Abstract
Under climate warming, plants will undergo novel selective pressures to adjust reproductive timing. Adjustment between reproductive phenology and environment is expected to be higher in arctic and alpine habitats because the growing season is considerably short. As early- and late-flowering species reproduce under very different environmental conditions, selective pressures on flowering phenology and potential effects of climate change are likely to differ between them. However, there is no agreement on the magnitude of the benefits and costs of early- vs. late-flowering species under a global warming scenario. In spite of its relevance, phenotypic selection on flowering phenology has rarely been explored in alpine plants and never in Mediterranean high mountain species, where selective pressures are very different due to the summer drought imposed over the short growth season. We hypothesized that late-flowering plants in Mediterranean mountains should present stronger selective pressures towards early onset of reproduction than early-flowering species, because less water is available in the soil as growing season progresses. We performed selection analyses on flowering onset and duration in two high mountain species of contrasting phenology. Since phenotypic selection can be highly context-dependent, we studied several populations of each species for 2 years, covering their local altitudinal ranges and their different microhabitats. Surrogates of biotic selective agents, like fruitset for pollinators and flower and fruit loss for flower and seed predators, were included in the analysis. Differences between the early- and the late-flowering species were less than expected. A consistent negative correlational selection of flowering onset and duration was found affecting plant fitness, i.e., plants that bloomed earlier flowered for longer periods improving plant fitness. Nevertheless, the late-flowering species may experience higher risks under climate warming because in extremely warm and dry years the earlier season does not bring about a longer flowering duration due to summer drought.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
Phenology evolves to match environmental conditions and maximize individual fitness (Stanton et al. 2000; Donohue 2005). By adjusting to the abiotic and biotic environment where growth and reproduction take place, flowering phenology can dramatically affect the selective environment to which plants are exposed, with the resulting consequences on plant fitness. An adequate match between phenology and environment is especially critical in high-mountain plants because the growing season is usually very short and unpredictable (Stinson 2004; Alatalo and Totland 2003; Totland 1999; Hegland et al. 2009).
The climate in high mountains is warming up at an exceptionally high rate (Nogués-Bravo et al. 2007). This temperature increase primarily affects snowmelt and, consequently, the onset and length of the growing season (Beniston et al. 1997; Dye 2002; Inouye 2008). In fact, high-mountain plants are among those organisms for which warming is thought to be especially critical (Theurillat and Guisan 2001; Inouye et al. 2002; Thuiller et al. 2005). Upward and poleward range shifts (Grabherr et al. 1994; Klanderud and Birks 2003; Kelly and Goulden 2008) together with phenotypic plasticity and even local adaptation to novel conditions seem to be the only available alternatives to ensure long-term persistence (Jump and Peñuelas 2005; Giménez-Benavides et al. 2007b). Phenological shifts in high-mountain plants have been documented in both observational surveys (Inouye et al. 2002; Kudo and Hirao 2006; Inouye 2008) and manipulative warming experiments (Price and Waser 1998; Dunne et al. 2003). However, to our knowledge, it is not known whether these phenological shifts are the result of evolutionary changes or simple plastic responses to warming (Bradshaw and Holzapfel 2006; Burgess et al. 2007). Moreover, an assessment of the evolutionary potential of phenological shifts in high-mountain plants to face current climate change challenges is long due.
Most field observations on high-mountain plants suggest that an early-flowering strategy is always advantageous regardless of flowering timing (Galen and Stanton 1993; Kudo 1991; Stenström and Molau 1992; Totland 1994, 1997; Dunne et al. 2003). However, there is no agreement on the relative benefits of early-flowering in early- vs. late- flowering species under a global warming scenario. Some authors sustain that early-flowering species, so-called pollen-riskers, will not have a significant advantage because flowering too early may affect reproductive success due to pollinator scarcity (Molau 1993; Alatalo and Totland 2003; Totland 1997; Kudo and Hirao 2006). Moreover, flowering earlier can expose plants to a greater risk of early-summer frosts and subsequent bud mortality (Inouye 2008 and references therein). On the other hand, late-flowering species are usually exposed to risks during seed ripening due to early snow or frost events in late summer/early autumn (Molau 1993; Galen and Stanton 1995), the so-called seed-riskers. These plants may be favoured by warming scenarios because of the extended growing season (Molau 1993; Alatalo and Totland 2003; Molau et al. 2005). However, these hypotheses were mainly made on observations of arctic and temperate alpine species. The growing season may not extend in a warming world in high-mountain plants of Mediterranean climate regions if summer drought also increases (Giménez-Benavides et al. 2007a). Climate trend projections for Mediterranean mountains predict greater warming rates than those reported for other mountain regions, together with lower precipitation mainly during spring (Nogués-Bravo et al. 2008). We therefore hypothesized that more intense summer droughts in Mediterranean high mountains might impose novel selective scenarios not considered in other mountain areas. Our conceptual scheme is represented briefly in Fig. 1. We expected that early-flowering plants in Mediterranean high mountains would benefit from milder temperatures and the snowmelt water of spring - early summer, while late-flowering plants would face more stressful conditions imposed by soil water shortage and higher temperatures. In a warming scenario in Mediterranean mountains, earlier snowmelt would result in an earlier onset of both the growing season and summer drought. Thus, we hypothesized that both early- and late- flowering species would show a selective pressure towards early blooming. More specifically, we hypothesized that late-flowering species would present stronger selective pressures towards early onset of reproduction, because some individuals may not be able to flower before the arrival of summer drought.
