Lower dormancy with rapid germination is an important strategy for seeds in an arid zone with unpredictable rainfall

Seed germination traits are key drivers of population dynamics, yet they are under-represented in community ecology studies, which have predominately focussed on adult plant and seed morphological traits. We studied the seed traits and germination strategy of eight woody plant species to investigate regeneration strategies in the arid zone of eastern Australia. To cope with stochastic and minimal rainfall, we predict that arid seeds will either have rapid germination across a wide range of temperatures, improved germination under cooler temperatures, or dormancy and/or longevity traits to delay or stagger germination across time. To understand how temperature affects germination responses, seeds of eight keystone arid species were germinated under laboratory conditions, and under three diurnal temperatures (30/20°C, 25/15°C and 17/7°C) for 30 days. We also tested for decline in seed viability across 24 months in a dry-aging treatment (~20°C). Six of the eight arid species studied had non-dormant, rapidly germinating seeds, and only two species had physiological dormancy traits. Seed longevity differed widely between species, from one recalcitrant species surviving only months in aging (P50 = <3 months) and one serotinous species surviving for many years (P50 = 84 months). Our results highlight the importance of understanding the reproductive strategies of plant species in arid environments. Rapid germination, the dominant seed trait of species included in this study, allows arid species to capitalise on sporadic rainfall. However, some species also exhibit dormancy and delayed germination; this an alternative strategy which spreads the risk of germination failure over time.


Introduction
Seed traits and germination strategies drive plant community dynamics and provide insight into species' adaptations to environmental filters [1] and community composition [2]. Despite this, seed traits are under-represented in community ecology studies [3][4][5]. Knowledge  pulses [41]. Further empirical evidence is required from a greater suite of species (particularly perennials), and from a greater range of environments, to test the effect of rainfall predictability on seed traits and to determine the prevalence of dormancy in the arid zone. This study investigates the seed traits and germination strategy of eight Australian, arid zone species. The species selected are considered as keystone plants, as they are the dominant species and the only prevalent woody species in the ecosystem. We focused on traits that may be critical to species' persistence in arid ecosystems, and that are essential to understand if we are to successfully restore these species using seed. We hypothesise that seeds will be categorised by one of three strategies beneficial to recruitment in arid zones. To cope with stochastic and low rainfall, we predict that species adapted to the arid zone produce seeds that will either 1) have low seed mass but with rapid germination, across a wide range of temperatures, to ensure seedling establishment before soil moisture evaporates, 2) show improved germination under cool, winter temperatures where soil evaporation rates are lowest, or 3) have dormancy and/or longevity traits to delay or stagger germination, therefore spreading the risk of germination failure across time. Specifically, we measure seed dormancy and embryo traits, germination responses under different temperature regimes, and seed longevity under ambient aging conditions. We also test if seed mass, and other traits, are related to germination strategy and whether seed traits can be used as a proxy for germination strategy. We highlight important germination strategies of plants from an arid zone with stochastic rainfall, and discuss the evolution of seed traits that favour seed survival and germination when rainfall is infrequent and unpredictable. Wilson and Maireana pyramidata (Benth.) Paul G.Wilson. All species in this study are targeted for restoration after mineral sand mining. Vegetation communities prior to clearing include three major plant community types; shrublands dominated by M. sedifolia, woodlands dominated by C. pauper and M. platycarpum and woodlands with shrubland understory, dominated by C. pauper and understory shrubs M. sedifolia and M. pyramidata [42]. Less dominant tree species occur as small, scattered patches across the landscape and include H. tephrosperma, A. oleifolius and G. parviflora.

Study species and seed collection
The climate of the study area is arid (250 mm mean annual rainfall) although average annual rainfall can often fall below 200 mm for consecutive years. Temperatures range from 2˚C to 47˚C, with cooler mean daily temperatures from May to August (Fig 1A; [43]). Across 60 years of climate data, average monthly rainfall was 24 mm and, unlike most arid zones across the globe, there is no distinct wet season (Fig 1B; [43]).
Mature seeds of each species were collected from remnant populations adjacent to a mine site (33˚22'05"S, 142˚13'36"E) at which restoration of these species is planned, however, for species with infrequent seeding events (H. tephrosperma, A. oliefolius and A. rhagodioides), seeds were collected within a 200-km radius of the mine site. Seeds were either collected by hand, donated from Ecotypic Pty Ltd and Tronox Mining, or purchased from Ogyris Pty Ltd. Where possible, seeds were collected within the six months prior to testing. However, due to lack of seeding events, it was necessary to use seeds stored for over one year in some species (seed ages shown in Table 1). Seeds were manually cleaned and stored in paper bags, under cool dark conditions. Bracts and seed covering structures were removed prior to seed weight measurements and germination experiments.

