We acknowledge that our grazing data do not fully represent the spatial extent of livestock grazing in the region since we were unable to obtain data for actively grazed lands managed by tribal nations (approximately 25% of the total area analyzed).Thus, any grazing estimates described in this study are underestimated.

Climate Suitability Models
We created climate suitability models for the 28 commercially available materials for our focal species using the USGS' Native Plant Seed Mapping Toolkit (https://rconnect.usgs.gov/seed-toolkit/).The tool has a variety of uses including the ability to take Euclidean Distance information and generate climate similarity maps between the origin of a particular restoration material and any user-selected geographic area.species or materials if relevant information exists (e.g. common garden performance data, genomic data, etc.).Given that few data exist for our focal species and that a primary goal of our study is to illustrate our methodological pipeline, we equally weighted each of the seven climate variables included in the Native Plant Seed Mapping Toolkit (i.e.mean annual temperature, diurnal range, temperature seasonality, temperature in the wettest quarter, mean annual precipitation, precipitation seasonality, and precipitation of the warmest quarter).Note that future studies may benefit from adjusting variable weights based on output needs and system-specific knowledge of environmental drivers (e.g. Warren & Seifert 2011).We then downloaded binary raster datasets representing 80% climate similarity models for each commercially available native plant material.A detailed, usable protocol for building climate suitability models is provided as Supplement S2.We quantified the total estimated area of disturbed habitat for each of our 12 focal species in ArcMap 10.6.We also estimated disturbance area for energy production, livestock grazing, and wildfires separately for each species.We then compared the climatically suitable area of each commercially available material again using the Native Plant Seed Mapping Toolkit; this time using the Climate Partitioning Tool to identify potential areas where future native plant materials would ideally be sourced from.We delineated three climate partitions for each species for illustrative purposes and provide a detailed, usable protocol for building climate partition layers in Supplement S2.Last, we calculated the total estimated area of disturbed habitat for which there are no climatically suitable materials.All data were projected in the GCS WGS 1984 Geographic Coordinate System and the USA Contiguous Albers Equal Area Conic USGS projection was used for all spatial calculations.All data and generated layers are publicly available in USGS's ScienceBase Data Catalog (Sterner et al. 2023).

Results
Twenty-eight plant materials are commercially available for 6 of the 12 priority restoration species examined in this study as of October 2022 (Table S1; Fig. 1).Five of the twenty-eight commercial materials were sourced from areas on the Colorado Plateau (Elymus elymoides: Massadona, Achnatherum hymenoides: Chipeta, Ouray, Star Lake, White River) with the rest sourced from across nine states, spanning ca.1.4 million km 2 (Fig. 1).

Disturbance-Generated Restoration Need
Fifty-four percent of the Colorado Plateau is disturbed by one or more of the following three disturbance types: livestock grazing (190,858 km 2 , 50% of the Colorado Plateau), wildfires that have burned in the past 20 years (20,273 km 2 , 5% of the Colorado Plateau), or energy production from oil and gas wells, natural gas pipelines, and coal mines (21,970 km 2 , 6% of the Colorado Plateau; Figure 2).Total disturbance covers 51-56% of each species' habitat with livestock grazing covering most of this area (Table 2; Fig. 2).Livestock grazing covers approximately 50% of each species' habitat, energy production covers 5-8%, and wildfires burned 1-6% of each species' habitat during 2000-2019 (Table S2).

Climate Suitability of Existing Materials
Climate suitability maps created using the Native Plant Seed Mapping Toolkit (using an 80% climate similarity to the material's origin location) indicate that of the 28 commercial materials, only 3 are climatically similar to more than 50% of the species modeled habitat on the Colorado Plateau (Table S1).Sixteen commercial materials are climatically similar to less than 10% of the species habitat, and 9 commercial materials are climatically similar to 20-40% of the species habitat (Table S1; Fig. S1).Across all commercial materials, 15-93% of the species ranges are not climatically similar to any plant materials (Table 2).The remaining seven species did not have a commercial material available (= 100% of species range without climatically suitable materials for that species; Table 2).Seventy-five percent of Bouteloua gracilis habitat and 70% of Machaeranthera canescens habitat is unsuitable to currently available plant materials (Table 2). A. hymenoides and E. elymoides are the only two species with commercial materials whose germplasm was originally sourced from the Colorado Plateau (Fig. 1).Eighty-five percent of A. hymenoides habitat and 77% of E. elymoides habitat was climatically suitable for those materials (Table S1).Note that the number of commercially available materials ranges from 0 to 13 across our 12 focal species.Still, across all species 47-56% of habitat identified as unsuitable for existing materials is also disturbed (Table 2; Fig. 3).Climate partitions can be designed for each species to identify potential regions from which to source future native materials (Fig. 4).These partitions can be adjusted based on user needs if, for example, restoration mangers are following already-existing seed transfer zone boundaries (Fig. S2) or seek a higher resolution reference to minimize seed transfer distance.