To explore this hypothesis, we selected two co-occurring Mediterranean high-mountain specialists with contrasting flowering strategies: Armeria caespitosa, one of the first bloomers in the orophyllous cryophilic pastures above the treeline (pollen-risk species), and Silene ciliata, an extremely late-flowering plant (seed-risk species). Concerning photoperiod control, previous studies (Giménez-Benavides et al. 2007a; García-Camacho and Escudero 2009) suggest that A. caespitosa is a neutral plant that flowers immediately after snowmelt, whereas S. ciliata may be considered a genuine long-day (LD) plant that flowers only after a particular day length is reached, as recorded for other Silene species (Lang 1957; Talon and Zeewaart 1990; Donnison and Francis 1994). Each species initiates its reproductive cycle under very different environmental conditions, so that potential effects of climate warming are likely to differ (Molau 1993; Totland 1997; Alatalo and Totland 2003). Hence, these species constitute ideal model systems to test our working hypothesis.
We used an altitudinal gradient approach to evaluate how phenotypic selection on flowering time may be modulated by climate conditions, because study sites along altitudinal gradients may be a good proxy of how plants respond to warming scenarios (Körner 2007). Since pollinators and predators can act as selective agents on flowering time and exert opposing or fluctuating selection on flowering phenology (Elzinga et al. 2007; Ehrlén and Münzbergová 2009; Sandring and Ågren 2009; Gómez et al. 2009), we identified and included several sequential selective episodes in the reproductive cycle, from flower production and predation to fruit ripening and predation. With this altitudinal gradient approach and the highly contrasting flowering times observed in these two species, we aimed to assess: (1) whether shifts in flowering phenology may confer advantages to high-mountain plants, (2) whether differences in the strength and direction of selective pressures exist in early vs. late flowering species, (3) whether selective forces differ among altitudes within species, and (4) whether conflicting selection exists within populations when many potential selective agents (pollinators, predators) act simultaneously on the same phenological traits.
Materials and methods
Species and study area
Silene ciliata Poiret is a perennial cushion plant which inhabits the main Mediterranean mountain ranges in the northern half of the Mediterranean Basin reaching its southern latitudinal limit in the Iberian Central Range. Flowering stems reach 15 cm in height and bear 1–5 flowers. Although the species is self-compatible, it requires pollen vectors to produce a full seed set (Giménez-Benavides et al. 2008). A total of 11 insect species were recorded visiting S. ciliata flowers: hoverflies (Syrphidae), hawkmoths (Sphingidae), and butterflies and moths (Lepidoptera) (Giménez-Benavides et al. 2007c). Flowers are mainly predated by the blister beetle Mylabrissobrina Graëlls, 1849 (Coleoptera: Meloidae), and occasionally by nectar-thieving, florivorous ants (Hymenoptera: Formicidae) (Giménez-Benavides et al.,2008). Pre-dispersed seeds are consumed by the moth Hadena consparcatoides (Schawerda 1928) (Lepidoptera: Noctuidae), and rarely by weevils (Coleoptera: Curculionidae) (Giménez-Benavides et al. 2007c, 2008). H. consparcatoides acts as a nursery pollinator of S. ciliata (i.e. flowers are used for oviposition and as brood sites for the pollinator’s developing larvae). Therefore, this moth species can act as a mutualist or parasite depending on the trade-off between pollination services and seed predation losses and the presence of effective co-pollinators (Giménez-Benavides et al. 2007c).
Armeria caespitosa (Gómez Ortega) Boiss. in DC. is a high-mountain perennial dwarf cushion plant, endemic to the eastern Iberian Central Range, ocurring from 1,700 to 2,430 m. It grows in two interspersed but contrasting microhabitats: crevasses and ledges of granite and gneiss outcrops (hereafter, rocky microhabitat), and dry cryophilic Festuca curvifolia fellfields (hereafter, pasture microhabitat). It presents pink flowers grouped in short-scaped flowerheads. As in most species of the genus, A. caespitosa has a dimorphic self-incompatibility system (see García-Camacho et al. 2009 for more details). Flowers are visited by several species of hoverflies (Syrphidae), butterflies (Lepidoptera), honey bees and bumblebees (Hymenoptera, Apidae). Pre-dispersal seed predation is produced by larvae of Sibinia weevils (Coleoptera, Curculionidae), but there is no evidence of flower predation (García-Camacho, personal observation).
The study area was located in the Sierra de Guadarrama (Madrid, 40oN, 3oW), a W–E running mountain range characterized by a Mediterranean-type climate with an intense dry season from June to September (less than 10% of total annual rainfall) (Giménez-Benavides et al. 2007a). Vegetation in the tree line (1,900–2,000 m a.s.l.) is dominated by a low dense stand of Pinus sylvestris interspersed in a matrix of shrub species (Cytisus oromediterraneus and Juniperus communis subsp. alpina) and perennial grasses dominated by Festuca curvifolia. The Cytisus-Juniperus shrub formation is progressively replaced in altitude by the patchy xerophytic pasture dominated by F. curvifolia and several endemic perennial high-mountain cushion plants in the summit flat areas and crests (from 2,100–2,200 to 2,430 m.a.s.l.).