Seed traits: Embryo type, mass, viability and imbibition
For each species, the ratio of embryo length to seed length (E:S) was calculated based on longitudinal dissections and measurements of 50 fully-imbibed seeds [44]. Embryo type was classified according to Martin [45]. Seed viability was assessed in a 1% solution of 2,3,5-triphenyl tetrazolium chloride (TZ), except for H. tephrosperma, for which seeds were incubated on moist filter paper at 30/20˚C due to consistently poor TZ stain results; they gave a germination response of 100%. Embryos that only partially absorbed stain were scored as weakly viable (Fig  2A, 2B and 2C) and were classified as non-viable seeds. To ensure accurate TZ interpretations, results were frequently compared to tests for viability by germinating seeds at diurnal temperatures at 30/20˚C, and according to the germination methods outlined below. Seed mass was  determined using the mean of three replicates of 100 seeds, with results then divided by 100 to represent weight (g) per seed. For imbibition tests, four replicates of 25 seeds were weighed, placed on moist filter paper, and incubated under three diurnal temperatures of 30/20˚C, 25/15˚C and 17/7˚C, under a 12 hr light/dark schedule. Seeds were weighed when dry at beginning of experiment and, after 5 min on moist filter paper, seeds were removed and patted dry with a paper towel to absorb surface moisture, and weighed. To determine increases in seed weight, seeds were re-weighed at 10 min, 30 min and at 1, 2, 3, 6, 9, 24, 48, 72, 96, 120, 144, 168 and 192+ hrs. All bracts and seed coverings were removed for imbibition, germination and viability tests, but they were retained for the longevity experiment as this is how seeds are currently stored after collection. For one species (G. parviflora), additional tests were performed in an attempt to alleviate dormancy, including the move-along method (as described by Baskin & Baskin [46]) and soaking in boiling water and 90% H 2 SO 4 , both treatments for 1 min, 30 min and at 1, 4, 12, 24 and 48 hours.

Germination responses under diurnal temperatures
Prior to germination treatments, seeds of all species were surface sterilised by soaking in 1% sodium hypochlorite for one minute, then rinsed for 40 seconds with double distilled water. For each species, four replicates of 25 seeds each were used. Seeds were placed in 90-mm diameter petri dishes on filter paper moistened with distilled water and incubated at a 12/12-hr light/dark regime at daily alternating temperatures of 30/20˚C, 25/15˚C and 17/7˚C). Seeds were incubated in cabinets (Thermoline Scientific, temperature and humidity cabinet, Model: TRISLH-495-1-s.d., Vol. 240, Sydney, Australia) under cool-white fluorescent lamps with a 40 μmol.m -2 photosynthetic photon flux. To determine the effects of gibberellic acid (a plant hormone herein referred to as GA 3 ) on seed germination, species were incubated at 30/20˚C and the filter paper was moistened with a 350 ppm GA 3 solution. To prevent microbial outbreak and ensure constant hydration during germination tests, seeds were transferred to sterilised Petri dishes weekly, on new filter papers moistened with the same appropriate water/GA 3 solutions. Seed germination (when the radicle emerged to at least half of seed size) was recorded daily for 30 days, or until germination ceased for four consecutive readings across all treatments. Dormancy type was classified according to Baskin and Baskin [46]. For two species that exhibited dormancy traits (G. parviflora and A. rhagodioides) an after-ripening treatment was applied by storing seeds for one year, under constant dark and air-conditioned temperatures between 10-20˚C and 45-50% humidity. An additional four replicate plates at each diurnal temperature were included for both of these species using the after-ripened seeds. Seeds for the treatments with and without after-ripening were from the same seed lot.

