Temporal and intraclonal variation of flowering and pseudovivipary in Poa bulbosa

Abstract

Background and Aims

Versatility in the reproductive development of pseudoviviparous grasses in response to growth conditions is an intriguing reproduction strategy. To better understand this strategy, this study examined variation in flowering and pseudovivipary among populations, co-occurring clones within populations, and among tillers in individual clones of Poa bulbosa, a summer-dormant geophytic grass that reproduces sexually by seed, and asexually by basal tiller bulbs and bulbils formed in proliferated panicles.

Methods

Clones were collected from 17 populations across a rainfall gradient. Patterns of reproduction were monitored for 11 years in a common garden experiment and related to interannual differences in climatic conditions. Intraclonal variation in flowering and pseudovivipary was studied in a phytotron, under daylengths marginal for flowering induction.

Key Results

Clones showed large temporal variability in their reproductive behaviour. They flowered in some years but not in others, produced normal or proliferated panicles in different years, or became dormant without flowering. Proliferating clones did not show a distinct time sequence of flowering and proliferation across years. Populations differed in incidence of flowering and proliferation. The proportion of flowering clones increased with decreasing rainfall at the site of population origin, but no consistent relationship was found between flowering and precipitation in the common garden experiment across years. In contrast, flowering decreased at higher temperatures during early growth stages after bulb sprouting. Pulses of soil fertilization greatly increased the proportion of flowering clones and panicle production. High intraclonal tiller heterogeneity was observed, as shown by the divergent developmental fates of daughter plants arising from bulbs from the same parent clone and grown under similar conditions. Panicle proliferation was enhanced by non-inductive 8 h short days, while marginally inductive 12 h days promoted normal panicles.

Conclusions

Interannual variation in flowering and proliferation in P. bulbosa clones was attributed to differences in the onset of the rainy season, resulting in different daylength and temperature conditions during the early stages of growth, during which induction of flowering and dormancy occurs.

INTRODUCTION

In some temperate grasses, marginal conditions for flowering induce the development of plantlets and bulbils instead of normal florets in the inflorescence. These vegetative structures can establish and grow as separate plants under suitable conditions (Wycherley, 1953; Youngner, 1960; Beetle, 1980; Heide, 1988). This form of asexual reproduction is considered as a type of pseudovivipary, to distinguish it from true vivipary in which the plantlets develop directly from sexually or asexually produced embryos (Elmqvist and Cox, 1996). In this context, proliferation refers to the conversion of florets into plantlets in the grass inflorescence (Latting, 1972; Beetle, 1980; Ahmad et al., 2009). Proliferative plantlets from different species differ in morphological and physiological characteristics. In some grasses (e.g. Festuca vivipara), plantlets do not dehisce and continue to grow on the inflorescence until the culm collapses (Lee and Harmer, 1980). They are susceptible to desiccation, but may root and establish daughter plants under wet conditions. In contrast, plantlets in Poa bulbosa grow until bulbil formation and subsequently they dry and dehisce. These bulbils tolerate desiccation, are dormant during the dry summer and disperse by wind and ants (Youngner, 1960; Heyn, 1971; Lee and Harmer, 1980; Tsukanova, 1995). Seed-producing and viviparous forms of temperate grasses may represent the phenotypic expression of different reproductive traits under changing environmental conditions affecting the flowering process (Evans, 1964; Heide, 1988, 1989). This assumption is supported by several facts. Seminiferous and viviparous forms frequently coexist in the same area, as in F. vivipara and F. ovina (Wycherley, 1953; Heide, 1988; Chiurugwi et al., 2011), Poa alpina and P. alpina var. vivipara (Latting, 1972), Poa alopecurus ssp. alopecurus and P. alopecurus ssp. fuegiana (Moore and Doggett, 1976), and P. bulbosa (Heyn, 1971). Furthermore, transitions between pseudovivipary and normal seed production are often observed in the field in association with seasonal shifts in climatic conditions (Wycherley, 1954), as well as between years (Heyn, 1971; Harmer, 1984). This facultative pseudovivipary trait is under both genetic and environmental control (Chiurugwi et al., 2011), as shown experimentally by growing contrasting forms in a common environment or under controlled conditions (Youngner, 1960; Junttila, 1985; Heide, 1988, 1989). The balance between clonal and sexual reproduction is controlled mainly by daylength and temperature (Wycherley, 1952; Junttila, 1985; Heide, 1988, 1989; Ofir and Kigel, 2003). Generally, short days (SDs) and low temperature promote inflorescence proliferation, while long days and higher temperature lead to normal flowering and seed production (Youngner, 1960; Evans, 1964; Heide, 1988; Ofir and Kigel, 2003). Thus, pseudovivipary can be induced by conditions sub-optimal for flowering, while in proliferating forms sexual reproduction can be restored under conditions optimal for flowering (Heide, 1989). However, although versatility in the reproductive behaviour of pseudoviviparous grasses in response to growth conditions is well known, stability of flowering and proliferation across years in co-occurring clones within populations, as well as among tillers within a clone, has been less studied. Here we examined the interannual and intraclonal stability of flowering and pseudovivipary in clones of P. bulbosa grown in a common garden experiment and under controlled conditions.

