Thermal germination niches of Persoonia species and projected spatiotemporal shifts under a changing climate

Seasonal germination is critical in synchronizing seedling emergence with optimal conditions for survival but will be disrupted by climate change. Understanding how germination of threatened species with complex dormancy patterns will be affected by climate change is a priority for their management. By exploring the effects of temperature on germination for six Persoonia species ranging in rareness, this study aims to improve ex situ plant production and better understand the probable impacts of climate change on persistence of local populations.

Persoonia (Proteaceae) are endemic to Australia, with the greatest speciation occurring in south-eastern Australia, a region where vegetation clearing and habitat fragmentation is common. Persoonia mostly occur as part of sclerophyllous plant assemblages associated with sandstone-derived soils (Weston, 2003). In these nutrient-poor environments, they produce carbohydrate-rich fruits, nectar and shoots that are a food source for animals (Cannon, 1984;Rose, 1973;Snow, 1981). Nine of the approximately one-hundred Persoonia species in south-eastern Australia are threatened, endangered or critically endangered (IUCN, 2017;Unit, 1999). A number important populations of listed Persoonia species are currently threatened by clearing for mining and development (DEE, 2008;DEC, 2006;OEH, 2012). Yet, an inability to cultivate sufficient plant numbers for reintroduction, either by propagation or seed, has so far excluded Persoonia from restoration programmes (Bauer, Johnston, & Williams, 2004;Cambecedes & Balmer, 1995;Mullins, Koch, & Ward, 2002).
The pyrene or "seed" of Persoonia consists of an embryo with no endosperm, encased in a thick, woody endocarp, which is relatively durable compared to other species (Norman & Koch, 2008).
The endocarp is water permeable, but exerts physical dormancy, with germination occurring after its removal or weakening (Bauer et al., 2004;Chia, Sadler, Turner, & Baskin, 2016;Norman & Koch, 2008 (McKenna, 2007;Nield, Monaco, Birnbaum, & Enright, 2014), but climatic factors and seed burial also affect germination (Chia, Koch, Sadler, & Turner, 2015;Chia et al., 2016;Nancarrow, 2001). Seed bank longevity remains largely undefined with some indication that soilstored seeds of P. pinifolia have a half-life of approximately a year (Auld, Keith, & Bradstock, 2000), highlighting the importance of regularly occurring suitable germination conditions regardless of fire events. Sexual maturation (>7 years) and time to peak seed production (>12 years) are long compared to co-occurring species in pyrogenic communities, indicating that average inter-fire period >7 years is required for population persistence via seed production (Auld, Denham, & Turner, 2007;Auld & Ooi, 2008). Long juvenile development but short seed bank persistence suggests that inter-fire germination (occurring independent of fire events) is important in buffering population regeneration during periods of both infrequent and frequent fires. Temperature as an isolated germination signal remains unknown for many species in this genus, and identification of optimal temperatures could improve ex situ germination, predict impacts of future climates and assist in prioritizing conservation efforts.
In this study, we aim to characterize the relationship between germination probability and temperature for six significant and rare Persoonia species (Table 1). We do this by subjecting recently dispersed, dormant Persoonia seeds to ranging temperature regimes using thermogradient tables. We then modelled the probability of germination against average day temperatures (X day ), average night temperature (X night ), and the difference between day and night temperatures (X day − X night or DT δ ) using generalized additive models. Using predicted germination responses and continental-scale surface temperature models, we predict the germination niche for each species under historical  conditions as well as quantify the spatial and temporal shifts under projected climate scenarios (2080). We discuss results with respect to improving and informing current and future conservation efforts for three rare Persoonia species included in this study and provide insights into the challenges faced by the Persoonia genus under future climate conditions.

