Predicting the consequences of global warming on Gentiana lutea germination at the edge of its distributional and ecological range

Study area

Gentiana lutea subsp. lutea (hereafter G. lutea) is mainly present in the Central-Southern European mountain ranges (Renobales, 2012). In Sardinia (Italy), it grows on the edge of its distributional and ecological range (Fois et al., 2015). Its distribution and its individuals are considered as a panmictic population based on recent genetic studies (Dettori et al., 2018). In detail, the distribution of G. lutea in Sardinia consists of only one population and it is restricted to a small area above 1,200 m a.s.l. in the Gennargentu massif (Fois et al., 2015). The annual mean temperature of the Gennargentu Massif varies from ca. 12 °C (at ca. 1,000 m a.s.l.) to ca. 7 °C in the higher areas (at ca. 1,800 m a.s.l.), with a snowfall period lasting from 3 to 4 months (Secci et al., 2010; Cuena-Lombraña et al., 2016, 2018a). The two collection sites, named Is Terre Molentes (IS) and Trainu Murcunieddu (TM), constitute the representative Sardinia natural growing sites of this species. Both are found at an altitudinal range of 1,300–1,500 m a.s.l. Is Terre Molentes is exposed to the northeast and Trainu Murcunieddu to the northwest (Cuena-Lombraña et al., 2016). The sites are characterised by siliceous metamorphic substrate and by an open grassland vegetation type (Cuena-Lombraña et al., 2016).

Study species and seed lot details

Gentiana lutea is an herbaceous rhizomatous perennial plant. The rhizome develops an approximately 80-cm-high flowering stalk during late spring/early summer, with 4–8 oblong leaves, forming a basal rosette (Renobales, 2012). The fruits consist of many-seeded capsules that ripen during the summer, while seed dissemination takes place in August by anemochory (Cuena-Lombraña et al., 2018a). After dispersal, the seeds are exposed first to a slightly warm post-dispersal period before winter begins, and then to low soil winter temperatures (near 0 °C for ca. three months, due to the conditions under the snow cover), while germination occurs in the following spring (Cuena-Lombraña et al., 2016). G. lutea seeds have linear underdeveloped embryos and show a morphophysiological dormancy (MPD) with intermediate complex level (Cuena-Lombraña et al., 2018a).

For these experiments, ca. 30 mature fruits (capsules) per individuals of G. lutea were collected from at least 50 randomly selected plants in each site, during August 2016 (Authorisation reference: DPM/5D/2005/26104; Ministry of the Environment, Land and Sea; Italy). They were manually cleaned, and well-developed seeds were selected in the laboratory, discarding any visually malformed seeds, and maintained at room temperature (ca. 40% of relative humidity and 20 °C) until the start of the experiments. Seeds were sampled at the same time in two representative localities: one located at 1,460–1,575 m a.s.l. (Is Terre Molentes, IS) and the other one located at 1,324–1,372 m a.s.l. (Trainu Murcunieddu, TM).

Seed germination tests

There were five cold stratification pre-treatments at 1 ± 1 °C applied to G. lutea fresh seeds (600 seeds per treatment per site), following an experimental design that took into account the possible effect of the reduction of the cold period during winter on dormancy release. Concretely, the following pre-treatments were applied: (I) cold stratification for 0 days (C0), (II) cold stratification for 15 days (C15), (III) cold stratification for 30 days (C30), (IV) cold stratification for 60 days (C60) and (V) cold stratification for 90 days (C90). In order to simulate the snow cover period, the chilling pre-treatments were performed in dark conditions (0 h light/24 h dark) (Cuena-Lombraña et al., 2018a). Pre-treatments started at the same time within two weeks after harvesting. Four replicates of 25 randomly selected seeds (maintaining separate the seeds from the different collection sites) were sown in 90-mm diameter plastic Petri dishes for each experimental condition with 1% agar water substrate and incubated with a day/night cycle (12 h light/12 h dark) at 5, 10, 15, 20, 25 and 30 °C in germination chambers (Sanyo MLR-351) with white fluorescent lamps (FL40SS.W/37 70–10 μmol m⁻² s⁻¹). Following the methodology used for G. lutea in Cuena-Lombraña et al. (2018a), the germination was checked three times a week, and germinated seeds were scored when a radicle ≥1 mm long was visible. After a minimum period of 90 days, if no additional germination had occurred for two further weeks, non-germinated seeds were cut with a scalpel in order to determine the number of filled, viable and empty seeds (Bacchetta et al., 2006; ISTA, 2008). For germination details, see the Supplemental File.

