Seed-Specific Mass and Root Growth Relate to Perennial Bunchgrass Seedling Survivorship under Highly Limited Nutrient Supply

ABSTRACT Maternal provisioning to seeds is critical to later plant success, and identifying seed traits that predict successful plant establishment and resilience may facilitate improved plant material selection for rangeland restoration. Although size-standardized measurements are typically recorded for leaf traits, this is not the case for seed traits. In this study, we investigated whether an area-standardized metric, seed-specific mass (SSM), was better associated with seedling performance relative to simple seed mass. We germinated seeds of bluebunch wheatgrass (Pseudoroegneria spicata [Pursh.] A. Löve), crested wheatgrass (Agropyron cristatum L. var. Hycrest II), and Sandberg bluegrass (Poa secunda Presl.) on sterile agar and tracked seed mass, SSM, and subsequent seedling growth and mortality to investigate relationships between seed traits and seedling performance. SSM showed clear variation among species despite similarities in seed mass, and species-specific patterns of mortality most closely tracked variation in SSM. Crested wheatgrass seedlings that survived to 52 wk had significantly greater SSM than those that perished during the study, and surviving seedlings also had significantly longer roots at wk 4 than those that died. Seed mass and SSM each explained variation in seedling traits to some extent. Simple seed mass best predicted variation in early leaf area within species, while root length was best predicted by SSM across species. Our study is indicative that SSM warrants consideration in future studies investigating maternal energetic provisioning and seedling performance.


Introduction
Plant functional traits can provide insight into bottlenecks during important life stage transitions, thus providing valuable guidance for restoration management ( Larson et al. 2015 ). Furthermore, if traits vary in a predictable manner across demographic stages, then seed or seedling traits may be early indicators of plant performance or the ability to avoid or tolerate environmental stress and disturbance. Identifying seed traits that reflect life history strategies related to stress tolerance has the potential to improve plant material selection for restoration purposes, including in degraded rangelands of the western United States. Despite the possible strengths of this approach, few seed traits have been investigated with respect to predictive functional ecology ( Saatkamp et al. 2019 ).
Seed mass is a trait of interest in community dynamics and restoration, as it corresponds with other functional traits that confer early postgermination growth and subsequent emergence and survival ( Larson et al. 2020 ). Seed mass is assumed to reflect optimized maternal energetic investment in reproductive effort, which may strongly influence seed dispersal and subsequent seedling growth, resource assimilation capacity, and competitive and reproductive ability once established ( Greene and Johnson 1993 ;Turnbull et al. 1999 ;Fenner and Thompson 2005 ;Daws et al. 2007 ; Moles and Leishman 2008 ;Larios et al. 2014 ). Recent work in dryland range grasses has shown that parental performance and seed mass can influence demographic transition probabilities among germination, emergence, and seedling survival, as well as seedling competitive ability with exotic and invasive grasses ( Espeland and Hammond 2013 ;Larson et al. 2015 ;Drenovsky et al. 2016 ).
In general, species that produce larger seeds are thought to have greater parental investment and stronger parental effects on subsequent seedling emergence and performance than smaller seeded species ( Greene and Johnson 1993 ;Reader 1993 Fenner and Thompson 2005 ;Moles and Leishman 2008 ). Within a species, parental growth conditions affect reproductive effort, and individuals that produce larger seeds often have greater seedling establishment, in part because greater parental energetic investment allows for more rapid initial growth from intrinsic seed resources ( Kromer and Gross 1987 ;Huxman et al. 1998Huxman et al. , 1999Moles and Leishman 2008 ;Larios et al. 2014 ). Greater seed reserves in high mass seeds may also allow faster initial seedling growth independent of light and soil nutrient resources, resulting in earlier establishment and light utilization (i.e., autotrophy) and offering an early competitive advantage ( Kitajima and Myers 2008 ). Improved performance associated with high seed mass may be associated with changes in resource allocation to roots or leaf material, depending on environmental conditions ( Westoby et al. 1996 ). However, seed mass does not necessarily equate with seedling performance in situ. For example, one field study illustrated how microsite differences can modulate seedling performance before and after emergence via hindering or augmenting a seedling's innate ability to establish and persist ( Merino-Martín et al. 2017 ).