We selected populations of each species covering the local species’ altitudinal range. Special care was taken to include the lowest and highest populations, considering only functional populations and non-isolated individuals, as well as the different main microhabitats colonized by each species. Therefore, we selected three populations of S. ciliata growing in a pasture microhabitat (at 1,976, 2,256, and 2,428 m a.s.l., hereafter low, medium and high populations, Table S1). Previous demographic study showed that long-term persistence at the low population is critically reduced (Giménez-Benavides et al. 2010). For A. caespitosa, four populations were selected, two growing in a rocky microhabitat (at 1,750 and 2,365 m a.s.l., hereafter low rock and high rock) and two in a pasture microhabitat (at 2,000 and 2,365 m a.s.l., hereafter low pasture and high pasture), corresponding to the lowest and highest altitudinal boundaries of each microhabitat (Table S1). Approximate snowmelt date at each site was estimated from landscape digital images taken at 10-day intervals.
Data collection and determination of selective episodes
We tagged 60–160 plants in each population before the onset of flowering (see Giménez-Benavides et al. 2007a and García-Camacho and Escudero 2009 for further details on sample sizes), and tracked each plant individually every 5–8 days from flowering to seed dispersal. At each census, we recorded the number of open flowers and mature fruits, as well as the number of predated flowers and fruits. The study was performed for two consecutive years for each species, from 2002–2003 for S. ciliata and from 2004–2005 for A. caespitosa. (see Giménez-Benavides et al. 2007a and García-Camacho and Escudero 2009 for further details). Hence, we analyzed six different selective scenarios for S. ciliata (3 altitudes × 2 years), and eight selective scenarios for A. caespitosa (2 altitudes × 2 microhabitats × 2 years). The comparison between the two species may be somewhat limited because they were not monitored in the same years, due to logistic restrictions. However, this is compensated by the fact that both species included an average year (2002 and 2004) and an extreme year (2003 and 2005) in terms of high temperature and low precipitation in their study.
The following phenotypic traits were recorded for each individual: flower number (total flower production over the entire reproductive season); flowering onset (number of days from 1 day before the first flowering date in the population to the first flowering date in the individual), and flowering duration (number of days the plant remained in bloom). Previous studies showed that these phenological variables are good predictors of female reproductive success in both species (Giménez-Benavides et al. 2007a, García-Camacho and Escudero 2009). Traits were standardized separately for each population and year.
Plant fitness was estimated using individual fruit production in each season as a proxy. We identified three sequential selective episodes (k th) in the life-history stages of the species’ reproductive cycles, from initial flower production to fruit production. The fitness components (w′ k ) associated to these episodes were: w′1 = proportion of flowers surviving predation by blister beetles and nectar-thieving, florivorous ants (only for S. ciliata); w′2 = proportion of non-predated flowers developing a mature fruit (fruit set); and w′3 = proportion of mature fruits surviving predispersal seed-predation by moths and weevils. Each fitness component was relativized as the individual absolute fitness divided by the average fitness in the population (Wade and Kalisz 1989). Only individuals reaching a given transition were included in the calculation of the corresponding fitness component, so w′ k s are independent (Arnold and Wade 1984a, b). The independent method was used rather than the additive one, because the immediate aim was to determine the selection occurring at each stage independent of selection at other stages, as well as to explore potential evolutionary trade-offs (Koenig et al. 1991).
Statistical analysis
Selection during each episode was estimated by selection differentials and gradients. Selection differentials indicate the strength and direction of the total selection (s) acting on each phenotypic trait during an episode, and were estimated by the covariance between that trait and the fitness values (Lande and Arnold 1983; Arnold and Wade 1984a, b). Selection gradients reveal the strength and direction of the direct selection (β) acting on each trait during a selective episode, independent of the other traits included in the model (Lande and Arnold 1983). Selection gradients were estimated from the standardized partial regression coefficients of linear multiple regression analyses using each fitness component as the response variable and standardized trait values as explanatory variables. Selection differentials and gradients were estimated using SAS v.9.0 (SAS Institute 1996). Statistical significance of selection differentials was obtained through permutation tests with 1,000 iterations using a script in R statistical package (R Development Core Team 2010). To quantify non-linear and correlational selection, we then performed new multiple regression models including quadratic terms and/or interaction terms. However, in most scenarios the R 2-values of the selection models increased only marginally, suggesting that selection was predominantly directional. We, therefore, only report linear selection gradients with main effects.
