Burning seeds: germinative response to heat treatments in relation to resprouting ability

Authors


Correspondence author. E-mail: pausas@gmail.com; juli@ceam.es

Summary

  • 1In Mediterranean fire-prone ecosystems, plant species persist and regenerate after fire by resprouting, by recruiting new individuals from a seed bank (post-fire seeding), or by both resprouting and post-fire seeding. Since species with resprouting ability are already able to persist in fire-prone ecosystems, we hypothesize that they have been subjected to lower evolutionary pressure to acquire traits allowing or enhancing post-fire recruitment. Consequently, we predict that the germination of non-resprouters is more likely to be increased or at least unaffected by heat than the germination of resprouters.
  • 2To test this hypothesis we compiled published experiments carried out in Mediterranean Basin species where seeds were exposed to different heat treatments. We compared the probability of heat-tolerant germination (i.e. heated seeds had greater or equal germination than the control), the probability of heat-stimulated germination (i.e. heated seeds had greater germination than the control) and the stimulation magnitude (differences in proportion of germination of the heated seeds in relation to the untreated seeds, for heat-stimulated treatments) between resprouters and non-resprouters.
  • 3Non-resprouters showed higher probability of heat-tolerance, higher probability of heat-stimulation and higher stimulation magnitude even when phylogenetic relatedness was considered. Differences between life-forms and post-fire seeding ability do not explain this pattern.
  • 4Non-resprouters appear to have a greater capacity to both (i) persist after fire by means of recruiting (greater heat-tolerance) and (ii) increase their population after fire (greater heat-stimulated germination), than resprouters.
  • 5Synthesis. Our results contribute to understanding the factors that condition the evolution of fire-persistence plant traits and support the hypothesis that resprouting and post-fire recruitment are negatively associated in Mediterranean Basin flora.

Introduction

Fire is one of the most predictable disturbances faced by plants in Mediterranean ecosystems (Di Castri et al. 1981). Plant species have two basic mechanisms for persisting in Mediterranean fire-prone ecosystems (Keeley & Zedler 1978; Pausas et al. 2004): by regenerating their above-ground biomass (i.e. resprouting) by plants that survived the fire (persistence at the individual and population-level) and by recruiting new individuals from a fire-resistant seed bank (persistence at the population-level only). Resprouters and seeders are the typical terms used to refer to the species that regenerate after fire primarily through the first or the second mechanism, respectively (Keeley & Zedler 1978). In fire-prone ecosystems, resprouters and seeders coexist; in addition, many species may have traits that allow them to both resprout and recruit after fire (e.g. Gill 1981; van Wilgen & Forsyth 1992; Enright & Goldblum 1999; Pausas et al. 2006). In the Mediterranean Basin, there is a negative correlation between the traits associated with post-fire resprouting and those associated with post-fire seeding such that post-fire seeding traits appeared later in evolution than resprouting traits and were mainly acquired by non-resprouting lineages (Pausas & Verdú 2005). These results suggest that both strategies may be efficient for persistence in Mediterranean fire regimes and that having one of them may reduce the probability of acquiring the other.

Post-fire recruitment may succeed because the heat of the fire may break seed dormancy or quiescence, and enhance germination (e.g. Thanos et al. 1992; Herranz et al. 1998; Keeley & Fotheringham 2000). Increased permeability of the seed coat and, to a lesser extent, induction of physiological processes and denaturing of seed coat inhibitors, are implicated in fire-triggered seed germination (Keeley 1991; Thanos et al. 1992; Bell et al. 1993; van Staden et al. 2000). Furthermore, in some species, seeds resist the heat produced by fire, but germination is not fire-stimulated (Trabaud & Casal 1989). Other germination cues related to fire, such as smoke, charcoal and nitrogenous compounds, have also been identified (Keeley & Bond 1997; Keeley & Baer-Keeley 1999; Clarke et al. 2000; van Staden et al. 2000; Wills & Read 2002; Pérez-Fernández & Rodríguez-Echevarría 2003); nevertheless, to respond to any of these cues, seeds must first survive the high temperatures released during a fire.