Data analysis

For each germination trial, the final germination percentage was calculated on the basis of the filled seeds, excluding the empty seeds, and reported as the mean of the four replicates (±SD). All non-dormant seeds that germinated after different cold stratification periods were considered for the thermal time analyses.

On the basis of the germination rate responses obtained from 10% to 90% of germination, the time estimates (t_g, days) required to reach different cumulative germination percentiles (g; with increments of 10% germination) were interpolated from the germination progress curves (Covell et al., 1986). Using a linear model, the germination rate (1/t_g) was regressed according to equation ‘1/t_g (days⁻¹) = (T_g – T_b)/θ’ as a function of temperature (García-Huidobro, Monteith & Squire, 1982). In order to establish the T_b for each pre-treatment, a mean (±SD) of the x-intercept among percentiles was calculated under the suboptimal temperature range, for example when the germination rate increases linearly with temperature to an optimum temperature (T_o) and, above which, the germination rate starts to decrease (Ellis et al., 1986; Pritchard & Manger, 1990). Due the very low germination percentages obtained after C0, C15 and C30, it was possible to fit the linear regression model only for C60 and C90. Accordingly, the base temperature was possible to estimate only for seeds that germinated after the pre-treatments C60 and C90. The T_b of each locality and pre-treatment, estimated as the mean value of the x-intercepts of the percentiles for which regression lines presented a P < 0.05, were considered. The simple linear regression for the germination rate for each percentile, constrained to pass through T_b, were recalculated (Hardegree, 2006). The best-adjusted model resulting in the smallest residual variance was considered (Covell et al., 1986). Thermal times (θ, °Cd) were calculated as the inverse of the suboptimal regression equations (Covell et al., 1986).

Statistical analyses

Generalised linear models (GLMs) with a quasi-binomial distribution and logit link function were used to evaluate the effect of pre-treatments, temperatures and localities (independent variables) on germination percentages (dependent variable), while a GLM with a quasi-Poisson distribution and logarithmic-link function (log-link) was used for analysing the base temperature (T_b). On the subsequent ANOVA, F tests with an empirical scale parameter instead of chi-squared was used (Crawley, 2007). Then significant differences among treatments, temperatures and T_b values were analysed by a post-hoc pairwise comparisons t-test (with Bonferroni adjustment). Statistical analyses were conducted using R version 2.14.1 (R Development Core Team, 2015).

Projections for future seed germination under climate change conditions

This study also evaluated the potential changes for the future seed germination of G. lutea across the different scenarios proposed by the IPCC Team, Pachauri & Meyer (2014). In detail, we considered the four RPC emission scenarios (i.e. RCP2.6, RCP4.5, RCP6.0 and RCP8.5), which represent the different estimated greenhouse gas emissions (IPCC Team, Pachauri & Meyer, 2014). For each of them we considered the low and the high global mean surface temperature predicted, understood as optimistic projections (OP) and less optimistic projections (LOP), respectively. Concretely, we used (I) for RCP2.6, the temperature +0.3 °C and +1.7 °C; (II) for RCP4.5, +1.1 °C and +2.6 °C; (III) for RCP6.0, +1.4 °C and +3.1 °C and (IV) for RCP8.5, +2.6 °C and +4.8 °C. In order to obtain these data and to try to estimate the effect of temperature rise on seed dormancy release and germination of G. lutea, we analysed data recorded by soil data loggers (TidbiTw v2 Temp logger; Onset Computer Corporation, Cape Cod, MA, USA) sowed in both the natural localities from 2013 to 2016 (Fig. 2). Using the onset HOBOware PRO (software for HOBO data loggers & devices, Version 3.7.8, Onset Computer Corporation 2002–2016), we extracted the number of days with temperatures ≤1 ± 1 °C recorded for each locality (i.e. the temperature detected for dormancy release; Cuena-Lombraña et al., 2018a). Afterwards, we increased the temperature following the RCPs as detailed above and estimated the number of days with temperatures ≤1 ± 1 °C for each of them in the two sites under study, in order to understand if the temperature required for dormancy release persisted under the predicted increase temperature conditions.