| Fruit collection
Seeds of six Persoonia species were collected from sites in New South Wales, Australia ( Figure 1). These species range in rareness (Table 1) and were chosen because they are currently effected by mining and development, yet are not included in restoration projects due to low levels of ex situ germination. Following the guidelines for seed collection of New South Wales threatened plant species, we collected <5% of available seeds in each population (Offord & Meagher, 2009).
These species were thus also chosen as they produced large enough fruit sets for collection of sufficient seeds without endangering wild is listed as endangered and grows at 5-158m ASL in the alluvial soils of the Hawkesbury River floodplains of Western Sydney (National Herbarium of NSW, 2018). It's geographic range is now significantly fragmented, and the majority (roughly 95%) of the population is restricted to an area of <2 km 2 (DEC, 2006;Robertson, Matthes, & Smith, 1996). P. pauciflora (P.H. Weston) is critically endangered and occurs at two sites (49-95m ASL); however, one population was recently cleared for development (OEH, 2012). P. levis ((Cav.) Domin), P. pinifolia (R.Br.) and P. linearis (Andrews) are all commonly occurring species with broad, overlapping distributions that range in altitude ( National Herbarium of NSW, 2018). All six species rely to varying degrees on the formation of a soil-stored seed bank to regenerate following disturbance, but P. acerosa, P. nutans, P. pauciflora and P.

| Seed preparation
We used x-ray imagery, tetrazolium staining, cut tests and germination experiments to quantify the viability of the collections and TA B L E 1 Conservation status from the Environmental Protection and Biodiversity Conservation Act (Commonwealth of Australia 1999): critically endangered (CE), endangered (E), vulnerable (V) and not listed (NL), disturbance strategy (Benson & McDougall, 2000): facultative resprouter (FR), obligate seeder (OB) and flowering time of the six studied Persoonia species (Benson & McDougall, 2000).   (Daws, Garwood, & Pritchard, 2006;ISTA, 1985;Leist, Krämer, & Jonitz, 2003). Using these methods, we found that for maximum seed viability, Persoonia fruits must remain attached to the parent plant until fully matured, signified by natural abscission.
We placed mesh bags around developing fruits and then returned  Table 2). For P. acerosa, P. nutans and P. pauciflora, fewer than five populations had a large enough seed bank to tolerate collection of seeds. In order to obtain sufficient seeds for germination experiments, particularly from these rare species, we pooled seeds from all collection sites for each of the six species, assuming seeds were representative of seed and climatic variability.

| Experimental preparation
The mass ratio of seed coats to seed is a measure commonly used to predict tolerance of desiccation and orthodox storage conditions (<15% humidity ~ 4°C) (Daws et al., 2006;Gold & Hay, 2008;Hong et al., 1996). Using a routine seed coat ratio test (Appendix 1), seeds of Persoonia species were found to be desiccation tolerant, indicating that viability is maintained with orthodox storage (Daws et al., 2006;Gold & Hay, 2008;Hong et al., 1996;International Seed Testing Association, 1985). Fruits were spread on trays, stored at 15% humidity and 15°C for up to six months.
Germination trials were run on a feasible subset of fresh seeds and compared to stored seeds (Baskin, Thompson, & Baskin, 2006), for which germination rates did not differ significantly when compared to seeds post-storage. We acclimatized seeds to laboratory conditions for 24 hr before initiating the germination protocols.
The seeds, still encapsulated in fruit, were soaked in water with Ultrazyme ® (0.3 g/L) for approximately 24 hr before separating the mesocarp from the pyrene, consisting of a woody endocarp encasing an embryo (Tieu, Turner, & Dixon, 2008). While intact, the fleshy mesocarp and woody endocarp exerts a mechanical constraint restricting moisture to the seed (Chia et al., 2016) and

TA B L E 2
Mean annual (± standard deviation) temperature, diurnal range in temperature, rainfall (Hijmans et al., 2005) (McIntyre, 1969;Mullins et al., 2002;Nancarrow, 2001). Taking into account that GA3 may enhance germination, it was included to isolate and explore the role of temperature on germination, with the assumption that all other conditions required for optimal germination have been met.
In order to extrapolate realistic germination responses from laboratory based studies, any pre-treatment of seeds must resemble preparatory processes for germination in natural contexts (Baskin et al., 2006). In line, each of these steps does so; the application of Ultrazyme ® representing the natural breakdown of fruit through ingestion by animals or microbes; the removal of the endocarp resembling the breakdown of the endocarp through repeated wet and drying or microbial degradation (Chia et al., 2016); and the application of GA3, resembling the build-up of naturally occurring growth hormones in order to optimize germination conditions. Likely required conditions, including the provision of natural growth hormones, such as GA3 that stimulate germination, moisture and removal of physical dormancy in the form of the endocarp, were therefore met to isolate and observe the effects of temperature on germination (Mullins et al., 2002;Norman & Koch, 2008).