Fire-stimulated germination is common in most Mediterranean-climate shrublands although it seems to be taxonomically aggregated, most common in the Cistaceae and Fabaceae, and in some clades of the Rhamnaceae and Malvaceae (Thanos et al. 1992; Bell et al. 1993; Keeley & Bond 1997; Keeley & Fotheringham 2000). Nevertheless, it has been suggested that this trait is less frequent in the Mediterranean Basin than in other Mediterranean ecosystems (Keeley & Baer-Keeley 1999), and a substantial amount of post-fire recruitment is probably the result of the seeds being tolerant to fire rather than being stimulated by it (Buhk & Hesen 2006; Luna et al. 2007).

Both fire-resistant and fire-stimulated germination require the presence of a persistent seed bank or the production of heat-resistant seeds just before fire (Pausas et al. 2004). Dormancy, which is strongly correlated with persistence of seed banks (Thompson et al. 1998), is a widespread heritable trait among angiosperms (Baskin et al. 2000). However, great inter- and intra-variability in seed dormancy and heat-stimulated germination has been found in several species (Keeley 1991; Pérez-García 1997; Herranz et al. 1999; Cruz et al. 2003a; see Tieu et al. 2001 for variability in the germinative response to smoke). Because natural selection acts on the variability of heritable traits, the evolution of heat-stimulated germination could be determined by the intensity of the selective pressure.

Whereas the resistance of seeds to fire only ensures the persistence of the plant population, fire stimulated germination permits a large number of offspring to be produced after fire and increases the population size compared with the pre-fire population (Roy & Sonie 1992; Ladd et al. 2005). Rapid post-fire recruitment enhances plant fitness by accessing more resources (Bond & van Wilgen 1996) and thus growing faster, shortening the time to reach maturity after fire and increasing the probability of storing a large amount of seeds before the next fire (Le Maitre & Migdley 1992; Verdú & Traveset 2005). Consequently, post-fire seedling emergence is under strong selective pressure.

We hypothesize that because species with resprouting ability are able to persist in fire-prone ecosystems, they may have been subjected to lower evolutionary pressure to acquire traits that enable or enhance their recruitment after fire. On the contrary, species with no resprouting ability can only persist in fire-prone environments if they have an efficient post-fire recruitment process. Consequently, we expect a higher germinative response to heat treatments in non-resprouters than in resprouters. That is, we predict that the germination of non-resprouters is more likely to be increased or at least unaffected by heat than the germination of resprouters. To test this hypothesis we performed an exhaustive compilation of published experiments in which seeds from Mediterranean Basin species were exposed to different heat treatments (combinations of temperature and exposure time), and then we compared the germinative response of the resprouters and non-resprouters. The germinative response was characterized as: (i) heat-tolerance, if germination of the heated seeds was greater or equal to that of the control; and (ii) heat-stimulation, if germination of heated seeds was greater than that of the control. The first case is related to the potential persistence of the population regardless of changes in population size, while the second indicates a potential increase in the post-fire population size. So, specifically, we tested whether heat-tolerance and heat-stimulation are higher for non-resprouters than for resprouters.

Methods

data set

We compiled a data set on heat experiments, that is, experimental germination tests under different temperatures and exposure times, performed on species of the Mediterranean Basin. We considered only dry-heat exposure experiments (cases using scarification with boiling water were not included) and only those in which the temperatures applied were ≥50 °C. For each study and species, we compiled the percent germination of the control and the percent germination of each heat treatment (combination of temperature and exposure time). Cases in which both the control and the treatment showed 0% germination were excluded. For experimental designs in which seeds of the same species and population were subjected to different storage conditions (e.g. with and without stratification) and/or incubation regime (e.g. in dark and in light conditions), we included only the germination test in which the untreated seeds showed the highest germination. This is a conservative criterion for avoiding pseudo-replication in the data set. Experiments performed on seeds from serotinous species were discarded because seeds inside cones are protected from the heat of a fire (Habrouk et al. 1999). In the Mediterranean Basin, serotiny is only found in some conifers (Pausas 1999), and the amount of serotinous cones varies greatly among populations (Goubitz et al. 2004; Tapias et al. 2004); furthermore, the selective pressure of fire on serotinous species may act on different traits from seed traits (e.g. cone traits). To ensure consistency, all coniferous species were excluded from the data set; our analysis was thus restricted to angiosperms.