Germination behaviour

Table 1 shows the final germination percentage obtained at the end of each germination test at each temperature tested for seeds collected from the two localities. Seeds pre-treated for 90 days at 1 ± 1 °C (C90) achieved high germination percentages (more than 80%), while, under the other pre-treatments, the germination percentages decreased proportionally with the duration of the pre-treatment (Fig. 3). In the case of C60, the germination percentages were lower than 60%, in the C30 pre-treatment, they were lower than 40%, and, in C15 and C0 pre-treatments, they were lower than 20% (Fig. 3).

In general, the germination of G. lutea seeds decreased remarkably at high temperatures (25 and 30 °C, Fig. 3), and, regardless of the pre-treatments, these conditions differed statistically from the temperature range comprised among 5 and 20 °C (Fig. 3).

GLM results indicated that pre-treatments and temperatures factors and their interaction had a statistically significant effects on germination (Table 2), highlighting that there was an interaction between the explanatory variables. The results allowed to highlight that the seeds of G. lutea responded differently to the pre-treatments and temperatures tested. On the specific analysis carried out on the pre-treatments (C0, C15, C30, C60 and C90), GLM highlighted that significant differences were among them (Table 2). The post-hoc analysis indicated that there were no statistical differences in the germination response between the treatments C0 and C15 (Table 3), while the differences among C0 and C15 and the longer pre-treatments were significant in all cases (Table 3). In the other cold stratification pre-treatments, statistically significant differences were highlighted among C30, C60 and C90 pre-treatments (Table 3).

Base temperature for seed germination and thermal time requirements

The germination rate responses obtained from 10 to 50−60% and from 10% to 90% permitted to fit the linear regressions of the relationship between the temperatures and the rate of germination (1/t_g) for seed germinated after C60 and C90, respectively (Figs. 4A–4D). In accordance to the goodness of fit (r²), the sub-optimal temperature for TM encompassed only data up to 15 °C (Figs. 4A and 4C), while for IS, the temperature 15 °C (in C60) and 20 °C (in C90) were included in the sub-optimal temperature (Figs. 4B and 4D). The T_b for C60 were 3.34 ± 1.03 °C for IS and 2.15 ± 0.86 °C for TM (Figs. 4A and 4C), while the T_b for C90 were −1.96 ± 1.23 °C and −1.58 ± 3.74 °C for IS and TM, respectively (Figs. 4B and 4D). The mean T_b estimated values showed no statistical differences between localities (Fig. 4). However, significant differences were detected among pre-treatments (Fig. 4). In both localities, an increase of T_b with the decrement in the duration of the cold stratification was detected (Fig. 4). In IS, the thermal time required for 50% of germination (θ₅₀) was slightly greater in C90 than C60 (Fig. 5). In TM, θ₅₀ was slightly greater in C60 compared to C90 (Fig. 5). However, in general, θ₅₀ were similar for both localities and, after both cold stratification periods, with values that ranged from ca. 108 °Cd to ca. 130 °Cd (Fig. 5; Table 1).