| Germination
For each species, 630 embryos were sown individually into sterile glass vials with 0.7% agar media substrate infused with gibberellic acid, except P. pauciflora, for which 357 embryos were sown due to a smaller seed collection. Each species was tested on a bidirectional thermogradient plate (Grant Instruments, UK, model GRD1; Figure 2). Along both gradient axes, temperatures ranged from 8 to 45°C. Photoperiods alternated between light and dark every 12 hr coupled with a reversal along one temperature gradient axis.
The thermogradient table was partitioned into 36 compartments of 9 cm 2 , and average light (X day ) and dark (X night ) temperatures characterized each compartment with a unique thermo-photoperiod. The temperature within each compartment was recorded by a Thermochron iButton every 10 min throughout the duration of the experiment and used to calculate X day and X night temperatures. The 12-hr light/warmer and 12-hr dark/cooler regimes simulated diurnal conditions over 24-hr cycles (Cochrane, 2019;Fernandez-Pascual et al., 2015). Available testing equipment is incapable of simulating a more realistic, gradual ramping up and down of temperature and light in true diurnal patterns, possibly affecting extrapolation to ecological contexts. Half of the table had relatively warmer dark periods and cooler light periods and were therefore excluded, leaving 21 active thermo-period combinations. Aside from simulating an ecologically implausible scenario, this also minimized our use of seeds from rare species. Seeds were set on top of firm agar medium, exposed to air temperature for the most part. For each species, 30 vials, each containing one embryo, were placed within each compartment, except for P. pauciflora, for which 17 vials were used in each compartment. Germination was recorded every two or three days, and the test was terminated after at least two weeks of no recorded change, but the experiment was continued for a minimum of 89 days for comparability between species.

| Statistical analysis
We used generalized additive models (GAMs) to examine associations between germination probabilities and the DT∆ vectors of X day , X night and the range of difference between them (DTδ), with each model postulating these vectors as singular or additive explanatory variables. GAMs are a non-parametric extension of a generalized linear model (GLM) that do not require pre-specification of non-linear relationships. In GAMs, predictors depend linearly on unknown smoothing functions of some of the covariates (Hastie & Tibshirani, 1990). The degree of smoothing was selected by minimum GCV/ UBRE scores, controlling for over-fitting with a gamma multiplier of 1.4 (Wood, 2011). To avoid over-fitting and ease of interpretation, the number of knots (i.e. polynomial level) was limited in the GAMs to three. As third-order polynomial models, they are sufficient in contouring to three-dimensional response variables. While being flexible, they are restrained as additives, allowing for conventional regression analysis (Hastie & Tibshirani, 1990;Yee & Mitchell, 1991).
We used the "gam" function in the "mgcv" package with a binomial link function (Wood, 2017) in the R environment (R Development Core Team, 2014).
Prior to modelling, we evaluated the correlation between the three explanatory variables (X day , X night and DT δ ) and those to be less than r = .5. We developed a unique model for each of the six species, only considering possible main effects of the three potential predictors without interactions. We examined two response package (Barton, 2018). The corrected Akaike's information criterion (AICc) was used to identify plausible models having AICc scores not greater than two relative to the lowest score (Appendix 2; Burnham & Anderson, 2002;White & Burnham, 1999). We then used a model averaging approach, weighted by model AICc scores, to quantify the average strength of association between the two response models, that is, germination probability and number of seeds, and the three explanatory variables. We also evaluated the fit of averaged models by calculating Efron's pseudo R 2 (Efron, 1978).