For the compilation, species names were standardized following the European Science Foundation – European Documentation System (ESFEDS) which is largely based on Flora Europaea (Tutin et al. 1964–80). The ESFEDS database was checked in the Global Plant Checklist of the International Organization for Plant Information (IOPI 1996–2005). Other criteria were followed in cases where some important taxonomical updates had been carried out (e.g. Talavera et al. 1999 for Genisteae) or when some species were missing in the ESFEDS database (e.g. Greuter et al. 1984–89 for Cistaceae). For family names, the Angiosperm Phylogenetic Group standards were used (Bremer et al. 2003).

We found 53 references reporting appropriate data on heat experiments for 164 species or subspecies included in 28 families and 85 genera (Appendix S1 in Supplementary material). This yielded a total of 1684 entries. Most of these studies (78%) report the statistical test between the control and the treatment.

For each species in the data set, we compiled its post-fire response based on bibliographic references and personal field observations. Because the germinative response to fire cues is related to the life-form (Keeley & Bond 1997; Clarke et al. 2000; van Staden et al. 2000), we also compiled this trait, mostly from local floras. The life-forms considered were: woody plants, perennial herbs (including hemicryptophytes, geophytes and some plants slightly lignified at the base) and annuals. We considered that a species had the ability to resprout (R+) if there was field evidence of resprouting after 100% scorching (Gill 1981; Pausas et al. 2004); otherwise, it was considered a non-resprouter (R–). Species with field evidence of post-fire seedling emergence from a soil seed bank or from seeds produced just before fire were considered seeders (S+); otherwise they were considered non-seeders (S–). Note that species recruiting after fire from outside populations (arriving by dispersal) were not regarded as S+ as their seeds are not subjected to fire. Therefore, four post-fire strategies are possible: R+S–, R+S+, R–S+ and R–S– (Pausas et al. 2004). Life-form was assigned to all species, but resprouting and post-fire seeding was obtained for 120 and 105 species, respectively; thus, when testing the resprouting or seeding factor, the total cases used were lower (1399 and 1451, respectively). In general, most species were woody plants (61%), with similar frequency among R+ and R–, whereas most of the perennial herbs were R+. S+ species were more abundant than S– and their frequency was higher among R– than among R+ (80% and 41%, respectively).

analysis

Heat experiments involve very different combinations of temperature and exposure time. To estimate the accumulated heat dose received by the seeds in each treatment, we multiplied the temperature (in degrees Celsius) by the natural logarithm of the exposure time plus one (in minutes). This heat index (H) is analogous to the concentration-time product used in toxicology (e.g. Wang et al. 2004 and references therein). In the compiled data, essayed temperatures ranged from 50 to 300 °C, exposure times ranged from 1 min to 24 h and the heat index (H) ranged from 34.7 to 727.3 units (Fig. 1).

Figure 1.

Relationship between temperature and exposure time for different heat index (H) values (contour lines).

For each treatment, the germinative response was first classified as greater, equal or lower than the control (i.e. stimulated, unaffected or inhibited by heat, respectively) on the basis of the significance of the statistical test provided in each study. A set of cases (22%) did not include any statistical test. These cases include both resprouting and non-resprouting species (51% and 49%, respectively). In order to avoid any loss of information, we classified the germinative response of statistically non-tested cases by comparing them with the tested experiments. For the statistically tested experiments, we plotted the frequency distribution (and the smoothed density distribution) of the difference in percent germination between control and treatment for the cases with greater, equal and lower germination than the control. The intersections between the three smoothed density distributions were –13 and 14. We then used these thresholds to classify the statistically non-tested treatments as stimulated (differences in percent germination between control and heated seeds higher than 14%), inhibited (differences lower than –13%), or unaffected (in between) germination. If we apply these thresholds to the statistically tested treatments we observe that the percentage of misclassified cases is 10%. This classification error was mainly (6.9%) due to the failure to detect a significant difference when it was different (type II error or false negative); therefore, the classification is conservative. Furthermore the frequency of treatments that stimulate, do not affect or inhibit germination was not significantly different between the tested and non-tested treatments, both when considering all species (χ2 = 0.54, d.f. = 2, P = 0.762) or each resprouting ability separately (R+: χ2 = 4.99, d.f. = 2, P = 0.083; R–: χ2 = 0.67, d.f. = 2, P = 0.715). The close-to-significant χ2-test for R+ is due to a slight over-representation of inhibited germination cases in the statistically non-tested data set, and thus caution should be taken only if the results comparing heat-tolerance between resprouting abilities (see below) were not strongly significant. It is notable that the whole data set includes more treatments (29%), more species (8%) and more treatments per species (19%) than the tested subset, and a wider overall range of heat index (34.7–727.3 for the whole data set vs. 34.7–456.7 for the tested treatments).

