A recruitment niche framework for improving seed‐based restoration

As larger tracts of land experience degradation, seed‐based restoration (SBR) will be a primary tool to reestablish vegetation and ecosystem function. SBR has advanced in terms of technical and technological approaches, yet plant recruitment remains a major barrier in some systems, notably drylands. There is an unmet opportunity to test science‐based approaches to seed mix design and application, based not only on diversity or local provenance, but on the unique recruitment strategies of species. We lay out a framework that uses a quantitative representation of species' recruitment niches to match them to targeted goals (e.g. drought or invasion resistance) and methods (e.g. precision tools and technologies) in SBR. We first describe how to quantify the recruitment niche with seed and seedling traits tied to observed recruitment responses to environmental factors. We then show how a quantified recruitment niche framework can serve as the foundation to address three major restoration challenges: (1) designing forward‐looking seed mixes that increase resilience to future climate and disturbance, (2) accounting for natural recovery in SBR planning, and (3) applying precision seeding practices to maximize restoration success. Finally, we demonstrate these ideas with existing data and discuss key challenges to adoption in SBR practice. While the ideas in this framework are based in ecological theory, they will require substantial testing and refinement by scientists engaged in SBR efforts. If this framework is integrated into research agendas, we believe it has the potential to unify and advance diverse elements of seed‐based restoration ecology and improve restoration outcomes.


Introduction
Restoring terrestrial ecosystems is a billion-dollar global industry that relies on the reestablishment of vegetation to prevent or reverse land degradation (Menz et al. 2013; United Nations Environment Programme [UNEP] 2021). Seed-based restoration (SBR) is a valuable tool to introduce plants and rebuild ecosystem function, but low rates of establishment among seeded species remain a fundamental challenge, especially in drylands (Shackelford et al. 2021). In the last decade, hundreds of research articles and several special issues have been dedicated to identifying challenges and solutions in SBR, including the need to integrate more diverse species in Author contributions: JL conceived the ideas and led the writing of the manuscript; all authors contributed critically to the drafts and gave final approval for publication. the seed supply chain and address recruitment barriers with tools and technologies that increase establishment (e.g. Kildisheva et al. 2016;Cross et al. 2020; Plant Conservation Alliance [PCA] 2021). However, research often focuses on one species or barrier at a time, making it difficult to synthesize this knowledge to improve planning and outcomes for multiple species in SBReach of which may have unique requirements for establishment. The plant recruitment niche concept provides a foundation to understand and utilize the ecological differences among species in decision-making. When described in a universal way, species' recruitment niches could be used to guide forward-looking seed mix design, account for natural recovery, and inform precision seeding practices to meet restoration goals ( Table 1).
The "recruitment niche" is the expression of abiotic and biotic conditions required for germination, emergence, seedling survival, and establishment (a subset of the regeneration niche; Grubb 1977). The recruitment niche is often applied as a conceptual tool to describe when and where species establish in SBR (Young et al. 2005;Kildisheva et al. 2016) and to inform aspects of planning like appropriate seeding time (fall or spring) or species selection for heavily-invaded sites (e.g. based on seed dormancy or germination speed, respectively; Gioria et al. 2018;Shaw et al. 2020). However, without a systematic approach to characterize and apply the recruitment niche across restoration species and challenges, many of its informative aspects may go unmeasured or underutilized. Recent efforts to quantify the environmental niche of adult plants in restoration demonstrate how a standard approach to niche characterization (in this case, based on species' geographic distributions) enables community-level planning in SBR (e.g. developing "seed menus" of climate-ready species for restoration, Shryock et al. 2022). Still, these efforts have not directly considered species' establishment barriers-a critical gap for seedbased restoration, since adult occurrence is not always a precise or useful indicator of recruitment suitability (e.g. Bell et al. 2014). The recruitment niche could also be quantified in a standard fashion to reflect the requirements of seeds and seedlings.
Our objective is to demonstrate how the recruitment niche concept can be quantified in a way that provides a universal, ecological foundation for more precise seed mix design and seeding decisions, building on existing approaches and leading to stronger community-level outcomes (Table 1). We first describe two complementary approaches to characterizing species' recruitment niches (Fig. 1), then use this as the basis for a decisionmaking framework to address three major challenges in SBR-designing forward-looking seed mixes, accounting for natural recovery, and applying precision seeding practices (Fig. 2). We also develop a case study with trait and recruitment data from a semiarid grassland (Colorado, U.S.A.) to demonstrate how quantitative models could be built to guide decisions through multiple restoration challenges (Box 1). This framework relies on robust ecological theory and growing experimental evidence, but several uncertainties remain, and ideas are largely untested. In the final section, we explore potential limitations for implementation practice and call on scientists to pursue next steps in the development and testing of this framework across a wide range of systems. As a research agenda, it has Table 1. The recruitment niche contributes to and unifies existing approaches to overcome several key challenges in seed-based restoration when quantified as the expression of abiotic and biotic conditions required for a seed to germinate, survive, and establish (or related trait proxies, Fig. 1). the potential to unify diverse efforts in restoration science and build a common foundation to guide decision-making as a wider range of species and technologies become accessible in SBR.