| Current and future germination predictions
Soil temperature models do not currently exist given the complex interactions between soil and air temperature (Parton & Logan, 1981;Zheng, Hunt, & Running, 1993); therefore, in concurrence with our experimental design, we used modelled continental-scale air temperature to predict germination-temperature responses (Cochrane, 2019;Fernandez-Pascual et al., 2015). Using modelaveraged responses of binomial models, we predicted the geographic distribution of seed germination probability (Pgerm) under historic  conditions using monthly X day , X night and DTδ

| RE SULTS
Germination niches for the six Persoonias were species-specific, presenting complex responses across diurnal regimes. In the experiments, observed optimal seed germination probability for all species occurred with average day temperatures of ≅ 22°C, except for P. acerosa, which had the highest germination probability at average day temperature of ≅ 15°C (Figure 3). Observed optimal night temperatures, however, varied between all the species. Explained variance by binomial individual seed germination models was relatively low and varied among species, ranging between 17% (P. acerosa and P. levis) and 44% (P. linearis; Table 3). When modelled at a population scale, as the number of seeds germinated per cell, explained variance was higher, ranging between 63% (P. pauciflora) and 86% (P. pinifolia; Table 3).
Models predicted declining germination probabilities for all of the six species in response to warmer day or night temperatures, and three (P. levis, P, nutans and to a lesser extent P. pauciflora) species were also sensitive to cooler day or night temperatures. Cooler X day and X night temperatures were most favourable for P. acerosa, which linearly declined with warmer temperatures (Figure 3a). The responses of P. nutans and P. pauciflora, respectively, were driven by convex negative associations with X night and X day temperatures ( Figure 3d,e). P. linearis, P. levis and P. pinifolia had convex negative germination probabilities responses to warmer X day temperatures, generally displaying linear negative associations with warmer X night (Figure 3b,c,f). Declines in germination probabilities driven by cooler or warmer temperatures are reflected in both temporal and spatial predicted outcomes. For example, declines with warmer temperatures resulted in peak P germ during winter months and reductions in spatial germination probabilities under warmer conditions, while germination sensitivities driven by cooler temperatures resulted in the opposite outcomes (Figure 4, Figure 5).

| Temporal models
Germination models predicted likely temporal shifts and reductions in germination probabilities in summer months under the three RCPs for five species, and slight increases in germination for P. nutans across the year (Figure 4). Relative temporal and spatial shifts were associated with negative linear responses for P. acerosa, with peak germination probability during February (summer) decreasing with higher RCP climate scenarios, and no significant change during the winter months (Figure 4a). Warmer climate scenarios also resulted in temporal shifts for P. pauciflora, P. linearis and P. levis, with reduced germination probabilities under warmer temperatures causing a contraction of peak germination into the cooler months, while germination probability for P. levis increasing notably in cooler months with warmer temperatures. P. pinifolia maintained an historic germination preference for cooler months; however, a slight reduction in germination probability with warming temperatures was still observed. Predicted sensitivity of P. nutans to cooler temperatures relates to the lower germination probabilities in winter months and a slight increase under warmer climate change scenarios (Figure 4d).

| Spatial models
Across bioregions, spatial shifts in germination probabilities largely aligned with species-specific temperature sensitivity ( Figure 5, Table 3). For P. acerosa, P. pauciflora and P. linearis, the effects of warming temperatures were reflected in consecutively predicted declines in germination probabilities under climate change scenarios ( Figure 5, Appendix 3, Table 3). In these instances, areas supporting the highest 40% germination probabilities decreased between 90% (RCP 8.5) and 56% (RCP 2.6) for P. acerosa, 33% (RCP 2.6) to 60% (RCP 8.5) for P. linearis, 60% (RCP 2.6) to 98% (RCP 8.5) for P. pauciflora and 33% (RCP 2.6) to 71% (RCP 8.5) for P. pinifolia (Appendix 3). In contrast, the spatial extent of high germination probabilities of P. nutans considerably increased with warmer RCP scenarios with the highest 40% increasing by 50% under RCP 2.6 and by 240% under RCP 8.5 (Figure 5d), while those of P. levis did not significantly change (Figure 5b, Table 3). When protected areas were considered, possible areas of high germination probabilities of P. nutans and P. pauciflora were further restricted, while P. acerosa, which is already largely restricted to protected areas, while the ranges of the three common species were fragmented and restricted to varying extents (Appendix 3, Table 3).