Structural equation modeling (SEM) with the independent fitness components was used to detect the relative importance of each selective episode and to clarify the causes of selection occurring throughout the reproductive cycle (Conner 1996; Scheiner et al. 2000; Gómez 2008). An a priori saturated SEM model (Fig. 2) was constructed connecting all plant phenological traits to each independent fitness component, and these components to our surrogate of maternal fitness (i.e., fruit production). We also included flower number as a variable in our SEM models to discriminate between direct and indirect selection through covariation with flower number (Gómez 2008; Ehrlén and Münzbergová 2009). We then performed several alternative nested models, in which some of the causal paths of the saturated model were sequentially constrained to zero (i.e., the least significant paths, those reaching the highest p-values in all selective scenarios, see Table S2). The models were solved through an iterative process that used generalized least squares (GLS) as the discrepancy function, as this procedure is known to provide additional precision in parameter estimates when sample size is small (Henning Olson et al. 2000). We also used GLS estimation on the variance–covariance matrix to test the goodness of fit of the models and to calculate the Akaike Information Criterion (AIC). In order to select the best-fitting model, we chose the one with the lowest chi-square and AIC values (Shipley 2000; Grace 2006). Significance of individual path coefficients was estimated from 95% confidence intervals by bootstrapping on 2000 random samples. SEMs were performed with AMOS v.5.0.1 (Arbuckle 2003).
We calculated the effects of direct, indirect and total selection of each phenological trait on the fitness components from the selected best-fitting models. Direct selection (β′, to differentiate it from selection gradients) is any forward connection between a trait and fitness, and is equivalent in path analysis terminology to the sum of direct and indirect (mediated) effects. Significance of direct selection was also computed by bootstrapping. Indirect selection is any backward connection between a trait and fitness, including correlational effects. Finally, total selection (s′) is the sum of direct and indirect selection (Scheiner et al. 2000). Total selection (s′) is somewhat equivalent to the selection differentials (s), so we can use path analysis to partial out the selection differential into direct selection (direct + indirect effects) and indirect selection (often called non-causal effects). For further details see Scheiner et al. (2000).
Results
Flowering phenology was markedly different between the two species. Flowering onset occurred in May–June in A. caespitosa, and in July–August in S. ciliata (Table S1). Armeria caespitosa came into bloom soon after snowmelt in all populations, behaving as a typical pollen-risk bloomer. In contrast, S. ciliata came into bloom a month after snowmelt or later (Table S1). Our data suggest that S. ciliata never flowers before the end of June, even in years when snowmelt is very early (see snowmelt dates for 2003 in Table S1).
Correlations between phenotypic traits and fitness components
Flowering onset and duration in S. ciliata were negatively correlated in most cases, but this correlation varied from significant to non-significant depending on the year (Table S3). A similar pattern was found in A. caespitosa. Flowering onset and duration were negatively correlated at the lower altitudinal range for both years and microhabitats, while this correlation varied from significant to non-significant at the higher limit.
Few significant correlations were found between fitness components. In S. ciliata, w′1 was positively correlated to w′2 at the low and high populations in 2002. However, no correlation was found in extremely dry 2003. In A. caespitosa, a positive correlation was only found between w′2 and w′3 in the rock microhabitat at the higher altitude in 2004 (Table S3).
Selection differentials and gradients
Selection analyses showed that flowering onset and duration were selected during different episodes of the reproductive process, although not consistently between years, altitudes or microhabitats. In both S. ciliata and A. caespitosa, the most consistently selected trait was flowering onset at the selective episode corresponding to w′2 (non-predated flowers setting fruit) (Table 1). In contrast, at w′1 (flowers escaping predation) and w′3 (fruits escaping predation) selective episodes, selection differentials and gradients were significantly different from zero in very few selective scenarios.
Plant fitness (fruit number) showed significant selection for phenological traits in many scenarios. Flowering onset was negatively selected in half the selective scenarios of each species, whereas flowering duration was positively selected in all scenarios in S. ciliata and in six of eight scenarios in A. caespitosa (Table 1). Comparison of the significance and strength of selection differentials and gradients on plant fitness showed that selection was indirect in most cases (Table 1). Flowering onset was only under direct selection in one selective scenario of A. caespitosa (2004, low pasture), and flowering duration was only under direct selection in one case in both S. ciliata (2003, high population) and A. caespitosa (2004, low pasture). Finally, flower number greatly determined plant fitness in all scenarios, but was weakly selected for the partial fitness components (Table 1). When non-linear selection gradients were analyzed, only the quadratic term flower number substantially improved the adjustment (R 2) of linear models, but usually with lower coefficients than its linear form (results not shown).
Trade-offs and conflicting selection on phenological traits
Phenological traits were selected in opposite ways through some selection episodes depending on the combination of altitude, species and year. Thus, at low and medium altitudes of S. ciliata in 2002, selection differentials indicated that flowering onset was positively related to w′2, but had a final negative effect on plant fitness. That is, the later flowering starts, the greater fruit set is, but the lower plant fitness becomes (Table 1). In S. ciliata, selection differentials also showed that flowering duration had a robust positive effect on fitness across most selective scenarios, but on w′2 (2002, high population) and w′3 (2003, medium population) the effect was negative. Flower number also had a very strong positive effect on plant fitness, but it had a negative effect on w′1 (2002, high population), w′2 (2002, medium and high populations and 2003, high population) and w′3 (2003, medium population).
In Armeria caespitosa, opposing selection was less frequent. Flowering onset was positively related to w′3, but had a final negative effect on plant fitness (2005, low pasture), while flowering duration was negatively related to w′2, but had a positive effect on plant fitness (2004, high pasture). Finally, flower number had a general strong positive effect on plant fitness, but a negative effect on w′2 (2004, high pasture).