Longevity
Seed longevity was assessed to understand the relationship between longevity and dormancy classification. Seeds of species in this study are currently collected for restoration purposes and stored at the study location in air-conditioned shipping containers at~20˚C, hence we tested the effects of these aging conditions on seed longevity. Seeds were manually cleaned to remove excess organic matter from the seed batch, and stored in paper bags, under constant dark and air-conditioned temperatures between 10-20˚C and 45-50% humidity. Temperature and humidity of aging treatments were monitored and recorded once a fortnight for 24 months. To test for viability loss with aging, 100 seeds were extracted at 0, 3, 6, 12, 18 and 24 months, and TZ stained as per viability testing methods listed above. Due to seed shortages in A. oleifolius, only 60 seeds were tested for viability at each longevity test.

Data analysis
Imbibition was calculated using increase in seed weight after 72 hours: where W i is mean mass of imbibed seeds and W d is mean mass of dry seeds [47]. Viability of seed batches was tested in the days prior to experiment, and Viability-Adjusted Germination (VAG, herein referred to as germination) was calculated using the following equation [48]: For each replicate dish, the time to minimum germination (t min ) was taken as the first day that germination was observed, time to 50% germination (t 50 ) was the first day that germination was recorded at �50%, and time to maximum germination (t max ) was the first day at which the maximum germination was recorded. Mean t min , t 50 and t max were calculated from the four replicates of each species at each diurnal temperature.
Loss of seed viability with aging was calculated as: where V 1 is mean viability at day 0, and V 2 is mean viability at day of test. A Pearson test was used to test the correlation between seed weight and t 50 , longevity (P50) and seed fill, and between embryo type (ranked according to E:S) and t 50 . One-way ANOVAs were used to test the effects of diurnal temperatures on maximum germination and t max . Shapiro-Wilk tests confirmed the normality of both maximum germination and t 50 data prior to ANOVAs. Tukey HSD post-hoc tests were used to make multiple comparsions of means among diurnal temperature treatments. Welsh's t-tests were performed to compare germination between water and GA 3 treatents, and to test the effects of after-ripeing in A. rhagadioides. A generalised linear model with binomial error and a probit link function was fitted to the seed longevity data (i.e. loss of viability over time), and thus fit the viability equation [49]; where v was the viability after p months in aging, σ is the standard deviation of the normal distribution of seed deaths in time, and K i is the initial seed viability. An estimate of the time taken for seed viability to fall to 50% (P50) was calculated by solving for p when v = 50%. The Pearson test, Shapiro-Wilk tests, ANOVAs and GLMs were all conducted in R [50].

Results
The heaviest seeds had the longest germination times, hence there was a positive correlation between seed mas and t 50 (S1 Fig; P = 0.034), albeit based on only six species with germination data. However, there was no correlation between seed mass and longevity (P50: R = 0.44, P = 0.28), between seed mass and seed fill (R = 0.23, P = 0.59), or between embryo type and t 50 (R = -0.41, P = 0.42).

Viability and dormancy
Each species in the study showed high germination in at least one diurnal temperature treatment within two weeks and without pre-treatment, except for A. rhagodioides and G. parviflora. No germination was observed for G. parviflora at any of the diurnal temperatures tested, nor through further treatments to relieve dormancy (boiling water and H 2 SO 4 soaks, GA 3 , after-ripening and the move-along method). Hence the dormancy characteristics of G. parviflora were not determined, and no further germination results for this species will be presented. Seed fill was high for all species, except M. platycarpum, which was also excluded from germination studies due to lack of viable seeds. Imbibition studies demonstrated that seeds of all species readily take up water within 72 hours of wetting ( Table 1).

Effect of temperature on seed germination proportions
Temperature had no significant effect on maximum germination for C. pauper, M. sedifolia and A. oleifolius (P � 0.2 across all temperatures). However, maximum germination was significantly lower at the warmest diurnal temperature of 30/20˚C for M. pyramidata (P = 0.01) and to a lesser degree for H. tephrosperma (P = 0.045) and A. rhagodioides (P = 0.047; S1 Table).