The probability of heat-tolerance (i.e. of having greater or equal germination than the control) and heat-stimulation (i.e. of having greater germination than the control) was analysed using a GLM model with a binomial error distribution and logit link function, and tested by an analysis of deviance (McCullagh & Nelder 1989). To analyse the extent to which germination was different between resprouting abilities, we used resprouting as a factor and the heat index (H) as a covariable. Since all life-forms and post-fire seeding abilities were not equally represented in the data set between R+ and R– (see above), we also compared the probability of tolerance and stimulation between life-forms and between post-fire seeding abilities (S). We expect a decreasing probability of heat-tolerance with H and a unimodal model of heat-stimulated germination probability in relation to H (i.e. low germination at low and high intensities); thus, we tested a monotonic response of heat-tolerance to H and a quadratic response of heat-stimulation to H (i.e. H + H2). In all analyses, species was also included in the model, nested in either post-fire regeneration abilities or life-form.

For treatments that produced stimulation, we evaluated to what extent the heat treatment increased germination in a different manner between R+ and R–, by testing the differences in the proportion of seeds germinating between the heated and the control treatments (hereafter, stimulation magnitude) against H and resprouting using the analysis of variance. As for the stimulation probability, we assumed a unimodal model in relation to H and included species as a factor nested in resprouting.

Because both resprouting and post-fire seeding are phylogenetically structured in the Mediterranean Basin (Herrera 1992; Pausas & Verdú 2005; Verdú & Pausas 2007), our differences in germination response between resprouting abilities could be driven by species relatedness. To evaluate this, we performed a test of the effect of resprouting ability on the probability of heat-tolerance and heat-stimulation, and on the stimulation magnitude, considering the phylogenetic relatedness among the species in our data set. We first assembled a phylogenetic tree with branch lengths for our species with the help of the Phylomatic software implemented in Phylocom 3.41 (Webb et al. 2007) and using the angiosperm megatree version R20050610.new. Then, we classified all the studied heat treatments in the following H classes: very low (H < 100), low (100 < H < 200), medium (200 < H < 300), high (300 < H < 400) and very high (H > 400). To conduct the phylogenetic analyses we recalculated the three dependent variables as follows: for each species and each H class we counted the cases with successful heat-tolerance and heat-stimulation, and computed the average magnitude. Finally, for each H class we tested the effect of resprouting ability on each of the dependent variables using a generalized estimating equation and including the phylogenetic relatedness among species (obtained from the phylogenetic tree) as a correlation matrix in the model (see Paradis & Claude 2002). As in the non-phylogenetic analyses, we assumed a binomial error distribution and logit link function for heat-tolerance and heat-stimulation, and a normal error distribution for magnitude. The advantages of using generalized estimating equation rather than other comparative methods is that it permits the use of qualitative variables (such as resprouting ability), the presence of polytomies in the phylogenetic tree, and the use of binomial error distributions (for germination probabilities). This analysis was performed using the ape software (Paradis et al. 2004).

Results

heat-tolerance: germination greater than or equal to the control

The probability of seed heat-tolerance was very high (close to 1) at low H values, decreasing with H either progressively or suddenly depending on the life-form and the post-fire regeneration strategy (Fig. 2). The probability of heat-tolerance was higher for R– species that for R+ ones (Fig. 2a). Both the H and its interaction with resprouting ability were significant (Table 1), suggesting that the probability of heat-tolerance decreases with H, and the decreasing pattern was quicker for R+ than for R– species (Fig. 2a). Considering the phylogeny, there were no differences between R+ and R– for the very low and low H classes (P = 0.484 and P = 0.207, respectively), but differences were significant and marginally significant for the moderate and high H class (P = 0.039 and P = 0.053, respectively). Nevertheless, differences between resprouting abilities for the very high H class were non-significant (P = 0.109) and thus attributable to the phylogenetic arrangement of the species at the higher end of the H gradient.