Structural equation models
A synthesis of the most parsimonious SEM models is shown in Fig. 2b, c and Table S2. The results of the SEM models were similar to those found with selection differentials and gradients. Of all fitness components initially considered, only the proportion of non-predated flowers developing a mature fruit (w′2) significantly and consistently affected plant fitness in both S. ciliata and A. caespitosa. The other fitness components did not significantly affect fitness in spite of presenting significant variability within populations. The most frequent and strongest paths in both species were: flowering duration to flower number, flower number to fitness, and w′2 to fitness, all of which had positive coefficients (Fig. 2b, c). Other frequent paths found in S. ciliata with negative coefficients were flowering duration and flower number to w′2. The rest of the significant paths were very context-dependent, emerging in less than half of selective scenarios in each species. SEMs also support that flowering onset and duration were under conflicting selection in the same scenarios as those found with selection differentials and gradients (Fig. 3 in appendix, where complete path diagrams and model results of each scenario are shown).
Direct, indirect and total selection emerging from the most parsimonious SEM models is shown in Table 2. Both species showed positive total selection on flowering duration, usually with moderate to large values. In S. ciliata, direct selection (direct + indirect effects) on flowering onset was always non-significant and very weak. However, indirect and especially total selection had high negative values in 2002, but it is worthy to note that these cannot be statistically tested. In A. caespitosa, direct selection on flowering onset was significantly positive in only two scenarios (2004, low pasture and 2004, high rock, mainly by its indirect effects through flower number). However, once again, indirect and total selection resulted in negative values in most scenarios, especially at low populations in 2005.
Discussion
The understanding of phenotypic selection on flowering phenology is essential for evaluating the potential risks of climate warming, although it has scarcely been explored in high-mountain plants (Stinson 2004; Alatalo and Totland 2003; Sandring et al. 2007). To our knowledge, this is the first study integrating two high mountain plants of contrasting flowering timing in a suite of spatial and temporal selective scenarios.
Results obtained across the different altitudes and microhabitats did not provide a clear pattern. Thus, selection responses for earlier flowering were not stronger at the lower places (i.e., with higher temperature and lower precipitation) as we might have expected. This implies that other local factors, in addition to altitude and microhabitat, affect reproductive success (Giménez-Benavides et al. 2007a, García-Camacho and Escudero 2009). On the other hand, the effects of temporal variation seemed to predominate over those of spatial variation. The pattern observed in the climatically average years (2002 and 2004) adjusted well to what is expected to happen in high-mountain environments. In general terms, there was a selection towards earlier onset of flowering in both species due to correlational selection through flowering duration (Tables 1 and 2). Nevertheless, the limiting condition brought about by summer drought that we had hypothesized was clearly seen in the late flowering species in 2003, an extreme year with high temperatures and low precipitation (Schär et al. 2004). In this year, snowmelt date occurred approximately 22 days earlier than in 2002, an average year, and flowering onset was also early (Table S1). At the same time, flowering duration was notably lower at all altitudes due to the harsh summer drought. Consequently, earlier flowering onset was not followed by longer flowering duration (Table 2) and early flowering was not selected for plant fitness (Table 1). Thus, early-flowering plants could not take advantage of prolonged flowering in this year (Giménez-Benavides et al. 2007a). In a similar way, a lack of correlation was also found at the high-altitude sites in A. caespitosa in 2005, when high temperature and low precipitation patterns were reproduced. In contrast to the favourable outlook for alpine plants (Molau 1993; Alatalo and Totland 2003; Molau et al. 2005), these results support the idea that some plants of Mediterranean high mountains may be highly vulnerable to extreme episodes related to climate warming.
The response of our two model species was more similar between them than expected. We hypothesized that early-flowering species would experience lower selective pressure towards earlier flowering onset because flowering too early might increase pollen limitation due to lower insect activity (Kudo 1991, 1993; Molau 1993; Kudo and Hirao 2006), and because drought would not be a strong selective pressure at the beginning of the growing season, even under Mediterranean conditions. However, our pollen-risk plant did not show evidence of pollen limitation at the beginning of the flowering season (expressed by direct positive effects of flowering onset on fruit set), nor evidence of frost damage during our study period (personal observation). Thus, if our observations were representative of the general pattern experienced by this species over the years, A. caespitosa could benefit from longer growing seasons in a climate warming context. On the other hand S. ciliata, the late-flowering species, also experienced selective pressures towards early-flowering phenotypes. However, its photoperiodic control on flowering time might preclude the use of the most favourable part of the summer for reproduction in the forthcoming warming conditions. Consequently, the different phylogenetic background of the two species affects the ability of the species to respond to these selective episodes.