Characterizing the Recruitment Niche
Applying the recruitment niche concept in SBR requires a common vision for how it is characterized across diverse plant species and varieties. Hereafter, we use "species" to describe the main ecological unit used in seed mix design but emphasize that niche variation also occurs among varieties and populations of the same species (e.g. Treurnicht et al. 2020;O'Brien et al. 2021). We also emphasize examples from dryland systems experiencing substantial recruitment limitation (Shackelford et al. 2021), though the framework can be applied in other systems.
Why Do We Need the Recruitment Niche?
The concept of the plant recruitment niche was first popularized by Grubb (1977), who emphasized that different recruitment requirements, alongside other aspects of regeneration (seed production, dispersal), are critical to explaining how species persist in different environments and plant communities. Although SBR delivers seed onto the land to bypass limitations related to seed production and dispersal, recruitment can still fail due to limited seed persistence, germination, seedling emergence, or establishment (James et al. 2011), and the probability of success can vary widely among species (Larson et al. 2015). The recruitment niche describes the different abiotic (e.g. light, water, temperature) and biotic (e.g. competition and soil biota) conditions that support a species' transition from seed to established plant. If we can quantify and compare recruitment niches in a standard, repeatable way across species, we can shift from singular approaches (understanding one species' requirements at a time) towards a framework that aids decisionmaking at the community-level-developing diverse seed mixes in which species' recruitment strategies are better matched to environmental conditions, restoration goals, and seeding support tools. Restoring diverse species assemblages is often desirable because species that respond differently to environmental change support stability across multiple ecosystem functions (A) (B) (C) Figure 1. We describe two possible methods to quantify the variation in the recruitment niche as a function of either observed recruitment responses or related trait proxies. (A) The recruitment response niche space could be represented as a multi-dimensional space in which each axis directly captures variation in recruitment sensitivity to key environmental or biotic factors. For example, species (i.e. points) may vary in their capacity to recruit under drought (blue dimension), frequent disturbance (gray dimension), or competition (purple dimension). A species' niche is characterized by its position along each dimension, indicating the direction (positive or negative) and degree (e.g. more or less positive) of its response to that factor. Species' response scores could be estimated as the raw difference in recruitment between two experimental conditions (e.g. wet vs. dry), the slope of recruitment change over a gradient, or as a threshold if recruitment responses are nonlinear. While points represent a species' mean response, variability could also be quantified (i.e. black error bars around red point; analogous to niche optimum and width, Treurnicht et al. 2020). A community range (represented by the gray cloud surrounding the collection of points) could also be quantified as the unidimensional or multidimensional niche space occupied by a collection of species (e.g. in a seed mix) as an indicator of recruitment response diversity. (B) Functional traits may serve as proxies of these niche dimensions when species' attributes are correlated to their response, as suggested for disturbance (gray dimension, e.g. Archibald et al. 2019;Dobarro et al. 2010;Wigley et al. 2020), drought tolerance (blue dimension; Larson et al. 2020) and competition (purple dimension; e.g. Gioria et al. 2018). For response to competition, favorable traits may depend on the trait(s) of the dominant species (e.g. early or late phenology). Furthermore, it is possible for multiple strategies to convey success in this (or any) dimension, leading to nonlinear responses (e.g. competing via early phenology, or avoiding via later phenology). This is not an exhaustive list of relevant traits, and multiple traits may contribute to a given response.
(C) Once traits are identified as proxies of each dimension, the recruitment niche space could be quantified as a multidimensional functional space, with species' mean trait values serving as indicators of recruitment response to corresponding environmental and biotic factors. (e.g. Sasaki et al. 2019). However, summarizing the recruitment niche concisely for many species is challenging: differences in how species develop seed banks, respond to germination cues, and subsequently grow as seedlings each contribute to recruitment outcomes in different contexts (Donohue et al. 2010). Here we describe two complementary approaches that could be used to represent species' recruitment niches within different multidimensional spaces (Fig. 1)-one based directly on observed recruitment responses to environmental and biotic factors (i.e. "recruitment response niche") and another that uses species' traits as a proxy of these responses (i.e. "recruitment functional niche"). In either approach, a species' recruitment niche is represented quantitatively by its position along multiple response or trait dimensions. The total space occupied by an entire species mix (i.e. "community range") can also be summarized to describe recruitment niche diversity in a species mix (e.g. following approaches developed for functional diversity metrics, Mammola & Cardoso 2020). This framing of the recruitment niche thus allows strategic planning for SBR at both the species and community level.

Recruitment Response Niche
The most direct quantitative expression of the recruitment niche would be based on observed recruitment responses to abiotic or biotic factors that limit or differentiate a species' recruitment success (e.g. Wandrag et al. 2019;Fig. 1A). In this approach, each relevant abiotic or biotic factor represents a unique dimension of the recruitment response niche (e.g. drought and competition), with species' positions along each dimension indicating their degree of sensitivity to that factor. For example, in the illustrated case study (Box 1), species recruiting only under wet spring conditions would be positioned on the opposite end of a precipitation response dimension from species that recruit relatively well under dry spring conditions (Fig. B1B, top panel). Response dimensions could be quantified for any measurable abiotic or biotic factor that alters recruitment outcomes (e.g. weather, soils, fire, invasion level, grazing intensity). Species' responses to each factor could also be quantified in different ways, for example, as the average difference in recruitment success between two experimental treatments or the change in recruitment success along an environmental gradient over space and time (see also Fig. 1). However, care must be taken to select dimensions that represent key recruitment barriers or management goals. For example, emphasizing drought tolerance will be uninformative if overwinter freezing mortality is a stronger driver of recruitment during restoration (e.g. Baughman et al. 2022).