Widely used metrics of vegetative functional traits are typically expressed in terms of specific mass (i.e., g m −2 structural area); this makes sense, as it is well established that leaf-and rootspecific masses reflect construction costs, as well as the ecophysiological performance attributes related to resource uptake ability and environmental stress tolerance ( Eissenstat 1991 ;Hamerlynck and Knapp 1994 ;Ostonen et al. 2007 ;Feng et al. 2008 ;Melo et al. 2021 ;Bonifas and Lindquist 2009 ;Wellstein et al. 2017 ). From an energetic standpoint, therefore, it makes sense that resource allocation to seeds may also be better reflected on a mass per unit basis. Seeds of the same mass may vary significantly in terms of their volume/area, and this difference is captured by seed-specific mass (SSM), or mass per unit area. For example, a recent study of two perennial bunchgrasses reported significant differences between seed head specific mass (mass per total seed head area) of Elymus elymoides and Agropyron cristatum , despite these two species having similar seed head mass ( Hamerlynck and O'Connor 2021 ). This suggests that some species can more densely package energetic reserves, potentially without the need to reduce seed number. Specific mass may therefore reflect important differences in overall maternal investment not captured by simple mass and thus better predict seedling traits like germination, emergence, and the ability to overcome stress during early establishment. To date, however, no studies have directly assessed specific seed mass in relation to seedling growth and performance.
We sought to address uncertainties surrounding seed mass versus SSM, via a controlled experiment in which seed mass and SSM were recorded for individual seeds of three perennial bunch grass species of differing seed size. Subsequent growth, mortality, and persistence on agar plates were monitored for a year. Our goals were to 1) compare variation in seed mass versus SSM among the three species, 2) track species survivorship and mortality in relation to these seed traits, and 3) determine intrinsic seed performance, in terms of allocation to roots and leaves strictly from maternal energy stores within the seed, by providing minimal exogenous nutrient resources to the seedlings. We predicted that species would exhibit differences in SSM, even when seed mass did not differ. If SSM is indicative of seed energetic reserves, we also expected to observe positive relationships between SSM and survivorship and potentially between SSM and traits conferring greater seedling performance, specifically root length and leaf area. Finally, we expected crested wheatgrass, a species known to produce seed heads and seeds of higher specific mass than native bunchgrasses Hamerlynck and O'Connor 2021 ) to have higher survivorship than the two native study species.

Materials and Methods
Seed of the exotic bunchgrass, crested wheatgrass ( Agropyron cristatum L. var. Hycrest II; "AGCR"), and the native species bluebunch wheatgrass ( Pseudoroegneria spicata [Pursh] A. Löve; "PSSP") and Sandberg bluegrass ( Poa secunda Presl.; "POSE") was obtained from Maple Leaf Seed (Ephraim, UT); these commercially sourced seeds were cleaned of accessory structures. To capture a broad range of species-specific variation in seed mass, seed lots from each species were sorted into three size/weight classes using a custom-built vacuum-powered seed sorter. Thirty individual seeds were then randomly selected from each sorting class (30 seeds x 3 classes x 3 species = 270 total seeds) and then scanned on an Epson 10 0 0 0 Expression flatbed scanner calibrated for image analysis using WinRhizo v. 2.0 (Regent Instruments, Ste. Foy, Quebec, Canada) to determine individual seed projected surface area (mm 2 ). Scans were made at 1 200 dpi resolution for crested and bluebunch wheatgrass and 2 800 dpi for Sandbergs bluegrass. Seeds were then weighed to 0.0 0 0 01 g on a Mettler AT20 microbalance to measure individual seed mass (mg) and determine seed specific mass (mg mm −2 ).