Effect of temperature on seed germination rates
Non-dormant species germinated (t min ) rapidly and within six days in at least one diurnal temperature treatment (Fig 3). The two Maireana shrubs reached maximum germination quickest, within one to five days, while tree species were slower to germinate. Time to maximum germination (t max ) was significantly affected by temperature for all species except for C. pauper, with a general trend of decreasing germination times as diurnal temperatures increased (S2 Table). t max was significantly longer in M. pyramidata and A. oleifolius at 17/7˚C and 25/15˚C than at 30/20˚C (P � 0.014). The species with germination times least affected by temperatures were H. tephrosperma which had significantly delayed germination only at the coldest temperature (17/7˚C; P = 0.014), and C. pauper, which had no significant change in germination times across all temperatures (P = 0.25).

Germination responses to GA 3 and after-ripening
The only species with a positive germination response to GA 3 treatments was A. rhagodioides (P = 0.010; P > 0.5 for all other species). A. rhagodioides achieved t max in nine days with afterripened seed compared to twenty days with fresh seed (Fig 3B and S2 Table). Similarly, for A. rhagodioides maximum germination of after-ripened seed was significantly higher than for fresh seed (P � 0.04 for all temperatures), and it eliminated the effect of GA 3 treatments (P = 0.36) and widened temperature ranges for maximum germination, indicating dormancy loss.

Seed longevity
Myoporum platycarpum was the only species to exhibit characteristics of recalcitrant seeds, meaning they do not survive desiccation or aging. Seeds of M. platycarpum were freshly collected and most had no embryo, and of the few that had an embryo most were non-viable. Seed freshly picked from adult trees showed only 13% viability, which fell to half that within one month of aging, and was close to 0% viability within six months of aging (Fig 4). Most species showed a substantial decline in viability (>50%) within 12 months of aging, with the exception of H. tephrosperma, A. rhagodioides and C. pauper. These three species experienced less than 20% decline in viability within 12 months. The only species to show a decline in seed viability of <10% during 24 months of aging was H. tephrosperma. Seed longevity, or P50, for each species was (in order of longest to shortest lived, in months): H. tephrosperma, 84

Discussion
Arid species in this study were generally categorised by two types of adaptive strategies to facilitate seed germination in sporadic rainfall: 1) rapid germination across a wide range of diurnal temperatures, or 2) dormancy and/or long-lived seeds to temporarily delay, or stagger, germination. Our study demonstrates that rapid germination is a common, alternative and important strategy in seeds from arid zone species, which allows seeds to capitalise on sporadic rainfall. Seeds of three species (M. pyramidata, A. rhagodioides and H. tephrosperma) had significantly higher germination at cooler temperatures, than at 30/20˚C, hence avoid germination when evaporation rates are highest [51]. Germination inhibition at hottest temperatures is considered an important germination strategy for seeds with dormancy or longevity in this study. We also demonstrated the tendency for seed weight and embryo type to be correlated to germination rate of the species in our study. Conversely, seed mass was unrelated to longevity and seed fill. We recommend further seed-mass germination studies, using a greater suite of indigenous species from similar bioregions, to better understand the role of seed morphology in prescribing germination behaviour.