Recruitment Functional Niche
Functional traits can serve as useful indicators of species' responses, offering an alternative foundation for representing species' recruitment niches (e.g. Treurnicht et al. 2020 Fig. 1B). These "functional traits" have been explored most within mature plants, but there is increasing evidence that traits related to seed persistence, germination, and seedling growth can also serve as indicators of recruitment function and responses in the field (Larson & Funk 2016;Saatkamp et al. 2019). In a functional niche approach, traits with the greatest relevance to key recruitment barriers become proxy dimensions of recruitment niche space, with species' mean trait values determining their position along each dimension (Fig. 1C).

Which Niche?
There are tradeoffs that may favor a response or functional niche approach in different scenarios, but the benefit of a framework linking the two approaches is that they can also be integrated as needed. The recruitment response niche approach is appealing Building on the starting vegetation, a target species pool can be developed including both pre-existing species (black points) and new species (red points) that fit within a targeted community niche range reflecting forward-looking restoration goals. For example, regions of the recruitment niche may be targeted that support continued recruitment under drought or competition from invasive species (green region). (C) Challenge 2. When disturbance has occurred, the recruitment niche can inform predictions of natural recovery and whether it is likely to support the target composition (panel B) without seeding. The recovery pool could be estimated as a function of species' initial abundances (i.e. starting vegetation), recruitment niche dimensions that impact seed banking and recruitment (green filter), and site or disturbance conditions that impact recruitment outcomes (e.g. degree of exotic invasion; purple filter). The difference between the composition of the target species pool and the expected recovery pool represents the recovery deficit-species that are desirable for restoration but less likely to recover naturally based on their recruitment niche. While some species may be prioritized for restoration based only on seed limitation (i.e. not abundant in the seed bank, light green region), others may be additionally limited by site conditions that pose barriers to recruitment (dark green region). (D) Challenge 3. Once species are targeted for restoration (i.e. a seed mix is designed), the recruitment niche can inform precision seeding practices-the directed use of tools and technologies-to overcome anticipated restoration barriers for each species. A seeding strategy can be developed by modeling likely recruitment barriers for each species as a function of their recruitment niche (green filter) in the context of forecasted environmental conditions, for example a dry planting year (blue filter) and competition from invasion (purple filter). If the functional niche reflects responses to these key factors (see Panel A), it can be used to identify species that are most susceptible to recruitment failure under drought (blue region) or invasion (purple region) and apply targeted solutions accordingly. For example, seed enhancements that support moisture retention and growth under stressful conditions (once commercially developed) could be applied to drought-limited species, while the decision to manage invasive species during restoration may be based on the portion of seeded species that are limited by invasion (purple region).

Box 1 Case study: Making restoration decisions based on the recruitment functional niche
To demonstrate implementation of the recruitment niche framework, we developed a set of models using data for 11 grass and forb species that are commonly found in semiarid, mixed grasslands near Boulder, CO, U.S.A. (data from Larson et al. 2021;Larson & Suding 2022). This case study is meant to illustrate how cumulative models can synthesize decision-making across multiple challenges in SBR, but we note that resulting seeding recommendations ( Fig. B1E) have yet to be evaluated in this system. Once relationships like these are tested and evaluated across a wider range of species and environments, we envision the development of regional decision support tools that could make such recommendations for land managers with a specified degree of certainty. We implemented all modeling in R (R Core Team 2021) and include underlying data and code in online supporting information (Supplement S1).

Data inputs
First, we compiled data on four traits to be tested and used as dimensions of the recruitment functional niche: seed mass, seed dormancy index, emergence timing, and life history ( Fig. B1A; full details on trait data collection can be found in Larson et al. 2021).
We also compiled observed recruitment, seed banking, and abundance data for each species to quantify the explanatory power of the recruitment functional niche. In relation to Challenge 1, we tested trait proxies against two recruitment response metrics capturing potential restoration goals in this system: drought response and disturbance bet-hedging. Recruitment drought response was estimated as the log response ratio of total recruitment (% of sown seeds) between wet and dry experimental treatments in a common garden (Fig. B1B, top panel). Bet-hedging potential was estimated as the mean portion of nonemerged seeds in the first year after sowing in the same common garden, taken as an indicator of seed persistence in soil (Fig. B1B, middle panel). Both metrics were derived from experiments in Larson et al. (2021), where seeds were buried in a protected area with minimal signs of seed loss to external factors.
In relation to Challenges 2 and 3, we also estimated two metrics related to natural recovery potential and competition response. Seed bank recovery potential was estimated as the log response ratio of a species' mean relative abundance in the seed bank versus the vegetation. Competition response was estimated as the linear correlation between each species' vegetative abundance and that of the most common, exotic species in this system. Both metrics were estimated from a dataset of species' abundances in local vegetation and seed banks (Larson & Suding 2022).

Approach
We quantified a simple recruitment functional niche from four traits (three continuous, one categorical; Fig. B1A), then used these trait dimensions to address the three identified restoration challenges in succession ( Fig. B1B-D). First, we estimated quantitative links between recruitment traits and the responses or functions they were expected to represent (i.e. upper panels in Fig. B1B-D). These included tests of the following trait proxies: seed mass as an indicator of drought tolerance (Fig. B1B, top panel), seed dormancy index as an indicator of bet-hedging potential ( Fig. B1B middle panel), seed mass and life history as indicators of seed-banking potential (Fig. B1C, top panel), and mean emergence time (ranked across species) as an indicator of response to a competitive, early-season dominant species (Fig. B1D).
We then used the parameters from each of the modeled relationships (slope(s), intercept(s)) to estimate predicted values for each of the four responses based on species' traits. For example, we used 9 of 11 species to estimate the relationship between seed mass and recruitment drought response (Fig. B1A, top panel), then extracted predicted values of water response for all 11 species (including those without response data). We used these predicted values to address each challenge.