Beginning on June 27, 2016, the selected seeds were inserted into ca. 12 mL of sterile agar in 100-mm diameter petri dishes. Dishes were oriented vertically, and seeds were planted below a notch cut from the receptacle and upper lid edges to allow aboveground growth ( Supporting Methods S1 ). Each dish was numbered, and seeds were randomly assigned to numbered dishes. Bundles of 10 dishes were wrapped in Parafilm (Bemis Company, Inc., Neenah, WI) and placed on trays in a growth chamber set to 15 °C and a 12-h photoperiod provided by fluorescent lights (Hoffman Manufacturing, Corvallis, OR). The lights were oriented on the back and front (i.e., inside of door) of the growth chamber to minimize any light gradient (Fig. S1, available online at …), and species were evenly distributed among bundles and racks within the growth chamber. After 2 wk, and every 2 wk thereafter, bundled dishes were removed, unwrapped, and assessed for condition (ungerminated, live, or dead). Any seeds that had not germinated by 6 wk were discarded and not used for the study. This resulted in three different-sized germinated seedling cohorts: 74 AGCR, 62 POSE, and 41 PSSP. Seedlings with any green aboveground tissue were considered live and scanned at 800 dpi on the flatbed scanner, and total seedling leaf area (cm 2 ) was determined from the images using WinRhizo. Before bundling and rewrapping, the agar of each dish was given a light misting of Nanopure deionized distilled water to keep the agar hydrated; any dishes that had fungal growth were wiped lightly first with a Kimwipe tissue and then lightly misted with a highly dilute chlorine bleach solution (1 capful bleach per 4 L Nanopure water). Dishes were kept within their initial bundle group and stored in the same growth chamber tray throughout the study, which ended 52 wk later in July 2017 after 26 total sampling periods.
Cohort survivorship was assessed using a Cox proportional hazards regression model in the R package survival ( Therneau 2022 ) to compare species-specific cumulative risk of a mortality event over the 52-wk study. For seedlings that died during the 52-wk study, differences in the timing of mortality were assessed using a Kruskal-Wallis nonparametric analysis of variance (ANOVA; R Core Team 2022 ) with week of observed mortality as the dependent variable; pairwise comparisons of species-specific mortality timing were made using a Wilcoxon rank sum test. Next, we quantified species-specific relationships between 1) seed mass and SSM and 2) seed mass and projected seed area using simple linear regression. We also used linear regression to quantify relationships among seed mass, SSM, and the two seedling traits of interest: projected leaf area and root length. We considered linear models both with and without species as a covariate to determine in- Table 1 Two-way analysis of variance F-tests comparing seed mass, seed-specific mass at the time of planting, and total seedling leaf area and root lengths of three bunchgrass species (AGCR, PSSP, POSE) that lived or died after 52 wk growth in sterile agar (end condition). * P < 0.05, * * P < 0.01, degrees freedom in parentheses following F-test result.  Figure 1. A, Cohort survivorship of crested wheatgrass (AGCR), bluebunch wheatgrass (PSSP), and Sandberg's bluegrass (POSE) seedlings germinated and growing in sterile agar. B, Box and whisker plots presenting variation around the median week of observed mortality for the three species; letters indicate significant differences between ranked week of mortality (Kruskal-Wallis one-way analysis of variance).
traspecific versus interspecific patterns. Finally, differences in seed mass, SSM, leaf area, and root length were assessed for surviving versus dead seedlings of each species using a two-way ANOVA ( R Core Team 2022 ). Leaf area and root length measurements from wk 4 of the experiment were used in this analysis because most initial growth occurred within 4 wk, and the majority of the three cohorts were alive at this point. F-test and α-adjusted post-hoc means testing (LSD) results were considered significant at an associated P value of ≤ 0.05.

Results
After 52 wk, the majority of AGCR (70.3%) and PSSP cohorts survived (58.5%), with a much smaller proportion of surviving POSE (24.2%; Fig. 1 a ). This represents a highly significant departure in proportional survivorship whereby the POSE cohort had 4.6x greater risk of mortality than AGCR and 1.6x greater risk than PSSP (likelihood ratio = 38.7; P < 0.01; df = 2). For seedlings that perished over the total study time, time to mortality was also species specific. The median date of observed mortality was soonest in POSE (16 wk) and then PSSP (28 wk), followed by AGCR (34 wk). PSSP mortality observations occurred throughout the study period, while POSE mortality occurred significantly earlier than AGCR mortality as indicated by the post-hoc comparison from a Kruskal-Wallis rank sum test ( χ 2 [2] = 8.96; P = 0.01; see Fig. 1 b).