Rapid germination across wide diurnal temperatures
All non-dormant seeds of species in this study exhibited rapid germination rates, which suggests that rapid germination is particularly well suited to this environment and may be at a selective advantage over other seed germination strategies. For temperate species, seeds that germinate in less than four days are considered to have rapid germination [27] although, for annual species, rapid germination means germinating within 24 hours [52]. Considering all species in this study are slow-growing, perennial, woody species, we herein refer to rapid germination as species that germinate (t min ) within 4-6 days. All species, except for dormant seeds of G. parviflora and A. rhagodioides, commenced germination within six days upon wetting and displayed 50% germination in �9 days at the highest temperatures tested. Rapid germinating species are able to take advantage of single rainfall events, whereas slow germinators require several rainfall events persisting across days [39], which rarely occurs in the arid zone studied here [51]. All species achieved >50% germination across all temperatures tested, suggesting they have wide thermal ranges for germination. Wide thermal ranges for germination allow seeds to take advantage of stochastic rainfall events that occur during any season. While this may enable opportunistic plant recruitment, additional studies should quantify hydrothermal niches for seed germination to further understand recruitment timing across different seasons.
Germination strategies, particularly those that affect the timing of germination such as dormancy, may be important determinants of population dynamics in arid ecosystems. For example, vegetation classification in this arid zone prior to disturbance was Belah-Rosewood  Table 1. https://doi.org/10.1371/journal.pone.0218421.g004 Low dormancy with rapid germination for arid seeds Woodland and Belah-Bluebush Woodland [53], of which the dominant species include the non-dormant, fast germinators C. pauper and Maireana shrubs. The two species that have seeds with physiological dormancy, A. rhagodioides and G. parviflora, and non-dormant seeds of A. oleifolius, had the slowest germination speeds, are less dominant [54] and appear as scattered individuals throughout the landscape [53]. Although the age of seeds in this study varied and pre-storage components may have contributed to aging, longevity or dormancy loss in some species, reports of non-dormancy here are consistent with the findings of Callister [55] for all species, except M. platycarpum. Further studies are required to confirm dormancy in fresh seeds of C. pauper, H. tephrosperma and A. oleifolius. The lack of seeding events in these species during this study (potentially due to drought conditions) suggests that acquiring the quantities of seed required for their restoration may become more challenging under climate change.
Rapid germination appears to be unrelated to other morphological seed traits, except for seed mass and embryo type. All seeds in this study had fully developed embryos without an endosperm, or with embryos that are coiled and larger than the seed, and these are traits that are generally thought to indicate a rapid germination strategy [28,29]. Germination for seeds with peripheral embryos (including Atriplex and Maireana) is typically very fast because it involves merely the uncoiling of the spiral embryo upon imbibition, which ruptures the seed coat [56]. Many such species with peripheral embryos inhabit high-stress environments, where the rapid exploitation of temporarily favourable conditions for germination is more important [40]. Our results showed that the size and development of embryos was not a consistent predictor of germination strategy.
The species in this study with the fastest germination rates (C. pauper, M. sedifolia and M. pyramidata), are dominant species from the region, and produced small seeds that were easier to obtain due to frequent and prolific seeding events. High seed production requires high maternal input but the risk of population crashes are mediated because these species are less dependent on high seed survival rates [1]. Other studies report strong evidence for survival advantages associated with larger seeds under stressful conditions [57][58][59] but, considering the variable size of arid seeds and the trade-off associated to increased seed production in small seeded species, survival advantages of large-seeded species does not appear large enough to counterbalance the advantage of small-seeded species during seed production [11,60,61]. We found that small seeded species were amongst the fastest to germinate, however the negative relationship between germination speed and seed size was weak, suggesting that seed size may not always be a reliable proxy for germination rate. Future studies are required to test a larger number of species of a greater order of magnitude of seed mass variation, prior to making general assumptions about the relationship between seed mass, germination rate and success.
While the germination responses of M. platycarpum were not classified in this study, Callister [55] shows that it germinates to 50% within 4-5 days following wetting, consistent with other non-dormant trees species in this study. Callister et al. [62] suggests that seed-dormancy, not seed viability, is the key factor limiting germination in M. platycarpum. In contrast, as observed in our study, examining embryo firmness and colour through dissection microscopy is not an appropriate method to assess seed viability in M. platycarpum because TZ results show that healthy looking embryo are often dead (Fig 2F). Alternative viability testing, by assessing seed metabolic rate, may provide more accurate viability assessment (see results in [63]) and should be considered for future testing. Our results suggest that seeds of M. platycarpum are non-dormant and that seed viability and short-lived seeds are the key factors inhibiting germination.