| D ISCUSS I ON
Germination in the six studied Persoonia species was found to display unique temperature sensitivity and optima, with distinct predicted responses to climate change. Under examined RCP scenarios, germination probabilities for five species were predicted to decline, raising concerns over the long-term viability of many populations and F I G U R E 3 Persoonia germination probabilities in response to average diurnal temperatures represented in three-dimensional plots and two-dimensional response curves (X day (red), X night (blue), DTδ (green) with 95% confidence intervals (dashed) based on averaged generalized additive models highlighting the need to prioritize conservation management of remaining populations. As habitat destruction is one of the biggest limitations to species' adaptive capacity and resilience to climate change (Travis, 2003), protecting remaining habitat and identifying suitable translocation sites within protected areas are critical to improve the likelihood of survival of Persoonia, particularly for those already listed as threatened.

| Implications for long-term viability
Rare or threatened species with distributions restricted naturally by altitude, small ecological niches or fragmented populations due to human activity are the most vulnerable to climate change and often the first to disappear (Abbott, Doak, & Peterson, 2017;Bellard, Bertelsmeier, Leadley, Thuiller, & Courchamp, 2012;Dirnböck et al., 2011;McLaughlin, Hellmann, Boggs, & Ehrlich, 2002;Parmesan, 2006). More the 90% of P. nutans occurrences are currently restricted to an area of <2 km 2 (DEC, 2006; Robertson et al., 1996) with no juveniles or seedlings found in this population. In addition, no recent regeneration was observed in P. acerosa nor in P. pauciflora populations which are now entirely restricted to within a few km 2 (OEH, 2012). Evident regeneration failure suggests a potential extinction debt, a time-delayed but inevitable extinction, occurring when unsuitable environments TA B L E 3 Explained variance (R 2 ) of germination models and predicted spatial extents of Persoonia germination probability within biogeographic subregions ("Range") (Geoscience Australia & Department of the Environment and Energy, 2018) and within protected areas (PA) (CAPAD, Commonwealth of Australia, 2017aAustralia, , 2017bAustralia, ) under historical (1970Australia, -2000 and future (2080) RCP 2.6, RCP 6.0 and RCP 8.5 emission scenarios (Hijmans et al., 2005) Species P germ Historic (km 2 ) RCP 2.6 (km 2 ) RCP 6.0 (km 2 ) RCP 8.

F I G U R E 4
Annual temperature plots of Persoonia germination probabilities and 95% confidence intervals (dashed) for each species as a response to monthly X day and X night under historical , RCP 2.6, RCP 6.0 and RCP 8.5 scenarios in the bioregions relevant to current distributions for each species arrest recruitment but maintain extant individuals (Tilman, May, Lehman, & Nowak, 1994), often identified in fragmented populations of species (Nield et al., 2014). Within the potential ranges of P. acerosa and P. pauciflora, germination probabilities were predicted to decrease under climate change, further enhancing recruitment deficits. In contrast, germination probability of P. nutans was predicted to increase within its potential range with warmer temperatures. This may initially increase much needed recruitment and help to address the extinction deficit in the currently ageing P. nutans populations. However, higher rates of germination could also destabilize populations in the long term. For instance, warmer temperatures may illicit higher germination, but are also expected to decrease seedling survivorship (Walck et al., 2011).
Fire frequency is also expected to increase with a warmer and dryer climate, and with long (7-12 years) sexual maturation periods, Persoonia species, particularly obligate seeders like P. nutans, rely on ungerminated, residual soil seed banks to persist through reoccurring fire events within the period of sexual maturation.
A possible cause of regenerative failure in P. nutans as well as P.
acerosa and P. pauciflora, however, may be due to lack of recent fires in many areas. While some populations of P. acerosa experienced recent fires and still show no signs of recruitment, most populations of these three species occur proximal to housing and infrastructure where natural fire regimes are suppressed. The role of fire in the germination of Persoonia species is therefore complex and requires more research.