Figure 2.

Tolerance probability in relation to heat treatments for the different resprouting abilities (a, d), life-forms (b) and post-fire seeding abilities (c). In figure d, only woody S+ species were considered. Fits include significant (P < 0.05) effects only (see Tables 1 and 2 for further statistical details). Dotted lines refer to SEs of the fitted line.

Table 1.  Analysis of deviance for the probability of heat-tolerant germination in relation to resprouting ability, life-form, seeding ability and heat index
Source of variationd.f.DevianceResidual d.f.Residual devianceP
  1. All species are included in the analysis. (Levels of significance: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; P values are shown when P ≥ 0.05).

Null  13981183.26 
Resprouting (R)  1  8.8013971174.46**
Heat index (H)  1147.7913961026.67****
R × H  1 12.3813951014.29***
Species in R118321.771277 692.52****
Explained deviance (%)  41.47   
Null  16831435.43 
Life-form (LF)  2  8.1816811427.24*
Heat index (H)  1165.0916801262.15****
LF × H  2 13.5116781248.64**
Species in LF167438.841511 809.79****
Explained deviance (%)  43.59   
Null  14501178.32 
Seeding (S)  121.6614491156.66****
Heat index (H)  1132.2214481024.44****
S × H  1  0.3614471024.090.55
Species in S104283.221343 740.87****
Explained deviance (%)  37.13   

Differences between life-forms were also significant, with perennial herbs showing the lowest probability of heat-tolerance (Table 1, Fig. 2b). The interaction between life-form and H was also highly significant (Table 1): woody plants and annuals showed some degree of tolerance even at high H, whereas for perennial herbs tolerance decreased rapidly (Fig. 2b). S+ showed a higher heat-tolerance probability than S–, although the pattern of tolerance probability with H was quite similar for both S+ and S– (no significant interaction; Table 1, Fig. 2c). Considering all these results, the previous differences detected between resprouting abilities could be due to the fact that, in our data set, S– and perennial herbs were more frequent in R+ species than in R– (see Methods) and both had a low heat-tolerance probability (Table 1, Fig. 2b and c). However, when this comparison was restricted to woody S+ species (i.e. comparison between R+S+ and R–S+), we still detected differences between R+ and R–, although for high H values only (significant resprouting ability–H interaction; Table 2, Fig. 2d).

Table 2.  Analysis of deviance for the probability of heat-tolerant germination in relation to resprouting ability and heat index considering only woody S+ species
Source of variationd.f.DevianceResidual d.f.Residual devianceP
  1. (Levels of significance: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; P values are shown when P ≥ 0.05).

Null  1055787.92 
Resprouting (R) 1  0.661054787.260.42
Heat index (H) 1 90.291053696.97****
R × H 1 11.821052685.15***
Species in R52143.411000541.75****
Explained deviance (%)  31.24   

heat-stimulation: germination greater than the control

The probability of heat-stimulated germination was different between resprouting abilities (Table 3). R– showed a higher stimulation probability than R+ (Fig. 3a). Moreover, the relationship between stimulation probability and H differed between them (Table 3): for low H values, the stimulation probability increased with H at a similar rate in R+ and R–, reaching their respective maximum values of probability for slightly different H values; stimulation was negatively affected by H sooner in R+ than in R–, in such a way that for very high H values, no R+ species was stimulated whereas some Cistus species (R–) were still stimulated (Fig. 3a). This pattern was mostly maintained when phylogeny was considered: differences in heat-stimulated probability between R+ and R– were detected for very low, low and moderate H classes (P = 0.0024, P < 0.0001, P = 0.0027, respectively), but they disappeared for high H (P = 0.334), indicating that such differences in high H values are attributable to phylogeny and not to resprouting ability. In fact, this was expected as the phylogenetic affiliations of the data in the high H class differed between R+ (Ancardiaceae and Rosaceae) and R– (Cistaceae and Fabaceae). This comparison was not conducted for very high H, because only R– species were stimulated, all of them pertaining to the genus Cistus.

Table 3.  Analysis of deviance for the probability of heat-stimulated germination in relation to resprouting ability, life-form, post-fire seeding ability and heat index
Source of variationd.f.DevianceResidual d.f.Residual devianceP
  1. All species are considered. Note that the heat index was fitted as a quadratic response. Levels of significance: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; P values are shown when P ≥ 0.05.