Relationships between projected seed area and seed mass varied considerably among species; AGCR had the strongest positive relationship between seed mass and seed area ( R 2 = 0.71, F 1,72 = 176.9, P < 0.01), while seeds of POSE had the most negligible gains in seed mass as seed area increased ( R 2 = 0.35, F 1,59 = 31.6, P < 0.01) ( Fig. 2 a ). Seed mass by projected area was also statistically correlated for PSSP, but the correlation coefficient for this relationship was relatively weak ( R 2 = 0.13, F 1,39 = 5.9, P = 0.02). Species-specific differences in seed area between AGCR and PSSP resulted in clear differentiation in SSM, despite these species having similar seed mass (see Fig. 2 b). The correlation coefficient for seed mass versus SSM was similar among species ( R 2 = 0.39 -0.49), although the slope of this relationship varied among species (see Fig. 2 b). POSE had the most positive slope (F 1,59 = 56.3, P < 0.01), followed by PSSP (F 1,39 = 37.7, P < 0.01) and AGCR (F 1,72 = 46.1, P < 0.01).
Seed mass and SSM performed differently in their ability to predict wk 4 root length and projected leaf area. Seed mass best predicted root length within species, although the strength of this relationship varied among species. Root length by seed mass relationships were best correlated for the low seed mass species POSE ( R 2 = 0.24, P < 0.01, F 1,55 = 16.7), followed by AGCR ( R 2 = 0.18, P < 0.01, F 1,71 = 16.7); there was no statistically significant relationship between PSSP seed mass and root length. In contrast to seed mass, SSM best predicted root length across species, with a correlation coefficient of 0.28 (F 1,167 = 64.8, P < 0.01). Seed mass and SSM both explained species-specific variation in wk 4 projected leaf area, although overall model fit was much better for the seed mass model. Seed mass was positively related to leaf area for all species and explained 71% of variation in AGCR leaf area (F 1,72 = 176.9, P < 0.01), 34% of variation in POSE leaf area (F 1,58 = 31.1, P < 0.01), and 11% of variation in PSSP leaf area (F 1,39 = 5.9, P = 0.02). In contrast, SSM only explained variation in PSSP projected leaf area ( R 2 = 0.14, F 1,39 = 7.4, P < 0.01), and the two variables were negatively correlated. These results are illustrated in the supporting information (Fig. S2, available online at …).
The PSSP cohort had significantly greater seed mass (3.89 mg ± 0.11 SE) than AGCR (3.37 mg ± 0.08 SE) and POSE (0.67 mg ± 0.02 SE; LSD < 0.05; from two-way ANOVA, Table 1 ). Pooled across species, there were no significant differences in the seed masses at time of planting for survivors or seedlings that died before 52 wk, with no significant species x survivorship interaction (see Table 1 ; Fig. 3 a ; see Table S3 ). In contrast to seed mass, SSM was significantly higher in AGCR (0.58 mg mm −2 ± 7.51 × 10 −3 SE) than PSSP (0.39 mg mm −2 ± 0.011 SE) and POSE (0.18 mg mm −2 ±4.43 × 10 −3 SE; LSD < 0.05; from two-way ANOVA, see Table 1 ). Moreover, there was a significant two-way species-bysurvivorship interaction (see Table 1 ). For PSSP and POSE, planting- Figure 2. Linear regressions of seed mass versus seed-specific mass ( A ) and seed mass versus projected seed area ( B ) for the three study species. Model fit for each species is reported at the upper left of each panel, and the gray shaded area shows the 95% confidence interval around each fitted regression line. Seed mass and seed-specific mass are inextricably linked, and seed mass consistently explained approximately half of the observed variation in SSM for each species (all P < 0.01). In contrast, relationships between seed mass and projected seed area vary substantially among species with the strongest relationship observed for AGCR (all P < 0.05). time SSM did not differ between seedlings that died or survived over the 52-wk study period, while surviving AGCR seedlings developed from seeds with significantly higher specific mass at planting (see Fig. 3 b; Table S3 ). Seedling leaf area at 4 wk differed among species, with PSSP (0.61 cm 2 + 0.050 SE) and AGCR leaf areas (0.65 cm 2 ± 0.033 SE) greater than those of POSE (0.18 cm 2 ± 0.011 SE; LSD < 0.05 from two-way ANOVA, see Table 1 ). However, survivorship was not associated with differences in seedling leaf area (see Table 1 ), and there was no significant species-bysurvivorship interaction (see Fig. 3 c; Table 1 ; Table S3 ). In contrast to aboveground growth, seedlings that died before 52 wk had significantly shorter roots at 4 wk compared with seedlings that persisted over the course of the experiment (see Table 1 ; Fig. 3 d; Table S3 ). As with seedling total leaf area, root lengths at 4 wk differed between species pairs. AGCR had significantly longer roots (27.4 cm ± 1.98 SE) than PSSP (20.0 cm ± 1.74 SE), which in turn were significantly longer than root lengths attained by POSE (10.4 cm ± 0.76 SE; LSD Table 1 ), and there was no significant species by survivorship interaction (see Table 1 ).