Projections for the future seed germination under climate change conditions

Using the RCP scenarios and extrapolating the soil data information from the studied localities (Fig. 2), the total number of days with temperature ≤1 ± 1 °C were calculated (see Table 4). Starting from the current data and analysing the temperature recorder by data logger from the year 2013 to 2016, the number of days with temperature ≤1 ± 1 °C was 76.33 ± 14.36 in IS and 78.33 ± 12.66 in TM (see Table 4). Incrementing the temperature following the RCP scenarios, it was possible to detect that under the predicted temperature change scenarios, the number of days will decrease in both localities. In detail, under the RCP2.6 scenario, seeds of G. lutea could break their dormancy (ensuring the germination) only in the case of the increment of +0.3 °C (optimistic projection), in which there will be present at least 60 days of temperature ≤1 ± 1 °C (Table 4). In the same scenario but with the increment of +1.7 °C (less optimistic), the number of days with temperature ≤1 ± 1 °C will decrease to less than 30 days (Table 4), which will lead to a condition that will not permit the completion of dormancy release, compromising the germination capacity of G. lutea seeds. Under the RCP 4.5 scenario, the increment of +1.1 °C predicted for the optimistic projection, the number of days with temperature ≤1 ± 1 °C will be ca. 35–40 days (see Table 4), which could permit too few seeds to germinate (ca. 30%; see Fig. 3). Under the others RCP scenarios (i.e. RCP6.0 and RCP8.5), both in the optimistic and less optimistic ones, the number of days with temperature ≤1 ± 1 °C will be very low (Table 4).

Effect of reduction of cold stratification period on seed germination

The responses to different durations of the cold stratification period indicated that G. lutea needs at least 60 days of cold pre-treatment in order to release seed dormancy and reach a successful germination (understood as more than 50% of germination). In detail, our study detected some critical issues related to the number of days with cold period or snow cover that allow us to hypothesise that the capacity of germination of G. lutea seeds will be drastically reduced in the future as a consequence of the global warming trend. Although it is true that the reproductive phase offers the plant species the opportunity to adapt gradually to environmental changes (Hedhly, Hormaza & Herrero, 2009), the changes that plants are currently experiencing are occurring too fast to have enough time to adapt to the temperature change (Hedhly, Hormaza & Herrero, 2009). In this context, the results obtained here for G. lutea appeared in contrast with that obtained for Rhamnus persicifolia Moris seeds, in which a reduction of the cold stratification period from 90 days to 30 days would not be detrimental per se for seed germination, although under 30 days of cold stratification, the seed germination was importantly reduced (Porceddu et al., 2013). This aspect permits the highlighting that different plants growing in similar climatic conditions may respond differently to the rapid climate changes and that some species may suffer the effects of climate change before others.

In the face of the current climate change, the migration and persistence of plants depends on their successful reproduction by seeds, a fundamental aspect of the life cycle of the plant that affects population performance, as well as plant species distribution and community assembly (Fernández-Pascual, Mattana & Pritchard, 2019). Consequently, G. lutea and other species that need minimum days of cold temperatures to promote their seed dormancy release and germination could be more threatened in the future, unlike species without such requirements. Species growing in the same Sardinian areas of G. lutea may have similar thermal requirements for seed dormancy release and germination, responding similarly to global warming. On the other hand, however, other species that occupy the same areas may require different conditions, like higher temperatures for dormancy breaking, causing consequently different germination behaviour with respect to the seeds of G. lutea. For example, Sardinian species growing at the same altitudinal range (e.g. from 1,200 to 1,350 m a.s.l.) may reduce the capacity of dormancy release under one month at a temperature of 5 °C (e.g. in R. persicifolia; Porceddu et al., 2013), while a warm condition as temperature of 25 °C for three months may promote the dormancy release allowing the seed germination (e.g. in Ribes sandalioticum (Arrigoni) Arrigoni and Ribes sardoum Martelli, Mattana et al., 2012; Porceddu, Fenu & Bacchetta, 2017). As reported by Briceño, Hoyle & Nicotra (2015) and Walder & Erschbamer (2015) some species could be indirectly favoured by a future warming climate, resulting in significant changes in the composition and dynamics of different plant communities. All this suggests that a deeper study on a wide scale may help in understanding if other species distributed with G. lutea will be affected by global warming in a similar way.