Null  13981608.12 
Resprouting (R)  1 45.6613971562.47****
Heat index (H + H2)  2 41.0913951521.38****
R × (H + H2)  2 11.9013931509.48**
Species in R118357.2812751152.2****
Explained deviance (%)  28.35   
Null  16831891.74 
Life-form (LF)  2 48.2816811843.46****
Heat index (H + H2)  2 45.4916791797.97****
LF × (H + H2)  4  3.8616751794.120.43
Species in LF166441.8815091352.24****
Explained deviance (%)  28.52   
Null  14501736.96 
Seeding (S)  1 42.0914491694.87****
Heat index (H + H2)  2 53.7014471641.17****
S × (H + H2)  2  2.4714451638.700.29
Species in S104293.8613411344.83****
Explained deviance (%) 22.58   
Figure 3.

Stimulation probability in relation to heat treatments for the different resprouting abilities (a, d), life-forms (b) and seeding abilities (c). In figure d, only woody S+ species were considered. Fits include significant (P < 0.05) effects only (see Tables 3 and 4 for further statistical details). Dotted lines refer to SEs of the fitted line.

The heat-stimulation probability was higher for woody plants than for herbs, and slightly higher for annuals than for perennial herbs (Table 3, Fig. 3b). The pattern followed by the stimulation probability with H was the same for all life-forms (non-significant life-form–H interaction; Table 3, Fig. 3b). As expected, S+ species showed a higher stimulation probability than S– ones, although the maximum stimulation probability was always < 0.4; the stimulation probability for S– was < 0.1 (Fig. 3c). The interaction between seeding ability and H was not significant (Table 3, Fig. 3c). To discount whether the differences between resprouting abilities in the probability of heat-stimulated germination were due to the heterogeneous distribution of life-forms and seeding abilities between R+ and R–, we repeated the comparison but considered only woody S+ species. For this data subset, differences between R+ and R– were detected and the interaction between resprouting ability and H remained highly significant (Table 4, Fig. 3d).

Table 4.  Analysis of deviance for the probability of heat-stimulated germination in relation to resprouting ability and heat index, considering woody S+ species only
Source of variationd.f.DevianceResidual d.f.Residual devianceP
  1. Note that the heat index was fitted as a quadratic response. Levels of significance: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; P values are shown when P ≥ 0.05.

Null  10551337.28 
Resprouting (R) 1 16.1410541321.14****
Heat index (H + H2) 2 46.1110521275.13****
R × (H + H2) 2 12.7010501262.43**
Species in R52173.92 9981088.51****
Explained deviance (%)  18.60   

The magnitude of the stimulation differed between resprouting abilities (Table 5), being higher for R– (R–: 39.7 ± 19.9 and R+: 31.8 ± 18.3, expressed as mean ± SD). The magnitude of stimulation was independent of H and the differences between resprouting abilities were unaffected by H (non-significant resprouting ability-H interaction; Table 5). In the phylogentic analyses, differences in the magnitude of heat-stimulation between resprouting abilities (higher in R–) were significant for the very low, low and moderate H classes (P = 0.0006, P < 0.0001, P < 0.0001, respectively), but non-significant for high H (P = 0.294), because stimulation was phylogenetically aggregated for these H values (see before). Differences between R+ and R– in the magnitude of stimulation for the very high H class were not tested because none of the R+ species was stimulated.

Table 5.  Analysis of variance for the difference in percent germination (for treatments that produced stimulation) in relation to resprouting ability and heat index
Source of variationd.f.SSResidual d.f.Residual SSFP
  1. Note that the heat index was tested as a quadratic response. Levels of significance: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; P values are shown when P ≥ 0.05.