| Compounding threats
Continuing habitat and population destruction is a major threat to the persistence of all three rare and threatened species (DEE, 2008;DEC, 2006;Robertson et al., 1996), decreasing their adaptive capacity by depleting the existing gene pools directly through destruction of individuals as well as by impacting genetic mixing of remnant populations by limiting dispersal between patches (Higgins & Richardson, 1999;Lowe, Boshier, Ward, Bacles, & Navarro, 2005). Furthermore, Persoonia fruit is known to be ingested by animals, and although the importance of zoochory for germination has not been established, animal ingestion is considered important for dispersal (Mullins et al., 2002;Nield, 2014;Rose, 1973). Habitat destruction will therefore not only result in a direct decrease in colonizable habitat and impact fire and rainfall regimes in remnant ecological systems, it will also disrupt dispersal mechanisms and the possible effects of zoochory on germination indirectly by impacting animal populations (Gill & Williams, 1996; F I G U R E 5 Density plots of spatial units (cells) representing spatial shifts in Persoonia germination probability within bioregions of each species under historical (1970-2000 monthly averages), RCP 2.6, RCP 6.0 and RCP 8.5 scenarios Nilsson & Berggren, 2000). As suitable habitats become scarcer and existing populations dwindle, adaptive capacity will continue to decrease while extinction debt increases (Tilman et al., 1994;Travis, 2003;Trent, 2013

| Biological limitations
Persoonia are notorious for low levels of germination, as observed in our study, for a number of possible contributing factors (Bauer et al., 2004;Cambecedes & Balmer, 1995;Mullins et al., 2002). Persoonia rely on symbiotic associations with soil-borne microbes for micronutrient fixation (Lambers et al., 2015), yet unlike other Proteaceae growing in nutrient-poor soils, Persoonia do not develop proteoid root nodules and instead are thought to rely on free-living microbes.
As a precursor to these unique arrangements, soil microbiology could also play a role in germination, an aspect which is yet to be investigated. These factors cannot be accounted for in laboratory experiments and may therefore explain the variance observed in germination probability of individuals (17%-44% of variance explained).
However, much higher levels of variance (63%-84%) were accounted for across population. Maintaining seed banks beyond a single season, through innate low germination rates of viable seeds, may also represent a critical adaptation in pyrogenic environments, acting as a buffer for reproduction during periods of high fire frequency, where a first fire event has the potential to destroy parent plants, and a subsequent fire could kill off a new cohort of seedlings or juvenile plants before seed bank replenishment has occurred. Compared to co-occurring species in pyrophytic environments, Persoonia have long maturation periods (>7 years, 12 years to peak seed production for some species), indicating an increased vulnerability to seed bank destabilization (Auld et al., 2000(Auld et al., , 2007. Fire is, however, also a critical evolutionary force, known to prompt germination in Persoonia (McKenna, 2007;Nield et al., 2014). Understanding inter-fire germination and its effects on population dynamics and persistence remains an important knowledge gap to address, particularly for the three rare species in fragmented environments that have recently experienced little or no fire.

| Predicted outcomes
Temperature is recognized as a major driver by which climatic change impacts population persistence (Aitken et al., 2008;Gilman et al., 2010;Jagadish et al., 2016). Under projected climate change, soil temperatures are expected to increase as a direct result of increased radiative force, but also as a result of increased frequency and intensity of fire events resulting in more exposed soil (Auld & Ooi, 2008;Ooi, 2012;Ooi, Whelan, & Auld, 2006).
With no accurate estimates for soil temperatures, our predictions may in fact represent a conservative model that does not account for the higher temperatures of soil and the compounded effect of increased fire events under climate change scenarios.

A PPE N D I X 1
Results of seed moisture content derived from a seed coat ratio testing, showing desiccation in a Persoonia species are desiccation tolerant.