Discussion
If seed mass reflected seed energetic reserves, we would expect to observe the greatest survivorship in the PSSP cohort under limited nutrient supply. However, we found that differences in seed mass were not necessarily associated with species-specific cohort survivorship. During the first 12 wk, all three species exhibited similar survivorship even though POSE seed mass was markedly lower at planting. However, POSE survivorship declined sharply as plants died and survivors initiated secondary growth (see Fig. 1 a). Although young seedlings are able to achieve autotrophy and net positive C balance within 5 d of imbibition under favorable conditions, the nutrient-poor agar growth medium appears to have restricted growth according to the amount of maternally supplied nutrients in the seed by about wk 14 ( McWilliam et al. 1970 ). Un-der this scenario, new growth is expected to yield diminishing returns due to increasing maintenance costs. Efficient nutrient resorption may reduce dependence on externally supplied nutrients, especially when resources are limited ( Killingbeck 1996 ;Luo et al. 2018 ). We hypothesize that long-term seedling persistence is influenced by both maternal investment in energy reserves within the seed and a seedling's ability to retain and recycle maternally provided nutrients to regrow photosynthetic tissue after die-back. In this experiment, we assume that since AGCR survivors had greater SSM than seedlings that died, this species received more maternal N and energetically richer investment than other species. Furthermore, although seedlings would receive a constant nutrient flux in a field setting, the advantages conferred by high SSM may also benefit seedlings under competition for resources. Follow-up studies are needed to investigate relationships between SSM and seed nutrient content, as well as the importance of nutrient-rich material for recovery following early seedling die-back.
Despite having similar seed mass, PSSP and AGCR had contrasting seed projected areas, which scaled to significant differences in SSM. This suggests that given a similar amount of space, AGCR can more efficiently package energetic reserves than PSSP. Higher SSM could reflect an optimized surface area to volume relationship and serve to enhance the effectiveness of mobilizing energetic reserves as the germinating seed imbibes water early in development ( McDonald et al., 1996 ). Unfortunately, we did not gather early seedling growth data at the fine temporal frequency required to determine if this drives early seedling growth and recommend that future studies consider tracking growth at a daily interval to determine if SSM influences aboveground or belowground growth rates. Increases in AGCR seed area were closely associated with increases in seed mass ( R 2 = 0.77; see Fig. 2 a and 2 b), whereas seed mass of the similarly sized native species PSSP varied in a less predictable manner as seed area increased ( R 2 = 0.13). The small seed mass species (POSE) had an intermediate correlation coefficient and lower slope, indicating that this species accumulates rel- Error bars indicate ± 1 standard error of the mean; letters differ significantly at P < 0.05 ( α-adjusted LSD; two-way analysis of variance). Means and standard errors are reported in Table S3 . atively less mass per unit gain in seed area. Seed size is affected throughout phenological stages, and overall maternal investment in seeds may be differentially affected by environmental conditions from early plant development, when total seed size and number are determined, through the grain filling stage, when storage compounds accumulate in the endosperm ( Sabelli and Larkins 2009 ). Seeds of similar size (i.e. , surface area or volume) may differ in seed mass due to differences in maternal environment during cell division when the total number of starch granules within the endosperm is determined ( Sabelli and Larkins 2009 ), or species may have genetic differences in their capacities for starch biosynthesis.