Discussion
Ecological restoration strategies have recently developed to address restoration needs at local (e.g.Edrisi et al. 2020), regional (e.g.Houseal & Smith 2000), national (e.g.Zuleta et al. 2015), and global scales (e.g.Brancalion et al. 2019; but see Chausson et al. 2020).The United Nations Decade on Ecosystem Restoration now stands as the world's largest coordinated call-to-action, with a goal of halting and reversing land degradation across more than 1 billion ha globally (UN 2021).As management guidelines increasingly prioritize the use of locally adapted, genetically appropriate plant materials, demand for such materials will likely continue to outpace supply (Jalonen et al. 2017;National Academies 2023).This is a pressing challenge to address given the scale of restoration needed in drylands at regional and global scales (>50% of drylands are degraded as of a couple decades ago and this number is surely larger today; UN Development Programme 1997).This challenge is also currently hindered by the potential need for plant materials for all ecologically, genetically, and/or climatically derived seed transfer zones (e.g.Doherty et al. 2017;Oldfield et al. 2019).Despite evidence suggesting the utility of genetic data in selecting seed sources (e.g.Breed et al. 2018;Bucharova et al. 2019), these data are relatively rare and/or cost-prohibitive in many restoration areas.Climate similarity matching is oftentimes the best, if not only, approach available for many systems.Our study utilizes the greater Colorado Plateau region of the western U.S. to model how our methodological pipeline can produce multiple, spatial products to guide restoration and land management practice in this and any region where restoration is needed and geospatial information is available.
There are approximately 25,000 plant species native to the United States, of which ca. 26% are commercially available (White et al. 2018).Many of these are likely not suitable for conserving existing genetic diversity or restoring diversity recently lost to disturbance on federally managed wildlands (Espeland et al. 2017).For example, the state of Utah alone currently uses ca.90 native plant species in seed-based restoration project on wildlands, of which there are ca.15 commercially available materials (A.Pilmanis pers.obs.).Approximately 98% of the plant materials (ca.800+) that would be needed to represent all species for all existing seed zones do not yet exist.The commercially available materials we included in this study include cultivated varieties developed through a genetically manipulated track with the intention of making them usable over wide geographic areas and natural track varieties selected primarily for their purported suitability for particular environments (Aubry et al. 2005).Provenance specific materials, or those which are available from known geographic seed sources, have not traditionally been produced by the USDA's Natural Resources Conservation Service, the predominant developer of plant materials over the last 80 years in the United States.These types of seed sources are rarer and oftentimes more challenging to produce and thus costly, but demand for such materials is increasing (Jones 2019;National Academies 2023).The five commercial materials in our study that were originally sourced from areas on the Colorado Plateau had some of the largest areas where materials had 80% or more climate suitability (E.elymoides: Massadona, A. hymenoides: Chipeta, Ouray, Star Lake, White River).Of our 12 focal species, we show that 15-93% of the species ranges did not have climatically suitable materials.Further, 47-56% of habitat we identified as unsuitable for existing materials is also disturbed.Future management efforts to source potential native plant materials could prioritize areas we identified as having no suitable materials that are also disturbed (Haidet & Olwell 2015).It may be useful to account for the influence of missing data (e.g.our grazing layer did not include Table 2. Total area (km 2 ) of modeled species ranges on the Colorado Plateau, area and percentage of disturbed habitat, area and percentage of habitat unsuitable for commercially available plant materials for each species, and area and percentage of unsuitable habitat that is also disturbed.Unsuitable habitat was determined as the area within a species habitat that does not meet an 80% climate similarity threshold match to the source location of any commercially available materials.
Red bold text indicates species with more than 90% of range with no suitable materials.).Additionally, disturbance severity may be an effective alternative to prioritizing areas to source native plant materials from (e.g.Barga et al. 2023) and/or to restore (e.g.Creutzburg et al. 2022).