Theoretically, regeneration processes are more suitable in the centre of the species’ distribution range than in the natural distribution limits (Lawton, 1993). For this reason, viability and regeneration evaluations of peripheral populations of plants under different climate change scenarios have become a conservation concern (Giménez-Benavides, Escudero & Iriondo, 2007). In this study, focused on the populations of G. lutea located at the edge of its distributional and ecological range, the germination responses highlighted a critical decrease of the germination percentage with the reduction of a cold stratification period. This could indicate that these populations, which are located at lower altitudes and at the edge of its distributional and ecological range, would be more threatened, if facing shorter winters. Seed germination results obtained in this work, in particular, the detection that seeds of G. lutea needed at the least 60 days of cold temperature to remove dormancy, are in agreement with previous studies (Fois et al., 2016; Cuena-Lombraña et al., 2018b) which reported that the most affected localities are those that are present at the lowest altitudes. Therefore, with the results presented here, it is possible to affirm that, while also considering the germination capacity of the seeds, climate warming will likely first affect the localities of G. lutea at lower altitudes.

Cardinal temperatures and thermal time

The optimal temperatures of non-dormant seeds of G. lutea detected in this work is assumed to be around 10−20 °C, as the best models in the sub-optimal range were obtained by considering a temperature from 5 to 20 °C. On the other hand, the temperatures >20 °C were included in the range of supra-optimal temperatures. The latter aspect highlights that the higher temperatures negatively affected the germination capacity of non-dormant seeds of G. lutea; it was already the case at 25 and 30 °C, that is the higher temperatures used in this work. This assumption is in accordance with the temperatures (from 10 to 20 °C) suggested in the optimal germination protocol by Cuena-Lombraña et al. (2018a) further confirming that the seeds of this species are characterised by spring germination. The germination of seeds of G. lutea occurs, in particular, after the snow melting period in which the risk of frost is low, the availability of water is high and the time of solar radiation is longer. The base temperature (T_b) in G. lutea seeds, regardless of the locality, varied from ca. 2 to 3 °C (after 60 days of cold pre-treatment) to ca. −2 °C (after 90 days of cold pre-treatment). Pre-treatments for dormancy release (C60 and C90) allowed the reduction of the T_b in G. lutea seeds. The effect of cold stratification in reducing the T_b has been also observed in seeds of Aesculus hippocastanum L. (Pritchard et al., 1999) and R. persicifolia (Porceddu et al., 2013). The reduction of the base temperature should increase the rate of accumulation of thermal units, as it happens in species like Vitis vinifera L. subsp. sylvestris (C.C. Gmel) Hegi (Orrù et al., 2012) and R. persicifolia (Porceddu et al., 2013). However, in seeds of G. lutea, the duration of cold stratification periods that promote dormancy release and seed germination (i.e. C60 and C90) do not cause a large variation in their thermal time estimates (θ₅₀), whose values were similar under the two conditions. In this sense, after a certain period of cold stratification, seeds of G. lutea start to accumulate thermal time units that allow them to germinate and with the reduction of duration of the overwintering season (<60 days), dormancy loss could be compromised. In addition, the thermal time (θ₅₀) requirements after brief winters could remain too high, thus potentially preventing seed germination.

Implications for seed germination under global warming

As far as the ecological implications under climate change scenarios are concerned, the future conditions may not meet the future requirements for breaking physiological dormancy in G. lutea seeds and will be detrimental to the proportion of seeds which germinate. Similarly, as reported by Orrù et al. (2012) for V. vinifera subsp. sylvestris, under the two simulated IPCC scenarios that they considered (e.g. an increase in temperature of +1.8 °C and +3.4 °C), the seed germination could be compromised because the future winters temperature may not be cold enough to break physiological dormancy. Contrary to these responses, an increase in soil temperature is predicted to alter the timing of germination of Polaskia Backeb, but this will not be detrimental to the proportion of seeds germination (Ordoñez-Salanueva et al., 2015).