Challenge 1-Designing forward-looking seed mixes
In the first challenge, we aimed to derive a "target species pool" of high-priority species for restoration based on two forwardlooking restoration goals: drought resistance and bet-hedging against disturbance. We used seed mass to estimate each species' recruitment drought response (Fig. B1B, top panel) and a seed dormancy index to estimate bet-hedging (based on nonemergence in the first year; Fig. B1B middle panel). We scaled species' predicted responses so that higher values of each (closer to 1) represented the desirable function (i.e. drought tolerance or bet-hedging capacity) and lower values (closer to 0) represented less desirable functions. We then multiplied scores for these two functions to generate a selection index-where high values indicate higher likelihoods of both drought tolerance and bet-hedging potential based on species' traits ( Fig. B1B, lower panel; note that this index could be weighted to prioritize different functions). This selection index could be used to identify a target species pool whose recruitment niches support forward-looking restoration goals (darker green points).

Challenge 2-Accounting for natural recovery
In the second challenge, we estimated a "recovery pool" reflecting which species are likely to be found in the seed bank, and thus recover naturally, without seeding. We used seed mass and life history to estimate each species' seed bank recovery potential (Fig. B1C, top panel). Because this reflects a ratio of abundance-with higher values indicating commonness in the seed bank relative to the vegetation-we then scaled this metric and multiplied it by species' mean abundances in local vegetation to estimate Continued Restoration Ecology September 2023 Figure B1. Demonstrating implementation of the recruitment niche framework using (A) trait data from 11 co-occuring grass and forb species in order to: (B) quantify links between species' traits and recruitment responses (upper two panels) and use these to identify a target species pool for seed mix design (lower lanel); (C) quantify links between species' traits and seed bank prevalence (upper panel) and use these to model likelihoods of natural recovery (lower panel); and (D) quantify links between species' traits and co-occurrence with a dominant, competitive species. (E) Results from these models (B-D) are combined to synthesize seeding recommendations in this system. See Box 1 text for full details on models and figures.
because it directly reflects species' environmental requirements for recruitment rather than relying on traits as proxies. This could lead to more direct application with greater certainty. For example, larger seed mass is a good-but not exact-indicator of stronger recruitment under drier conditions (Fig. B1B); some information on drought sensitivity is lost when using a trait proxy. However, depending on how they are quantified, recruitment response metrics also contain uncertainty. For example, it will be easier and faster to quantify recruitment drought responses for many species in controlled, common garden settings, but harder to translate those response patterns to field dynamics if unmeasured environmental variation alters expectations (e.g. Alba et al. 2017). In this case, networks of common gardens or restoration experiments utilizing the same species may be a practical solution making it feasible to collect field recruitment data over larger and longer scales (e.g. as in Havrilla et al. 2020).
Restoration databases that compile data on recruitment success from a wide range of field environments also offer a pathway to capture this variation and quantify more robust metrics of recruitment response (e.g. as in Shackelford et al. 2021). Still, accumulating enough data may represent a major time investment for each new species integrated into SBR. The functional recruitment niche approach offers an accessible and versatile alternative with different advantages. In a world of limited information on many restoration species, data accessibility is critical. The time investment to collect recruitment trait data is relatively low (weeks to months), and trait values can often be estimated under standard conditions for species or populations-facilitating data sharing across research efforts (e.g. Wigley et al. 2020;Leger et al. 2021). Functional niche space is also versatile: many traits are proxies for multiple aspects of recruitment and recovery, potentially reducing the number of dimensions that are needed to characterize major differences among species. For example, in the illustrated case study (Box 1), seed mass is useful as an indicator of species' precipitation responses as well as seed bank prevalence (Fig. B1C).
Finally, trait data are useful to many other applications, including process-based models that rely on germination and seedling traits to improve finer-scale forecasts of recruitment over space and time (e.g. Schlaepfer et al. 2014).
Both approaches have merit, and as research continues to establish robust links between the two, it will be feasible to draw on both direct responses and functional traits as they are available and useful. We emphasize a functional niche approach in following examples given its versatility to address all three of the challenges in SBR-including recovery from seed banksthat are discussed in the remainder of the paper.

Challenge 1: Designing Forward-Looking Seed Mixes
Changing climate and disturbance regimes are expected to impact most terrestrial ecosystems in the next century, with many species already facing poleward or upward migration, or risks of local extinction (Parmesan et al. 2022). Consequently, resilience under warming climates and more variable conditions-due to both extreme weather and increasing disturbance (e.g. fire and land use)-is a growing consideration in restoration planning (e.g. Butterfield et al. 2017;Barak et al. 2022). Species that cooccur locally still vary in recruitment response to environmental and biotic variation (e.g. Wandrag et al. 2019). As climate models and near-term weather and disturbance forecasts become more accurate (e.g. Hagger et al. 2018;Smith et al. 2022), a recruitment niche approach could make it possible to design forward-looking seed mixes with species that are both able to establish in the short term, and contribute to resilience in the long-term.