Our findings also showed that w′2 (surviving flowers setting mature fruits), a surrogate of pollinator success, is an effective selective episode in both species and the only fitness component from those initially studied that consistently affected plant fitness. It is well known that pollinators may drive the evolution of floral display and flowering phenology (Elzinga et al. 2007; Sandring and Ågren 2009). As expected, w′2 had a positive, direct effect on plant fitness in all scenarios. Nevertheless, determining how phenological traits are selected through w′2 is more complex and conflicting. In some scenarios, especially in S. ciliata, selection through w′2 favoured late flowering onset, short flowering duration and the production of few flowers. This suggests that pollen limitation was greater early in the flowering season, whereas low fruitset in plants with many flowers may be due to resource limitation and/or greater fruit abortion due to geitonogamous crossings (Albert et al. 2008). On the other hand, flowering onset, duration and flower number also exerted directional selection on plant fitness, sometimes leading to a suite of opposing selective forces (i.e. plants with early flowering onset, long duration and production of more flowers sometimes had lower w′2, but in absolute terms produced more fruits). In A. caespitosa, however, the constant positive effect of w′2 on plant fitness was almost independent of phenology and floral display (except in one of eight scenarios), suggesting that pollinator activity does not play a major role in determining flowering phenology in this plant.
Predator-mediated selection may also operate on flowering phenology evolution (Pilson 2000; Elzinga et al. 2007; Ehrlén and Münzbergová 2009; Sandring and Ågren 2009). As biotic interactions and selective forces may differ between altitudes (Gómez and Zamora 2000; Totland 2001; Sandring et al. 2007; Gómez et al. 2009), we expected differences in this fitness component in our studied populations. However, our results did not reveal significant predator-mediated selection on flowering phenology or floral display in either species. In S. ciliata 14% of plants were attacked by florivores and 8% by fruit predators, while in A. caespitosa 19% of plants were attacked by weevils (mean values across populations and years). However, predation events were fairly evenly distributed throughout the growing season instead of being concentrated in early or late periods, precluding the selection of early or late-flowering individuals.
Our analysis shows an important correlational component of indirect selection between flowering onset and duration that markedly influenced net effects on plant fitness. This correlation was negative and very consistent across species and was responsible for the strength and direction of indirect and total selection as shown in our SEM models (Table 1). In many scenarios, while direct selection (β′) indicated a weak phenotypic selection on flowering onset, total selection (s′) became negative for this trait after incorporating correlational effects (Table 2). This means that strong phenotypic selection towards early flowering onset might be explained by indirect selection processes. Our SEM diagrams graphically show that constant, moderate selection occurs through a preferential pathway: the negative effect of flowering onset on duration, the positive effect of duration on flower number and, finally, the positive effect of flower number on plant fitness (Fig. 2b, c). As Elzinga et al. (2007) pointed out, the common expectation of stabilizing selection on flowering phenology assumes that, although individuals vary in their first flowering date, they do not vary in their flowering duration. However, when the date of flowering onset is negatively correlated to flowering duration, correlated directional selection takes place.
Our observational approach could have detected spatial variation in selective pressures along the altitudinal gradient and microhabitats considered, indicating local variation in how plants adapt to environmental conditions that vary in space and time. Although phenological traits affected fitness in different ways among sites and years in these two species (see also Giménez-Benavides et al. 2007a; García-Camacho and Escudero 2009), our results suggest that this variation is not very relevant in evolutionary terms. In general, the earlier flowering started, the greater fitness was in both species, and in those scenarios where this pattern was not present, total selection on flowering onset was negligible (Tables 1, S3). Thus, selection of phenological traits was not fluctuating but directional in both species.
In conclusion, our results suggest that flowering onset and duration in these high-mountain Mediterranean species are under selection through an indirect path affecting flower number. When flowering starts earlier, flowering duration is longer and flower number is higher, resulting in a greater number of fruits. Thus, shifts in phenology seem to confer advantages to our model species. Phenotypic selection towards early flowering is consistent in both the early- and late-flowering species studied, but varies in intensity and is context-dependent. We did not detect conflicting selection because w′2 (surviving flowers setting mature fruits) was the only fitness components initially considered that consistently affected fitness. Future studies should assess whether the selective advantage of early-flowering individuals due to the significant correlation with flowering duration is strong enough to cause the predominance of early-flowering phenotypes in nature. Furthermore, our hypothesis of phenotypic selection on flowering timing should be tested in other high mountain Mediterranean species to explore the evolutionary potential of plants to overcome climate warming.