Low prevalence of seed dormancy
Most seeds of plants from the arid zone have dormancy traits that enable germination to coincide with periods of highest water availability [46,64], yet only two species in this study (A. rhagodioides and G. parviflora) showed dormancy traits. These two species presented physiological dormancy traits as they readily imbibe water and have fully developed embryos, but failed to germinate to maximum potential within thirty days without treatment [46]. A. rhagodioides also showed a positive response to GA 3 before after-ripening. The dormancy cues of G. parviflora remain unresolved as all treatments failed to relieve dormancy, and there was a lack of viable seed available for further germination treatments. Most species from the same family as G. parviflora, the Rutaceae family, produce seeds with physiological dormancy and are commonly associated with complex germination requirements [46,[65][66][67]. Fewer species showed signs of dormancy than was expected. There is potential that due to the age of seeds used for this study, some of the study species may have exhibited dormancy traits as fresh seed. We have classified four such species (A. oleifolius, C. pauper, H. tephrosperma and M. pyramidata) as non-dormant, and we do not believe there would be significant levels of dormancy in the fresh seed of these species. Callister [55] provides supporting evidence for this assertion for C. pauper and A. oleifolius. Local nurseries (including The Seeds of South Australia Database) and seed practitioners also support the claims of non-dormancy in these species (I. Sluiter, A. Quamby, T. Langdon, pers. comm. 2018), and there is evidence of non-dormancy in congeneric species within the genus [46]. Traits that increase the likelihood of germination coinciding with periods of highest water availability may be less important in environments with unpredictable wetting pulses.
Delayed germination through dormancy has been observed in many arid zone species [30][31][32]. Similarly, for A. rhagodioides, a small proportion of seeds germinate upon maturity, or upon seed dispersal, but germination rate and proportion improve as seeds age. This suggests that seeds can remain in the soil or canopy for years and stagger germination across seasons, which spreads the risk of recruitment failure through time and increases the probably that favourable conditions for seedling establishment will occur during the lifespan of a seed cohort [30,68]. Physiological dormancy in A. rhagodioides was relieved through a period of after-ripening, a trait also reported in other Atriplex species [69,70]. After-ripening also enables a wider temporal window for germination which suggests that germination opportunities increase as seeds after-ripen. Developing a short-term soil seedbank, before the onset of suitable rainfall events, is an important adaptation in response to the unpredictability of resource availability in arid ecosystems. However, A. rhagodioides and G. parviflora were the only dormant species in this study, and non-dormant and prolifically-seeding trees are the most dominant species of the region [71]. This again suggests that in arid zones without distinct wet seasons, playing it safe through dormancy may be less important than previously assumed.
The literature on bet-hedging predominately investigates annual species with short lifecycles (i.e. [72,73]), whereas this study focuses on long-lived perennial species. Bet-hedging is more likely in annuals because the consequences of death of entire seed cohorts are greater [72,73]. Comparative studies indicate that seed dormancy is higher in annuals than in perennials [74][75][76], although very few studies investigate bet-hedging in perennial shrubs [76,77], and no such studies exist for arid tree species. Perennial species (including C. pauper, M. platycarpum and M. sedifolia) can have many flowering and seeding events throughout their lifespan. Thus, losing an entire seed cohort is of little consequence, reducing the selection pressure to maintain a soil seedbank [78,79]. Additionally, C. pauper [80], H. tephrosperma [80], M. platycarpum [81] and A. oliefolius [82] have the capacity to regenerate via root suckers, which may further mediate the impact of losing entire seed cohorts for consecutive years, and possibly negate the need for dormancy. Our study supports the notion that perenniality selects against seed dormancy [74,83,84] however, we only tested dormancy in eight species and studies from other arid zones report dormancy in many perennial species that seed frequently. Therefore, we recommend further research testing the prevalence of dormancy for long-lived species that fruit prolifically.
The ability to delay seed release, through serotiny, may further mediate the likelihood of dormancy and bet-hedging in H. tephrosperma that has non-dormant seeds with exceptional seed longevity. Through delayed seed release, seeds of H. tephrosperma can likely persist is the canopy for many years following seed maturity, as reported in congeneric Hakea species [85]. Serotiny levels in H. tephrosperma have not been investigated to-date, although only one occurrence of seed release was observed in populations from the study site during the three years of this study (personal observation). However, the level of serotiny in Hakea is positively related to follicle mass [85], and hence the large and woody follicles of H. tephrospherma suggest it is strongly serotinous. While serotiny in Proteaceae is considered beneficial to surviving in fire-prone landscapes, it is also observed in fire-free landscapes [86] or, like this study site, where small bushfires occur at interdecadal scales. We suggest that seed release in H. tephrosperma is triggered by seasonal temperatures and/or humidity, rather than fire, and is a crucial survival strategy for this species because it is non-dormant and flowers infrequently when rainfall is limited.
Low serotiny and bet-hedging (with the exception of H. tephrosperma and A. rhagodioides) in species in this study suggest a strong selection for rapid flowering and seed dispersal in response to rainfall. For example, the two Maireana species in this study can flower at any time of the year, but usually do so following large rain events [79,87]. Furthermore, their seeds are encircled by a papery wing-a trait adapted for dispersal away from the adult plant by wind soon after it has matured, or when dehiscing. The two non-dormant tree species (C. pauper and M. platycarpum) also flower prolifically, have seeds that are rapidly dispersed by wind, and have no apparent adaptations for serotiny or seed banking. Conversely, fruiting in A. rhagodioides and H. tephrosperma may not occur every year, but they have the capacity to recharge their seed bank through flowering and fruiting during favourable periods. Thus, because dispersal ability, serotiny and bet-hedging all function to avoid the risk of seedling emergence during unfavourable conditions, they may substitute for one another so that selection for one may weaken selection for the other strategy [88][89][90]. For species that fruit irregularly with no obvious adaptations for delaying seed release and germination (A. oliefolius and G. parviflora), there will be limited recruitment opportunities with increasing aridity and rainfall variability. As such, we have concerns for the persistence of these species under climate change.
Seed longevity is an alternative strategy to spread the risk of reproductive failure across time, and may further reduce the requirement for dormancy and bet-hedging [75,83]. Seed longevity is also critical for species with dormancy traits, or serotinous species with canopystored seed, because seeds may need to survive multiple seasons. Our results are consistent with other studies that have found serotinous species typically have orthodox seeds that are long-lived [91], or tolerant to desiccation. However, longevity results in this study may only reflect seed-aging behaviour during ex situ storage and further studies should test seed persistence in situ to understand soil-seed banking dynamics. Species from arid environments are more likely to have comparatively long-lived seeds than those from cooler, wetter ecosystems [92]. Nonetheless, most species in this study had significantly higher longevity than expected [91][92][93] and we suggest that high seed longevity is a key adaptation to cope with the unpredictability of rainfall at the study site. Perennial plants that invest in seed longevity as a survival strategy increase their probability of encountering conditions favorable for seedling establishment [94,95], and our results support the notion that longevity can negate the need for dormancy. In summary, a perennial life history and long-lived seed together reduce the likelihood of bet-hedging for the species in this system, where dominant perennial plants appear to invest less in seed dormancy and bet-hedging, and instead select for seed longevity and increases to offspring numbers.