However, our knowledge of endosperm development in grasses is limited to species with agronomic value, primarily corn and wheat. Apparent differences in maternal provisioning to endosperm in rangeland grasses warrant future study, especially regarding fixed (e.g., genetic) versus plastic (e.g., maternal environment) components of starch biosynthesis in the endosperm.
Interestingly, the assumption of seed mass as a predictive functional trait may have appeared true if we had compared only POSE and PSSP. Under this scenario, the species with the greater seed mass at planting (PSSP) did better in the long term than the one with smaller seed mass, and SSM did not vary significantly be-tween individuals that died and those that survived (see Fig. 2 ). When AGCR was also considered, the timing of observed mortality and survivorship after 16 wk most closely mirrored speciesspecific differences in SSM (see Figs. 1 and 3 ). It remains unclear why the most dissimilar species in terms of SSM exhibited nearly identical survivorship during the first 12 wk of the experiment, although the rapid increase in POSE mortality after wk 12 suggests that seed resources became depleted shortly thereafter. Our experimental growth conditions may have also influenced the observed results; a recent study investigating seed mass-seedling growth relationships in bluebunch wheatgrass demonstrated that root traits were more strongly related to seed mass under cool germination temperatures than warm temperatures, suggesting environmental conditions may strengthen or weaken relationships between seed and seedling traits ( Mukherjee et al. 2019 ). Future studies considering SSM and germination under a range of temperatures may provide additional insights into how the germination environment influences relationships between maternal investment and early growth.
Patterns of tissue investment varied among species and between seedlings that died versus those that survived for 52 wk. Leaf area of the low SSM species POSE was significantly less than that of AGCR and PSSP, which were indistinguishable. Within species, seed mass was an excellent predictor of AGCR wk 4 leaf area and a poor to moderate predictor of leaf area for the two remaining species, while SSM was only statistically correlated with PSSP leaf area (see Fig. S2). The relationship between seed mass and early seedling leaf area most likely reflects the tendency of low seed mass species to have thin, high surface area tissues associated with rapid resource acquisition ( Larson et al. 2020 ). However, wk 4 leaf area was not related to intraspecific survivorship. In contrast, belowground allocation varied significantly for root length, which more closely aligned with species-specific survivorship (see Fig. 1 a) and differences in SSM (see Fig. 2 ). Within-species, root length at 4 wk for surviving seedlings was generally greater than root length of seedlings that died, but this difference was only statistically significant for AGCR (see Fig. 2 ). Overall, SSM was a slightly better predictor of root length than seed mass, although seed mass was able to explain intraspecific variation in root length for two of the species (see Fig S2). The different predictive power of these two related traits is intriguing and will require further experimentation to better resolve the relative utility of each as a predictive functional trait. Nonetheless, we suspect that larger roots were better able to sustain seedlings on the resource-limited growth medium (agar), as greater root investment is associated with mining for resources under stressful conditions ( Hallett et al. 2011 ).