References
Alatalo JM, Totland Ø (2003) Response to simulated climatic change in an alpine and subarctic pollen-risk strategist, Silene acaulis. Glob Change Biol 3:74–79
Albert MJ, Iriondo JM, Escudero A et al (2008) Dissecting components of flowering pattern: size effects on female fitness. Bot J Linn Soc 156:227–236
Arbuckle J (2003) Amos 5 user’s guide supplement. AMOS Development Corporation, Spring House, PA
Arnold SJ, Wade MJ (1984a) On the measurement of natural and sexual selection: Applications. Evolution 38:720–734
Arnold SJ, Wade MJ (1984b) On the measurement of natural and sexual selection: Theory. Evolution 38:709–719
Beniston M, Diaz HF, Bradley RS (1997) Climatic change at high elevation sites: an overview. Clim Change 36:233–251
Bradshaw WE, Holzapfel CM (2006) Evolutionary response to rapid climate change. Science 312:1477–1478
Brody AK (1997) Effects of pollinators, herbivores and seed predators on flowering phenology. Ecology 78:1624–1631
Burgess KS, Etterson JR, Galloway LF (2007) Artificial selection shifts flowering phenology and other correlated traits in an autotetraploid herb. Heredity 99:641–648
Conner JK (1996) Understanding natural selection: an approach integrating selection gradients, multiplicative fitness components, and path analysis. Ethol Ecol Evol 8:387–397
Donnison IS, Francis D (1994) Experimental control of floral reversion in isolated shoot apices of the long-day plant Silene coeli-rosa. Physiol Plant 92:329–335
Donohue K (2005) Niche construction through phenological plasticity: life history dynamics and ecological consequences. New Phytol 166:83–92
Dunne JA, Harte J, Taylor KJ (2003) Subalpine meadow flowering phenology responses to climate change: integrating experimental and gradient methods. Ecol Monogr 73:69–86
Dye DG (2002) Variability and trends in the annual snow cover cycle in Northern Hemisphere land areas, 1972–2000. Hydrol Proc 16:3065–3077
Ehrlén J, Münzbergová Z (2009) Timing of flowering: Opposed selection on different fitness components and trait covariation. Am Nat 173:819–830
Elzinga JA, Atlan A, Biere A et al (2007) Time after time: flowering phenology and biotic interactions. Trends Ecol Evol 22:432–439
Galen C, Stanton ML (1993) Short-term responses of alpine buttercups to experimental manipulations of growing season length. Ecology 74:1052–1058
Galen C, Stanton ML (1995) Responses of snowbed plant species to changes in growing season length. Ecology 76:1546–1557
García-Camacho R, Escudero A (2009) Reproduction of an early-flowering Mediterranean mountain narrow endemic (Armeria caespitosa) in a contracting mountain island. Plant Biol 11:515–525
García-Camacho R, Méndez M, Escudero A (2009) Pollination context effects in the high-mountain dimorphic Armeria caespitosa (Plumbaginaceae): neighborhood is something more than density. Am J Bot 96:1620–1626
Giménez-Benavides L, Dötterl S, Jürgens A et al (2007a) Generalist diurnal pollination provides greater fitness in a plant with nocturnal pollination syndrome: assessing the effects of a Silene-Hadena interaction. Oikos 116:1461–1472
Giménez-Benavides L, Escudero A, Iriondo JM (2007b) Reproductive limits of a late-flowering high-mountain Mediterranean plant along an elevational climate gradient. New Phytol 173:367–382
Giménez-Benavides L, Escudero A, Iriondo JM (2007c) Local adaptation enhances seedling recruitment along an altitudinal gradient in a high mountain Mediterranean plant. Ann Bot 99:723–734
Giménez-Benavides L, Escudero A, Iriondo JM (2008) What shapes the altitudinal range of a high mountain Mediterranean plant? Recruitment probabilities from ovule to seedling stage. Ecography 31:731–740
Giménez-Benavides L, Albert MJ, Iriondo JM et al. (2010) Demographic processes of upward range contraction in a long-lived Mediterranean high mountain plant. Ecography (in press)
Gómez JM (2008) Sequential conflicting selection due to multispecific interactions triggers evolutionary trade-offs in a monocarpic herb. Evolution 62:668–679
Gómez JM, Zamora R (2000) Spatial variation in the selective scenarios of Hormathophylla spinosa (Cruciferae). Am Nat 155:657–668
Gómez JM, Perfectti F, Bosch J et al (2009) A geographic selection mosaic in a generalized plant-pollinator-herbivore system. Ecol Monogr 79:245–263
Grabherr G, Gottfried M, Pauli H (1994) Climate effects on mountain plants. Nature 369:448
Grace JB (2006) Structural equation modeling and natural systems. Cambridge University Press, Cambridge
Hegland SJ, Nielsen A, Lázaro A et al (2009) How does global warming affect plant-pollinator interactions? Ecol Lett 12:184–195
Henning Olson U, Foss T, Troye SW et al (2000) The performance of ML, GLS, and WLS estimation in structural equation modeling under conditions of misspecification and nonnormality. Struct Equ Model 7:557–595
Inouye D (2008) Effects of climate change on phenology, frost damage and floral abundance of montane wildflowers. Ecology 89:353–362
Inouye DW, Morales MA, Dodge GJ (2002) Variation in timing and abundance of flowering by Delphinium barbeyi Huth (Ranunculaceae), the roles of snowpack, frost and La Niña, in the context of climate change. Oecologia 130:543–550
Jump AS, Peñuelas J (2005) Running to stand still: adaptation and the response of plants to rapid climate change. Ecol Lett 8:1010–1020
Kelly AE, Goulden ML (2008) Rapid shifts in plant distribution with recent climate change. Proc Natl Acad Sci USA 105:11823–11826
Klanderud K, Birks HJ (2003) Recent increases in species richness and shifts in altitudinal distributions of Norwegian mountain plants. Holocene 13:1–6