As reported in Table 3, under several RCP scenarios (sensu IPPC, 2014), the number of days with a suitable temperature for dormancy release in G. lutea decreased drastically, particularly under the less optimistic projections. As already mentioned in this work, after a certain duration of chilling, seeds of G. lutea start to accumulate thermal time units that allow them to germinate. If the future increase in temperatures, as predicted by IPCC, reduces the duration of the overwintering period, dormancy loss and, consequently, seed germination of G. lutea could be compromised. In addition, if the autumn/winter season shortens, the thermal time (θ₅₀) requirements could remain too high, potentially reducing the likelihood of seed germination. This statement is in agreement with what was reported in the ‘Model D1’ by Fernández-Pascual, Mattana & Pritchard (2019). The same authors reported the importance of considering the thermal memory effect of seed dormancy (see ‘Model D2’; Fernández-Pascual, Mattana & Pritchard, 2019). They suggest that, if a shortened winter causes a decrease in the amount of dormancy loss, thermal memory might be enough to reduce the value of θ₅₀ to a level that allows germination at an adequate time. However, in the case of G. lutea, we theorise that under several RPC emission scenarios (see Table 3), thermal memory (sensu Fernández-Pascual, Mattana & Pritchard, 2019) may not be enough to adjust its germination phenology to global warming.

In particular, for the populations of G. lutea located at the edge of its distributional and ecological range, we hypothesised that increases of more than +1.7 °C would significantly reduce the germination of this species in field conditions. In the literature, we find controversial results related to the germination behaviour of alpine or mountain species under climate change conditions. On the one hand, there are studies which showed that warming can either increase (Milbau et al., 2009; Shevtsova et al., 2009; Mondoni et al., 2015; Wang et al., 2018) or decrease (Jump et al., 2008; Hoyle et al., 2015) germination percentage. The response to the temperature increments varies depending on the plant species (Hedhly, Hormaza & Herrero, 2009) and for this reason, it is particularly important to study each species individually. The consequences of a warming environment on plant reproduction cannot be simply interpreted qualitatively as a series of absolute positives and negatives (Fernández-Pascual, Mattana & Pritchard, 2019). The effect of temperature rise should instead be understood in terms of the variation of the rates and thresholds that quantify the physiological processes underlying reproduction by seed (Fernández-Pascual, Mattana & Pritchard, 2019).

We might expect that a period of cold stratification would be an ordinary condition for seed germination to occur in mountain or alpine plant species, but this is not always the case. For example, in the studies carried out by Sommerville, Martyn & Offord (2013) and Hoyle et al. (2015) on 19 and 54 Australian species, respectively, only half of the taxa studied increased their seed germination performance as a result of a cold stratification period. Contrarily, in most of the 27 Japanese alpine species (Shimono & Kudo, 2005) and in alpine plants from the Central Chilean Andes (Cavieres & Sierra-Almeida, 2018), a cold period of stratification was needed to promote seed germination. Moreover, species living in environments with a contrasting snow cover period, as occurs in the Mediterranean mountains, indicate that there is a positive relationship between seed germination response and the length of the snow cover period (Meyer, McArthur & Jorgensen, 1989; Meyer, Kitchen & Carlson, 1995). This aspect is in accordance with what we detected for G. lutea, in which the seed germination increased with the increase in length of cold stratification. Infact, our study detected some critical issues related to the number of days with cold temperature, suggesting that the germination of G. lutea will be reduced in the future, as a consequence of global warming. In accordance with Briceño, Hoyle & Nicotra (2015) and Walder & Erschbamer (2015), we expected that a further decrease in seedling emergence will occur in the future due to global warming, thus leading to a shift towards asexual reproduction for G. lutea. The reproductive phase represents an opportunity for the plant to adapt to a changing environment. Therefore, new knowledge on plant adaptation, plant evolution and vegetation dynamics under a global warming might derive from the understanding of plant plasticity in response to prevailing temperatures and the associated potential changes (Hedhly, Hormaza & Herrero, 2009).