Why Do We Need the Recruitment Niche in Seed Mix Design?
One approach for climate-ready restoration is to select seed sources that are matched to future climate conditions based on current distributions. However, evidence for success is limited (Bucharova et al. 2019), and for long-lived species, species Box 1 Case study: Making restoration decisions based on the recruitment functional niche a recovery index (Fig. B1C, lower panel; relative abundance data from Larson et al. 2022). This approach can be used to identify seed bank-available species (darker blue points) as the likely recovery pool To prioritize species for seeding, the target species pool (Fig. B1B lower panel) was compared against the recovery pool ( Fig. B1C lower panel) to identify species that meet key restoration goals but are least likely to recover without seeding (the recovery deficit, Fig. B1E). To do this, we ranked species' scores along the selection index (high ranks = high desirability) and recovery index (high ranks = high recovery potential) and subtracted these to estimate a recovery deficit index. Species that rank higher in desirability than recovery are in the "recovery deficit" (i.e. positive scores in the dashed region, Fig. B1E) and priority species for SBR.

Challenge 3-Applying precision seeding practices
In the third challenge, we aimed to identify whether species within the recovery deficit (n = 5 species) would require additional tools or strategies to bolster restoration success in the context of a dominant, exotic grass species. We used species' emergence times to estimate their sensitivity to a ubiquitous, cool season grass, with negative correlation values indicating lower tolerance of competition ( Fig. B1D; based on cover data from Larson et al. 2022). We then used predicted values to identify species in the recovery deficit that are less sensitive to competition (points with lighter gray outlines; Fig. B1E). In practice, this approach could be used to identify a subset of more competitive species to sow in dominated areas. The degree of competitive intolerance across all seeded species could also inform site-level management decisions about whether to employ control measures for a dominant species prior to seeding. distributions may be more indicative of environmental suitability for mature plants than conditions necessary for successful recruitment. Seeds and seedlings can experience different conditions from mature plants given their smaller size (e.g. Poorter 2007) and may be susceptible to extreme conditions on much shorter timescales (e.g. 1-2 weeks dryspells, Larson et al. 2020). Because recruitment is a limiting factor for both short-term seeding outcomes and longer-term processes like range expansion and disturbance recovery (e.g. Walck et al. 2011;Shackelford et al. 2021), recruitment potential under changing conditions is a key consideration for forward-looking species selection. This could mean selecting species that can continue to recruit under gradually changing conditions (e.g. weighting mixes towards drought and invasion tolerance, Fig. 2B), but also selecting species to buffer against short-term environmental extremes and disturbance in the years after seeding (e.g. based on disturbance response or traits). Even when some management goals are external to recruitment-e.g. selecting species based on value for wildlife or carbon sequestration-contextualizing these species targets within recruitment niche space will help identify when desirable species are likely to be recruitment-limited and require further support (see Challenge 3).

Using the Recruitment Niche in Seed Mix Design
Identifying Focal Change Factors. A first step towards developing forward-looking species targets is to identify abiotic and biotic factors with the most likely impacts on recruitment as conditions change. Every ecosystem is different, but common threads point to a shortlist of change factors-and associated trait dimensions-that are worthy of consideration. First, changes in temperature and moisture can impact the degree and timing of seed germination and seedling establishment, with wide variation among species (e.g. Wainwright et al. 2012;Groves & Brudvig 2019;Baughman et al. 2022). Consequently, variation in traits reflecting seed temperature and moisture sensitivity or seedling stress tolerance may comprise relevant functional niche dimensions (Larson et al. 2020;Dalziell et al. 2022). In many systems, fire frequency and severity are also shifting in response to climate and land management (Smith et al. 2022). As disturbance becomes more frequent or severe, trait dimensions that are linked to avoidance of adverse conditions (e.g. seed persistence, dormancy, and light-sensitivity in soil) may be critical to consider for the resilience of restored plant communities (Archibald et al. 2019). Ultimately, the selection of focal change factors and corresponding recruitment niche dimensions should reflect both local management concerns as well as the best available science around future change.
Setting Species Targets. Based on focal change factors, relevant recruitment niche dimensions can be identified as criteria for species selection (e.g. drought-or disturbance-tolerance). By overlaying target ranges for each dimension, species can be selected whose niches support more resistant or resilient communities in the context of anticipated change (Fig. 2B, Fig.  B1B). One strategy is to target species whose individual niches support continued recruitment under a mean change in conditions. For example, if growing seasons become drier, seed mixes might be weighted towards species with drought-resistant traits. Another strategy is to target resilient species with the capacity to recover even if extreme events like drought render conditions temporarily inhospitable. For example, including additional species in seed mixes with higher levels of seed dormancy and persistence could support seed bank development and post-drought recruitment (but note that changing climate can also impact seed persistence in seed banks, Walck et al. 2011).
Finally, we can consider seed mixes that are resilient at the community level by focusing on the niche range across seeded species. Especially in scenarios where variability in conditions is the current or expected norm, targeting a wide community range across specific dimensions of the recruitment niche could increase the likelihood that some seeded species will resist or recover in any change scenario. Studies suggest that asynchronous responses of species to the environment are a critical driver of community stability through time, yet species richness is not always a useful proxy for response diversity and recruitment stability (e.g. Groves & Brudvig 2019;Sasaki et al. 2019). Targeting diversity in specific dimensions of the recruitment niche (e.g. seed dormancy, as a metric of disturbance response) offers a more directed approach to increase community resilience in variable environments.