Koenig WD, Albano SS, Dickinson JL (1991) A comparison of methods to partition selection acting via components of fitness: do larger male bullfrogs have greater hatching success? J Evol Biol 4:309–320
Körner C (2007) The use of ‘altitude’ in ecological research. Trends Ecol Evol 22:569–574
Kudo G (1991) Effects of snow-free period on the phenology of alpine plants inhabiting snow patches. Arct Antarct Alp Res 23:436–443
Kudo G (1993) Relationship between flowering time and fruit set of the entomophilous alpine shrub, Rhododendron aureum (Ericaceae), inhabiting snow patches. Am J Bot 80:1300–1304
Kudo G, Hirao AS (2006) Habitat-specific responses in the flowering phenology and seed set of alpine plants to climate variation: implications for global-change impacts. Popul Ecol 48:49–58
Lande R, Arnold SJ (1983) The measurement of selection on correlated characters. Evolution 37:1210–1226
Lang A (1957) The effect of gibberellin upon flower formation. Proc Natl Acad Sci USA 43:709–717
Molau U (1993) Relationships between flowering phenology and life-history strategies in tundra plants. Arct Alp Res 25:391–402
Molau U, Nordenhall U, Eriksen B (2005) Onset of flowering and climate variability in an alpine landscape: A 10-year study from Swedish Lapland. Am J Bot 92:422–431
Nogués-Bravo D, Araújo MB, Martinez-Rica JP et al (2007) Exposure of global mountain systems to climate change during the 21st century. Glob Environ Change 17:420–428
Nogués-Bravo D, Araújo MB, Lasanta T et al (2008) Climate change in Mediterranean mountains during the XXIst century. Ambio 37:380–385
Pilson D (2000) Herbivory and natural selection on flowering phenology in wild sunflower, Helianthus annuus. Oecologia 122:72–82
Price MV, Waser NM (1998) Effects of experimental warming on plant reproductive phenology in a subalpine meadow. Ecology 79:1261–1271
Sandring S, Ågren J (2009) Pollinator-mediated selection on floral display and flowering time in the perennial herb Arabidopsis lyrata. Evolution 63:1292–1300
Sandring S, Riihimäki M, Savolainen O et al (2007) Selection on flowering time and floral display in an alpine and a lowland population of Arabidopsis lyrata. J Evol Biol 20:558–567
SAS Institute Inc (1996) SAS/STAT software: changes and enhancements through release 6.11. Cary, NC
Schär C, Vidale PL, Luthi et al (2004) The role of increasing temperature variability in European summer heatwaves. Nature 427:332–336
Scheiner SM, Mitchell RJ, Calahan HS (2000) Using path analysis to measure natural selection. J Evol Biol 13:423–433
Shipley B (2000) Cause and correlation in biology. Cambridge University Press, Cambridge
Stanton ML, Roy BA, Thiede DA (2000) Evolution in stressful environments. I. Phenotypic variability, phenotypic selection and response to selection in five distinct environmental stresses. Evolution 54:93–111
Stenström M, Molau U (1992) Reproductive ecology of Saxifraga oppositifolia: phenology, mating system and reproductive success. Arct Antarct Alp Res 24:337–343
Stinson KA (2004) Natural selection favors rapid reproductive phenology in Potentilla pulcherrima (Rosaceae) at opposite ends of a subalpine snowmelt gradient. Am J Bot 91:531–539
Talon M, Zeewaart JAD (1990) Gibberelins and stem growth as related to photoperiod in Silene armeria L. Plant Physiol 92:1094–1100
R Development Core Team (2010) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing Vienna, Austria. Available from http://www.R-project.org (Accessed July 2010)
Theurillat JP, Guisan A (2001) Potential impact of climate change on vegetation in the European Alps: a review. Clim Change 50:77–109
Thuiller W, Lavorel S, Araújo MB et al (2005) Climate change threats to plant diversity in Europe. Proc Natl Acad Sci USA 102:8245–8250
Totland Ø (1994) Intraseasonal variation in pollination intensity and seed set in an alpine population of Ranunculus acris in southwestern Norway. Ecography 17:159–165
Totland Ø (1997) Effects of flowering time and temperature on growth and reproduction in Leontodon autumnalis var. taraxaci a late-flowering alpine plant. Arct Alp Res 29:285–290
Totland Ø (1999) Effects of temperature on performance and phenotypic selection on plant traits in alpine Ranunculus acris. Oecologia 120:242–252
Totland Ø (2001) Environment-dependent pollen limitation and selection on floral traits in an alpine species. Ecology 82:2233–2244
Wade MJ, Kalisz S (1989) The additive partitioning of selection gradients. Evolution 43:1567–1569
Acknowledgments
The authors thank the staff of Parque Natural de las Cumbres, Circo y Lagunas de Peñalara for permission to work in the area and C. F. Aragón, A. L. Luzuriaga and M. J. Albert for helping with field work. The authors also thank D. Palacios for providing the digital images for snowmelt date estimation, K. Donohue and M. de la Cruz for providing SAS and R scripts, respectively, and L. De Hond for linguistic assistance. This work was supported by projects ISLAS (CGL2009-13190-C03-01), SIL-HAD (CGL2009-08755) and LIMITES (CGL2009-07229) funded by the Ministerio de Ciencia e Innovación (Spain) and REMEDINAL2 funded by the Comunidad de Madrid.
Author information
Authors and Affiliations
Corresponding author
Electronic supplementary material
Below is the link to the electronic supplementary material.
Rights and permissions
About this article
Cite this article
Giménez-Benavides, L., García-Camacho, R., Iriondo, J.M. et al. Selection on flowering time in Mediterranean high-mountain plants under global warming. Evol Ecol 25, 777–794 (2011). https://doi.org/10.1007/s10682-010-9440-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10682-010-9440-z