Species
Our study provides a methodological pipeline that can be tailored to multiple needs and uses.For example our study utilizes 12 priority restoration species, though maps and spatial layers similar to those presented in our study can be developed for pooled communities (e.g.those species included in seed mixes) or can be based on functional trait data (Zeldin et al. 2020).This approach is possible if data exist for all species included in a seed mix.Users can also customize their management boundaries to include seed transfer zones based on genomic data, differently weighted climate variables, or a combination of both (e.g.Massatti & Winkler 2022).Disturbance layers can also be customized to include specific regional challenges including invasive species (Mainali et al. Figure 3. Areas within the Colorado Plateau where existing commercially available plant materials are suitable (purple), and areas where there are no existing suitable materials (yellow) as well as areas that have experienced disturbance (red lines).Six of the 12 species represented in this study did not have commercially available NPMs leaving the entire species habitat to be classified as unsuitable to germplasms.
Native plant materials selection tools 2015) or modified to consider suitability of plant materials under future climate and ecologically relevant drought conditions (McKay et al. 2005;Havens et al. 2015;Bradford et al. 2020).These alternative data can be used early on in our methodological pipeline, for example, to build SDMs using future climate scenarios (García-Valdés et al. 2020), modeled soil moisture histories related to a species' regeneration niche (Petrie et al. 2017), or to create layers based on wildfire severity or management history (Sterner et al. 2022).Incorporating additional planning tools and restoration criteria for source or site selection may ensure effectiveness of limited resources (Chambers et al. 2019) and ensure needs are met for multiple end-users or across multiple jurisdictional boundaries (Creutzburg et al. 2022).
Users can also adjust our pipeline to create climate similarity models at whichever threshold deemed appropriate.For example, existing commercial materials for A. hymenoides may be suitable for most restoration projects when assuming a loose climate matching threshold (80%) as we do in the present study.However, the climate space across the Colorado Plateau is highly Figure 4. Climate partitions within habitat areas that are not suitable for any commercially available plant materials, identifying regions where seed selection and increase could be used to source future materials.These partitions are created within habitat unsuitable to currently available materials as depicted in yellow in Figure 3.
variable and can be broken down into more than just a few categories when considering genomic (Massatti & Winkler 2022) and phenotypic patterns (Johnson et al. 2012) of A. hymenoides that may represent adaptation.Thus, as additional regionally appropriate plant materials are developed to meet restoration needs and rigorous field-based testing characterizes niche space, higher climate similarity threshold values (i.e.≥90%) can be used to infer seed transfer zones across landscapes (Johnson et al. 2012).Until then, lower threshold values seem appropriate if only to identify which of the currently available plant materials are relatively suitable for a particular restoration site.The protocols provided as Supplemental Materials, and our example spatially referenced data sources, enable users to customize each step of our described pipeline based on their needs and available data.
Matching environmental characteristics of a seed source site to a target restoration site (Doherty et al. 2017) is perhaps the second most used approach to selecting seed sources after simply using what is commercially available and/or affordable (National Academies 2023).Our approach illustrates the ease of using publicly available herbaria data to generate climatically derived SDMs that can be used to test the purported suitability of existing commercial materials.Further, we developed climate partition data products to guide future management action when selecting regions to collect and increase seed for native plant materials development.In a booming native seed market, our described pipeline can enable land managers and seed collectors to target existing gaps while considering site-or region-specific needs and challenges.

Figure 2 .
Figure 2. Distribution of lands on the Colorado Plateau recently disturbed by livestock grazing (A), wildfires (B), energy production (C), and all three disturbance types combined (D).We acknowledge that, due to data availability, our grazing data do not fully represent the spatial extent of livestock grazing.Thus, any grazing estimates described in this study are underestimated.Inset of panel A: location of the Colorado Plateau ecoregional management boundary (black polygon) highlighted in the southwest region of the United States (red polygon).

Table 1 .
Our methodological pipeline as it relates to potential project goals, where to see examples of potential products in this study, the names of publicly available, spatially referenced data sourcesgenerated in this study, and where relevant methodologies are described and usable protocols are available.
unsuitable habitat as the area within a species habitat that does not meet an 80% climate similarity threshold match to the source location of any commercially available materials.The ability to weigh individual climate variables in the Native Plant Seed Mapping Toolkit allows models to be tailored to individual