Challenge 2: Accounting for Natural Recovery in Seeding Decisions
SBR is often used following major disturbance to curtail invasion, habitat loss, and soil degradation-especially in drylands. However, active restoration is resource-intensive and not always an improvement upon passive regeneration (Jones et al. 2018). Assessments from multiple systems suggest that natural recovery performs as well or better than seeding efforts in some contexts-particularly when other tradeoffs are considered (e.g. unintended soil disturbance or invasion impacts, Crouzeilles et al. 2017;Duniway et al. 2015). Unfortunately, failure to act can also lead to undesirable vegetation if conditions do not support recovery (e.g. Farrell et al. 2021), and it can take decades to see the full effects of seeding relative to natural recovery (Ott et al. 2019). This can lead to "action bias," in which taking any action is perceived as favorable over inaction, even if the more rational choice is to wait and collect more information (Iftekhar & Pannell 2015). Decisions about whether to seed, and which species to seed following disturbance, should ultimately depend on whether natural recovery trajectories and timelines are likely to meet management goals. To determine this, there is a need to characterize ecological contexts that support natural recovery with greater accuracy and speed so limited resources can be allocated more efficiently in SBR.
Why Do We Need the Recruitment Niche to Account for Recovery?
Following disturbance, vegetative recovery can come from several biotic sources, including recruitment from in situ seed banks, seed dispersal, and vegetative reproduction. Here we focus on seed banks as a source of recovery that is directly informed by the recruitment niche, though it is also possible to characterize a broader regeneration niche with dimensions related to vegetative reproduction and dispersal potential. Species traits related to all three modes have been informative in recovery models (e.g. Shryock et al. 2014).
Seed banks can contribute meaningfully to revegetation, especially when aboveground vegetation experiences wide mortality (limiting spatial dispersal of seeds, e.g. Ziegenhagen & Miller 2009) or when ruderal species dominate seed banks and set trajectories (e.g. Farrell et al. 2021). Natural recovery patterns will depend on both seed availability as well as post-disturbance conditions (Ziegenhagen & Miller 2009). Predictive models of recruitment potential from seed banks-based on species' recruitment niches-could illuminate likely recovery trajectories and identify desirable species that are least likely to recover without seeding (i.e. the "recovery deficit," Fig. 2C). This insight could help prioritize sites for seeding (i.e. those with the largest recovery deficits) and guide seed mix design to target species based not just on desirability, but on likelihoods of recovery.
However, leveraging natural recovery in decisions will hinge on rapid predictions of both the degree and speed of recruitment from seed after disturbance. Slow rates of passive recovery can deter willingness to wait on seeding, even though active restoration may provide little improvement (Jones et al. 2018). A quantified recruitment niche space with dimensions related to seed-banking capacity as well as germination timing and seedling development could provide an ecological foundation to avoid "action bias" by estimating likelihoods of recovery over time (e.g. across different species and weather contexts) and space (across sites with different environmental conditions).

Using the Recruitment Niche to Predict and Mitigate the Recovery Deficit
After disturbance, both the availability of seed and potential for establishment can influence the likely recovery pool and inform which species are targeted for SBR (Fig. 2C). A recruitment functional niche approach is particularly well-suited to account for these multiple aspects of recovery while minimizing complexity. For example, seed mass is a trait that could serve as a general indicator of seed bank abundance as well as emergence and establishment likelihoods under drought, making it a useful "multi-tool" in modeling (see Box 1).
Estimating Seed Bank Abundance. Seed banks store future communities and often look different from extant vegetation, for example, with short-lived, seed-dependent plants being over-represented while longer-lived plants may be virtually absent (Thompson 2000). These biases have important implications for projected recovery, yet measuring seed bank composition on large scales is infeasible. Species' traits reflecting life history, seed production, and persistence in soil (e.g. seed mass and shape, light sensitivity) can be useful proxies of seed-banking capacity (Pakeman & Eastwood 2013;Larson & Suding 2022), making species' functional niches a useful tool to translate vegetative covers to expected abundances in the seed bank. In this approach, relevant trait dimensions could be used to up-or down-weight estimated seed bank abundances based on pre-disturbance vegetative cover (Fig. B1C).
Estimating Seed-Based Recovery in Space and Time. Equal seed bank abundances will not translate to equal recovery if recruitment potential varies widely among seed bank species. Just as recruitment niche dimensions could be used as indicators of suitability for future conditions (see Challenge 1), they may inform the likelihood and timing of successful recruitment in post-disturbance conditions. Disturbance generally creates a flush of resources and space that favor species with disturbance-ready, ruderal strategies (Fig. 1B). Consequently, traits reflecting germination delay mechanisms that are triggered by disturbance (e.g. seed dormancy, light sensitivity) and rapid resource access tend to be associated with early recovery (e.g. Dobarro et al. 2010;Navas et al. 2010). However, shifts in weather conditions could alter these expectations-especially those that fail to break dormancy (e.g. warm and dry) or create fatal hazards for early seedlings (e.g. freezing and drought; Wainwright et al. 2012;Walck et al. 2011). With the help of environmental forecasting tools, modeling tools could incorporate recruitment trait by environment interactions into recovery projections (e.g. as in Zirbel & Brudvig 2020).
Targeting the "Recovery Deficit" in SBR. Species' seed bank availabilities and recruitment potentials can be combined to model coarse recovery likelihoods for the extant species pool. These multi-species models could be used to highlight targeted areas on the landscape or temporal periods that are most strategic for seeding based on broad expectations for communitylevel recovery. However, they could also identify individual species that will benefit most from seeding: target species mixes (e.g. those derived from Challenge 1) can be compared against species' likelihoods of natural recovery to estimate the "recovery deficit"-the set of high-priority restoration species that are least likely to recover without seeding ( Fig. 2C; see also lower panels of Fig. B1). When combined with the next challenge-precision matching of recruitment-boosting tools and technologies-these efforts can begin to funnel resources more effectively and efficiently to the species that will benefit most in SBR.