Null  365142 012  
Resprouting (R) 1  4850364137 16215.86****
Heat index (H + H2) 2   543362136 619 0.890.41
R × (H + H2) 1   551361136 068 1.800.18
Species in R5542 470306 93 598 2.53****

Discussion

As predicted, the germinative response of seeds after heat exposure was greater in non-resprouters (R–) than in resprouters (R+), because R– seeds show higher heat-tolerance and heat-stimulation (the latter expressed as both probability and magnitude). These results were maintained even when phylogenetic relatedness was included in the analyses. Similar results were found in the Fynbos for smoke-stimulated germination, being more frequent in R– species (van Staden et al. 2000). In the Mediterranean Basin, R+S– and R–S+ are more frequent than could be expected by chance, even when phylogenetic relatedness is considered (Pausas & Verdú 2005). Therefore, differences between R+ and R– in the germinative response to heat could be due to differences between S+ and S–. Nevertheless, when the probability of heat-tolerance or heat-stimulation was compared in terms of S+ species only, the differences between resprouting abilities remained significant. Recently, a study carried out with a subset of the data included in our analysis found that the percent germination after heat-shock did not differ between R+ and R– (Luna et al. 2007), although germination after high temperatures (100 and 120 °C for 10 min) was higher for R–. The wider spatial and taxonomic range of our data set and the great amount of combinations of temperature and exposure time tested in our study increase the power of the analyses and thus highlight the differences between resprouting abilities.

As a result of post-fire germination, R– species have a higher potential than R+ species to persist and even to increase their population size after fire. Moreno & Oechel (1991a) found a Californian example in which the post-fire emergence of an obligate seeder (R–S+; Ceanothus greggii) was higher than for a facultative seeder (R+S+; Adenostoma fasciculatum), especially after high-intensity fires. The authors proposed differences in the germinative response and in the seed burial depth in the soil as possible explanations for this finding. Our results support the first hypothesis, although the latter cannot be discarded. Similar results have also been found in serotinous species from Western Australia (Enright & Lamont 1989). Nevertheless, effective post-fire recruitment also depends on the amount of seed availability and seedling survival. Regarding seed availability, obligate resprouters (R+S–) typically produce fewer seeds than obligate seeders (R–S+), although the differences between R+ and R– in seed production do not always occur in S+ species (Pausas et al. 2004). In the Mediterranean Basin, species with both resprouting and seeding ability are less frequent than species that persist by means of only one of the recovery mechanisms (Pausas & Verdú 2005); therefore, differences in seed production between R+ and R– are expected, with the latter being higher (Pausas et al. 2004). Moreover, congeneric comparisons of three co-occurring R+ and R– pairs have suggested that viable seed density in the soil is higher for R– species (Kelly & Parker 1990; see Enright & Lamont 1989 for similar results in serotinous species). On the other hand, seedling establishment is the most critical phase in the life-history of Mediterranean species, mostly limited by seasonal dryness (Mejías et al. 2002). Seedling survival under seasonal water stress is frequently higher for R– than for R+ species (Keeley & Zedler 1978; Zammit & Westoby1987; Davis et al. 1998; Enright & Goldblum, 1999). In fact, at least at leaf level, R– species show higher potential for structural resistance to drought and higher water-use efficiency than R+ (Paula & Pausas 2006). In conclusion, differences in seed availability, post-fire germination and seedling survival indicate that effective recruitment is higher in R– than in R+.

The magnitude of the stimulation was independent of the heat index. Hence seed populations are quite homogeneous in relation to their heat-sensitivity, showing both refractory and non-refractory seeds (sensu Keeley 1991), but virtually no seeds with intermediate heat resistance. Thus, a pulse of germination is produced only when refractory seeds are exposed to a certain heat threshold, and the amount of germinated seeds depends on the proportion of refractory seeds, which is higher for R– species. Moreover, this heat threshold is different between resprouting abilities, since the probability of heat-stimulation was significantly affected by H. Indeed, at high H values, the probability of heat-stimulation decreases quicker in R+ than in R–, and the maximum probability of heat-stimulated germination is reached at slightly higher H values for R–.