Although germination success is typically high in sagebrush steppe bunchgrasses ( Boyd and Lemos 2015 ;Rigby et al. 2018 ), transitioning from germination to emergence poses an additional recruitment bottleneck for some natives ( Larson et al. 2015 ). Previous research has invoked drought and fungal pathogens ( James et al. 2011 ;Gornish et al. 2015 ; but identifying why so many seeds germinate and develop into seedlings yet don't make it through the emergence filter remains difficult. Our results suggest that native bunchgrass seedlings with low SSM (i.e., maternal energetic and nutrient reserves) may have constrained root development relative to seedlings from seeds with high SSM (see Fig. 2 b and Fig. S2). This aligns with the assumption that root construction strategies in resource-limited environments, such as xeric soils, require energetically expensive physiological traits to avoid/tolerate desiccation, therefore necessitating large seed size ( Leishman and Westoby 1994 ). AGCR has demonstrated higher reproductive photosynthetic performance compared with natives ( Hamerlynck et al. 2019 ; Hamerlynck and Ziegenhagen 2020 ; Hamerlynck and O'Connor 2022 ), and this seems to be associated principally with producing seeds of high quality, not necessarily filling the greatest number of seeds ( Hamerlynck and O'Connor 2021 ). This greater investment in high-quality seeds may provide a critical edge for seedlings to survive extended periods without ready access to soil resources, and it may explain AGCR's ability to establish, persist, and outcompete native rangeland bunchgrasses . A recent study linking plant functional traits to drought-induced community change identified root length as the best predictor of seedling mortality and emphasized the importance of identifying tissueand life stage −specific functional traits to accurately predict responses to changing climate ( Harrison and LaForgia 2019 ). Our finding that SSM is positively related to root length suggests a functional link between SSM and extended seedling survival, but future research controlling or experimentally manipulating maternal plant environment is needed to determine the strength of the relationships among SSM, root growth, and variation in seedling performance.
The seedlots used in this study came from locations throughout the intermountain west and Pacific northwest, which may have introduced confounding sources of genetic and environmental variation. AFLP genotyping of 565 P. spicata samples from localities throughout western North America revealed a population structure of 21 distinct P. spicata groups, and gene flow among these regional groups has likely been restricted since the last Glacial Maximum ( Larson et al. 2004 ;Massatti et al. 2018 ). Thus, differences in seedlot source and localities used in this study may have resulted in genetically constrained trait values. Similarly, maternal environment can affect offspring traits. For example, the offspring of barley (Hordeum vulgare) from a maternally drought-stressed population exhibited increased root-to-shoot ratios and enhanced growth of thin roots under drought conditions, relative to drought-stressed offspring of parent plants from optimum soil water conditions ( Nosalewicz et al. 2016 ). Regardless, our results obtained from a commercial seed source are in agreement with studies demonstrating higher SSM in crested wheatgrass compared with native bunchgrasses ( Hamerlynck and O'Connor, 2021 ) and are consistent with this species' ability to establish successful seed cohorts under conditions that strongly limit native grass populations . To better resolve uncertainties stemming from population structure and transgenerational effects, future studies should consider comparing seed-seedling relationships for a collection of seeds from shared environments versus across stress gradients and/or from distinct populations.

Implications
Identifying informative seed traits, such as SSM, may contribute to improved restoration outcomes by allowing for development of arid land bunchgrass plant materials better adapted to drought and other stressors ( Merino-Martín et al. 2017 ). This is because maternal energetic provisioning to seeds influences germination, emergence, and seedling recruitment. Our finding that SSM corresponded with root length and AGCR seedling survivorship supports the recent call for a more rigorous assessment of seed functional traits to improve their utility in predictive applied ecology ( Saatkamp et al. 2019 ). While the TRY Plant Trait database contains > 200 000 observations for seed mass, seed surface area and volume are reported far less frequently with no explicit mention of SSM; this is in stark contrast with widely reported values for massand area-standardized values of leaf and root traits ( Kattge et al. 2020 ). Developing drought-tolerant perennial grasses for restoration has proven a challenging task, especially through the lens of identifying functional traits that confer stress tolerance ( Garbowski et al. 2021 ). Sagebrush steppe restoration from seed is highly variable, and native species experience especially high failure rates ( Davies et al. 2011 ;Clements et al. 2017 ;Svejcar et al. 2017 ). Identifying seed and seedling traits that predict the ability to overcome early life stress, in combination with novel approaches like applied predictive genomic modeling, may offer a valuable way forward in native plant material selection for rangeland restoration ( Jones et al. 2010( Jones et al. , 2022.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.rama.2023.04.005 .