Challenge 3: Applying Precision Seeding Practices
Even when seed mixes are weighted towards recruitment niches that are amenable to expected conditions (Challenge 1), there will be variation in species' responses to environmental factors and acute risks. There will also be species that are essential to restore for their unique functional or conservation value, even while their recruitment niche is not ideally suited to restoration conditions. The information stored in the recruitment niche offers a roadmap to match restoration species to targeted microsites, seeding practices, and technologies to overcome remaining barriers in heterogeneous environments (Fig. 2D). Custom application of seeding tools and technologies could increase the likelihood of establishment across species in a seed mix, leading to more resilient established communities that meet management goals.
Why Do we Need the Recruitment Niche for Precision Seeding?
In the coming years, SBR will have unprecedented access to new tools and seed enhancement technologies-from the application of seed coatings to drone-based microsite-matching Castro et al. 2022). Many restoration technologies originate from agricultural science where use is customized to individual crops, while restoration efforts are tasked with improving establishment for dozens of species with different recruitment niches. Without a framework to streamline the matching of seeds to appropriate tools, their application may remain limited to only the most common restoration species.
We propose that a recruitment niche approach can make precision seeding practices possible for more species by characterizing ecological differences that are directly related to the functionality of tools and technologies. For example, seed light sensitivity is a trait that impacts the likelihood of germination when seeds are covered (Dobarro et al. 2010), offering a rapid species-level indicator for appropriate burial depth during seeding and seed coating techniques (as coatings can obstruct light). Furthermore, a collective view of recruitment strategies in a seed mix (i.e. the community niche range) can guide decisions around site-level restoration methods that impact all species. For example, the decision to pursue invasive species management at a site could be based on the overall susceptibility of seed mix species to exotic competition (as indicated by the community niche range along a competition dimension).

Using the Recruitment Niche in Precision Seeding
Targeting Spatiotemporal Microsites. Deciding when and where to seed within a site-with respect to spatiotemporal heterogeneity in weather, microtopography, exotic cover, or other factors-can be a major driver of SBR success (e.g. Groves & Brudvig 2019;Shaw et al. 2020). Dimensions of the recruitment niche reflect continuous variation in environmental and biotic requirements that can guide these decisions at the species or community level. For example, heavily invaded areas on the landscape could be seeded with a subset of species from the target seed mix expected to be generally tolerant of competition (e.g. based on emergence phenology; Fig. B1D). A recruitment niche approach also provides a foundation to match species to finer-scale spatial microsites as more precise remote sensing and seed delivery technologies develop (e.g. based on spatial variation in soil moisture or light availability). This application extends to temporal matching as well. For example, variation in seed dormancy can inform seasonal seeding windows, while drought response dimensions could help optimize the use of forecasted wet or dry events in SBR scheduling based on species' water sensitivities (Hagger et al. 2018).
Applying Seed Enhancement Technologies. Even as our understanding of seed-microsite matching improves, it will be challenging to coordinate different seeding times and locations to match species' recruitment requirements at large scales. As an alternative, researchers are developing seed enhancement technologies that are applied prior to deployment to artificially create optimized "microsites" around seeds and/or to promote recruitment (Madsen et al. 2016;Pedrini et al. 2020). Major areas of development include coating, pelleting, or priming technologies that retain moisture, enhance beneficial microbes and seedling growth, or alter germination timing to avoid hazards (e.g. Chua et al. 2020;Zvinavashe et al. 2021). As technologies are scaled up, quantified recruitment niches will be an essential tool to match enhancements to appropriate species and settings, increasing the cost-effectiveness of their application. For example, moisture-related seed enhancements could be applied according to species' positions along droughtresponse dimensions and paired with near-term rainfall forecasts to prioritize application in water-sensitive species and drier contexts. Similarly, under conditions of extreme temporal variability, seed coatings that delay germination could be targeted towards species with fast germination (e.g. Baughman et al. 2022) in a way that spreads germination (and risk of post-germination hazards) across a wider range of temporal windows-giving a single seeding event a wider timeframe to succeed.