Conclusion
We discuss seed morphology and physiology, and the germination behaviour of seeds that may facilitate seed survival and growth in arid zones. Most species exhibit rapid germination across wide diurnal temperatures, often producing prolific quantities of seed, which enables species to take advantage of rainfall events that fall across all seasons. Fewer species tend to avoid unfavourable conditions by delaying germination through seed dormancy and possibly bet-hedging. High seed longevity under ex situ aging was observed in most species included in this study. While seed dormancy is an important survival strategy for an arid seed, rapid germination and high seed production rates may be alternative regeneration traits and have a crucial role in explaining population dynamics in arid ecosystems with unpredictable rainfall. In such systems, the benefits of rapid germination may outweigh those associated with delayed germination through dormancy, and the population structure of remnant ecosystems in the region (which are dominated by rapid germinators) may be a testament to this.

S1 Fig. Pearson correlation between mean seed weight and mean t 50 .
(TIF) S1 Table. Mean values for maximum germination for each species under three diurnal temperatures. Maximum germination is also shown for the GA 3 treatment (when diurnal temperature is 30/20˚C), and after the after-ripening treatment for A. rhagodioides. Letters indicate the results of Tukey pairwise comparisons among the three diurnal temperature treatments. Treatments that share a letter are not significantly different from each other. For the GA 3 and after-ripening treatments, asterisks represent significant differences compared to the control treatment. For the 'GA 3 + after-ripening' treatment, the comparison is to the GA 3 only treatment (n.s. = not significant; � 0.05 > p > 0.01; �� 0.01 > p > 0.001; p < 0.001). (DOCX) S2 Table. Time to minimum, 50% and maximum germination of seeds incubated at 30/ 20˚C, 25/15˚C and 17/7˚C (± standard error). Atriplex rhagodioides (AR) refer to seeds rendered non-dormant through a 12 month after-ripening. (DOCX)