The relationship found between heat-stimulated germination and life-form is consistent with the results obtained from other fire-prone ecosystems (Keeley & Bond 1997; Clarke et al. 2000; Keeley & Fotheringham 2000). Differences between life-form have also been found for other fire cues, such as smoke or charred wood (Keeley & Bond 1997; van Staden et al. 2000). We found the highest probability of heat-tolerant and heat-stimulated germination in woody plants and the lowest in perennial herbs, whereas the annuals showed an intermediate position. A predominance of species with refractory seeds has been found in the Californian chaparral and in the South African Fynbos (Keeley & Bond 1997). In these ecosystems, annuals are frequently stimulated by heat and, in many cases, also by charred wood or smoke, which are germination cues more specifically linked to fire (Keeley 1991; Keeley & Bond 1997; Keeley & Fotheringham 2000). In the Mediterranean Basin, annuals are not necessarily linked to burned sites and tend to colonize sites recently affected by any types of disturbance (Bonet & Pausas 2004; Buhk & Hensen 2006). Furthermore, the generation time of annuals is much shorter than the fire intervals in the Mediterranean Basin, and thus fire may represent a low selective pressure for these species. Consequently, the relatively low fire-enhanced germination that we found may be expected. Nevertheless, annuals are poorly represented in the data set (they represent only 10% of the species and 6% of the treatments) and, thus, future experiments with annuals may throw further light in this regard.

Perennial herbs showed the lowest probability of heat-stimulation and heat-tolerance. A possible explanation is that perennial herbs are normally resprouters (only 9% were R–), and their fire-germinative response is thus low (see comparison between R+ and R–). The mean frequency of heat-stimulation for perennial herbs is also lower than for woody plants when only R+ are considered (4 ± 14% for perennial herbs and 19 ± 24% for woody plants, expressed as mean ± SD). This low germinative response of perennial herbs to heat has been found in other fire-prone ecosystems (Keeley & Bond 1997; Clarke et al. 2000), and both the longevity of the seed bank and grazing pressure have been adduced to explain it (Clarke et al. 2000).

As in the Californian chaparral and the Fynbos (Keeley & Bond 1997), heat-stimulated germination in the Mediterranean Basin is quite common: of the 164 taxa included in our data set, 41% were heat-stimulated in at least one treatment and 28% in at least 25% of the treatments. Nevertheless, fire-stimulated germination appears to be phylogenetically aggregated in the Mediterranean Basin: stimulation was detected in 39% of the families, and 60% of the treatments that produced stimulated germination corresponded to a single family, the Cistaceae. Moreover, the few species stimulated at high H values were also Cistaceae. These results explain the high phylogenetic and phenotypic clustering found in communities under high fire frequency (Verdú & Pausas 2007; Pausas & Verdú, in press). We are aware that our data set is biased, and that the species for which heat-stimulated germination is expected were over-represented (e.g. 39% of the treatments were carried out with Cistaceae). Nevertheless, the broad taxonomic and phylogenetic spectrum included in this data set suggests that any possible bias in selection of the species studied may also represent the taxonomic aggregation of the heat-stimulated germination trait. This aggregation has been explained by the relationship between germination patterns and seed anatomy, with the latter being highly conserved at high taxonomic levels (see Keeley & Bond 1997 and references therein).

Our results highlight the importance of heat intensity for understanding post-fire germination; because it also affects resprouting ability (Moreno & Oechel 1991b; Lloret & López-Soria 1993; Cruz et al. 2003b), it is a major factor in the post-fire recovery of Mediterranean Basin ecosystems. In fact, germination ability of seeds in response to heat is missing in most global trait databases (e.g. Cornelissen et al. 2003) and could be a valuable trait to improve the quality and the predictive value of such compilations. Nevertheless, other fire-related factors may also stimulate seed germination (e.g. smoke, charred wood). These other germination cues have not yet been studied extensively and, at present, they seem to be relatively unimportant in the Mediterranean Basin (Keeley & Baer-Keeley 1999; Pérez-Fernández & Rodríguez-Echeverria 2003; Rivas et al. 2006; Crosti et al. 2006). However, given that the differences between R+ and R– observed in this paper are similar to those observed in smoke-stimulated germination for South Africa flora (van Staden et al. 2000), we expect R– species to be stimulated more by any fire-related germination cue. Further research is undoubtedly needed to understand fully the role of these mechanisms on post-fire regeneration.

Acknowledgements

We thank C. Pérez-Cervelló for his kind contribution to the database compilation, M. Arianotsou for providing access to Greek grey literature, and B. Moreira and two anonymous referees for their helpful comments and suggestions on the manuscript. This work has been financed by SINREG (REN2003-07198-C02-02/GLO) and PERSIST (CGL2006-07126/BOS) projects from the Spanish Government. CEAM is supported by Generalitat Valenciana and Bancaixa.

Ancillary