A Long-Term Vision for SBR: Limitations and Next Steps
By uniting research in seed-based restoration around the recruitment niche, we can build a common scientific foundation to guide the use of an increasingly diverse species pool in seed mix design and precision seeding. The ideas laid out in this framework are based in ecological theory, yet largely untested in realistic restoration settings. Here we describe several caveats limiting the implementation of a recruitment niche framework in restoration practice and illuminate a path forward for innovative and systematic research to test these approaches.
First, the diversity of species that can be utilized in SBR practice and research is bounded by the availability and cost of seeds (e.g. Palma & Laurance 2015; Barak et al. 2022). A narrower species pool could substantially limit the design of different seed mixes based on recruitment strategy and could pose a particular issue if some areas of niche space are disproportionately unavailable (e.g. species with high seed dormancy levels). However, a quantified recruitment niche framework is still useful for matching available species to more effective practices in restoration (Challenge 3) and to account for natural recovery dynamics in existing communities (Challenge 2). Furthermore, a recruitment niche approach could help identify species that should be targeted for seed production based on management goals (Challenge 1). Characterizing recruitment niches preemptively for a diverse range of current and emerging restoration species will make the approaches described here ready for wider implementation as seed supplies develop.
Another caveat is the complex nature of this framework, which aims to simultaneously address several complicated SBR challenges with the recruitment niche as a universal basis for decision-making. The resulting risk is a framework that is too complex to develop into a concise multi-tool for SBR planning, yet too simplified to meaningfully address any of the individual challenges described here. Before the framework can be implemented as a whole, its predictive capacity must be explored within individual challenges. With traits as a common foundation for inference, these independent efforts can subsequently be connected to build and test more complex predictions and seeding strategies. We demonstrate this in the illustrated case study (Box 1), where a single set of traits was applied to data from multiple studies quantifying recruitment potential under drought, seed bank storage, and co-occurrence with an exotic grass (Challenges 1, 2, and 3, respectively). Collectively, these data were used to synthesize a seeding approach that considered restoration goals, natural recovery, and best seeding practices. This versatile application of species' traits is one strength of a functional niche approach, which could reduce complexity by informing multiple aspects of decision-making.
To apply the recruitment functional niche, however, limitations must be addressed in each part of the framework. First, links between traits and recruitment response still need extensive testing to quantify predictive power and remaining uncertainty in different contexts. While early trait-based models are promising (e.g. Larson et al. 2015;Zirbel & Brudvig 2020), several advancements are critical. To design forward-looking seed mixes (Challenge 1), empirical tests must disentangle how species' recruitment niches mediate responses to both single and interacting environmental factors (e.g. drought and disturbance) to plan around changes occurring in tandem (e.g. Enright et al. 2014). Furthermore, expected changes in regional environmental factors could translate to variable recruitment patterns locally based on small-scale heterogeneity (e.g. due to microtopography or plant neighbors, O'Brien et al. 2021). Consequently, models that integrate environmental variation experienced by seeds at local scales could clarify how recruitment niches are linked to suitable restoration sites and microsites, for more precise species-matching.
Accounting for natural recovery from seed (Challenge 2) may be the most complex task in this framework, requiring an understanding of both storage in and recruitment from seed banks in post-disturbance environments. Trait-based models of both processes require substantial testing, ideally based on observations of recovery following real disturbance-from species' initial abundances in the vegetation and seed bank through their reestablishment over time (e.g. Ziegenhagen & Miller 2009;Abella 2022). In the long run, models based on recruitment niches need not capture every aspect of natural recovery to be useful, but enabling successful SBR in drylands requires that we identify and represent the factors that most influence recruitment.
As we build greater certainty around recovery, we can turn more attention towards improving outcomes for species in the "recovery deficit" (Challenge 3). A first challenge for precision seeding practices based on species' recruitment niches is the limited accessibility of scalable tools at present. Many technologies that will make microsite-matching or amelioration possible (e.g. drone seeding, seed coating enhancements) are early in development with testing mostly limited to common species. However, this also presents an early opportunity to shape forthcoming research and development-selecting study species based not only on their commonness in SBR, but on their recruitment niche (e.g. as described by Dalziell et al. 2022). For example, seed coatings that delay germination should be most effective when applied to species with nondormant seeds (e.g. Baughman et al. 2022), so testing should prioritize species spanning the dormancy dimension of the recruitment functional niche. This approach paves the way to validate the efficacy of new technologies while guiding their application to additional species based on their recruitment niche.
We do not expect the recruitment niche to explain all variation in SBR outcomes, and it will be critical to describe precision and uncertainty as these approaches are tested within each of the described challenges. Whether strategies derived from this framework are deemed "worth the effort" to adopt in practice will likely depend on accuracy of predictions as well as the current cost of restoration failure under the status quo. The effort of implementing this framework may not be worthwhile in all contexts (e.g. smaller-scale projects with ample resources for seed collection/grow-out or relatively high establishment success). However, for broadscale SBR efforts with high ecological value that cost huge amounts of effort, time, and money, even small increases in success resulting from recruitment niche-based strategies could become valuable. Clearly quantifying certainty and effectiveness (relative to existing approaches) will enable practitioners to make these valuations.
Testing and verifying this framework will require an immense amount of data on spatiotemporal recruitment patterns and traits for current and potential restoration species. This highlights a critical role for researchers to lead data collection efforts in a way that is comparable, accessible, and useful in the context of SBR. While it is challenging to collect enough data to understand even local recruitment dynamics, a spacefor-time approach that monitors recruitment across a range of controlled or natural environmental conditions may facilitate generalization. Expanding restoration databases may also enable syntheses across larger scales (as in Shackelford et al. 2021), but standardized data collection approaches must be established to facilitate sharing. At present, recruitment traits are only partially described in existing methods handbooks (e.g. Wigley et al. 2020), and seed bank assays can take many forms (e.g. Gross 1990).
Finally, individual researchers must work directly with managers to understand local concerns around data collection and use. For example, in some regions, it may be critical to differentiate between populations or seed sources of the same species that differ in local adaptation and suitability for restoration. There is also a critical opportunity for these approaches to complement and innovate-not replace-existing knowledge systems. For example, if practitioners have identified species that establish well under drought or competition, projects should include these species with the goal of quantifying, explaining, and sharing information on the recruitment success of these species in SBR.

Conclusion
The recruitment niche is a familiar concept in SBR: both researchers and managers draw on it to guide restoration science and practice. The vision laid out here urges researchers to invest in a standard, quantitative characterization of the recruitment niche as a common foundation to describe the recruitment requirements of species and incorporate them into precision seeding approaches. As seeds of more diverse species become commercially available, such a framework will be critical to guide their effective use in SBR-from their strategic selection for seed mixes to the appropriate matching of establishmentboosting tools and technologies. Given the breadth and depth of this research agenda, we expect it to be a long-term vision with efforts to address each challenge tested and refined at local scales first. However, with the recruitment niche as a shared foundation across studies and systems, incremental understanding can build more quickly towards cumulative understanding and progress in SBR. We encourage researchers to consider this vision as they work with managers to build long-term research programs and partnerships that center the recruitment niche in seed-based restoration.