Poa bulbosa is a summer-dormant, small geophytic grass with a compound reproductive strategy – it can reproduce sexually by seed, as well as asexually by basal bulbs developing in non-flowering tillers, and by bulbils produced in proliferated panicles (Heyn, 1962, 1971; Ofir and Kigel, 2003). Basal bulbs and bulbils are dormant during the summer and sprout at the onset of the winter rainy season. Annual production of basal bulbs allows continued spatial occupancy and lateral spread in colonized patches (Sheffer et al., 2007). In proliferating forms, bulbils detach and disperse from the parent plant. Field records show that non-flowering plants, as well as seed-producing and bulbil-producing flowering plants, co-occur in natural populations, although their relative frequency varies in different sites and across years (Heyn, 1971). To better understand the factors involved in the reproductive versatility of P. bulbosa, we studied variation in patterns of reproductive development at the population, clonal and intraclonal level. To this end, we examined (1) interannual stability of flowering and pseudovivipary in individual clones collected from populations along a rainfall gradient and grown in a common environment; and (2) intraclonal variation of these traits in daughter plants from pseudoviviparous clones in response to marginal induction of flowering under controlled conditions.

MATERIALS AND METHODS

Plant material

Fifteen populations of Poa bulbosa were sampled during April–May 2000, in sites along a South–North rainfall gradient in Israel (100–810 mm year⁻¹), latitudes 30°52′ to 33°03′N and 50–600 m altitude. Two additional populations were sampled in Thermessos, Antalya, Turkey (36°58′N, 1200 mm rain year⁻¹, 1000 m) and in Sardinia, Italy (40°06′N, 1100 mm rain year⁻¹, 450 m) (Table 1). In each location, 10–15 ‘tufts’ representing different clones were collected and planted in loess soil in 4 L pots. Clones were kept under similar outdoor conditions in a net-house at the Faculty of Agriculture in Rehovot, Israel (31°54′N, 530 mm rain year⁻¹, 60 m altitude). The net transmits 75 % of daylight, is freely permeable to rain and has little effect on air temperature.

Table 1.

Locations where Poa bulbosa populations were sampled, type of soil and long-term annual rainfall

Population code	Location	Soil	Rainfall (mm year−1)
19	Sde Boquer	Stony loess	100
8	Road Arad to Dead Sea	Stony loess	110
10	Mitspe Jericho	Stony calcareous lithosol	120
5	Tel Arad	Stony loess	180
11	Maale Edumim	Terra rossa	250
4	Har Amasa, plain	Loessial alluvium	220
1	Har Amasa, hilltop	Stony loess	276
2	Har Amasa, foothill	Stony loess	276
3	Har Amasa, ruins	Stony loess	276
17	Lehavim	Stony loess	307
15	Nahshon	Rendzina	514
13	Karei Deshe, meadow	Basaltic	520
18	Tirat Shalom	Calcareous sandstone	530
12	Park Rabin, hilltop	Terra rossa	620
14	Montfort, terraces	Terra rossa	810
16	Sardinia, Italy	Rendzina	1100
20	Antalya, Turkey	Terra rossa	1200

Experiments and growth conditions

Net-house experiment

Between- and within-population differences in patterns of flowering and panicle proliferation across years were studied by monitoring individual clones over 11 growing seasons (2000/01 to 2010/11). Flowering, number of panicles and number of proliferated panicles were recorded for each clone. Slow-release fertilizer (Osmocote-Plus, 2 g per pot) was added in October 2006 and again in October 2009, just before the rainy season and onset of sprouting. Monthly changes in daylength and interannual variation in rainfall and temperature conditions during the growing season are shown in Fig. 1. The growing season is defined as the period between the first effective rain event inducing bulb sprouting (>15 mm, early November to mid-December), and the last rain event before plants became dormant between late March and early April. Onset of the rainy season is the date of the first effective rain event. Climatic data are from the nearby Bet Dagan meteorological station (Israel Meteorological Service http://www.ims.gov.il/ims/).

Fig. 1.

Climatic conditions at the site of the common garden experiment. (A) Monthly changes in daylength and maximum–minimum temperatures. Temperature values are averages for the experimental period (2000/01–2010/11). Bars indicate the maximal deviation of temperature from the average in years with the warmest and coldest winters, respectively. Rectangle indicates the range of onset of the growing season during 2000/01–2010/11, as triggered by the first effective rainfall event (≥15 mm). (B) Amount of rainfall during the growing season and duration of the growing season. The growing season is the period between the first effective rainfall event inducing bulb sprouting (>15 mm, early November to mid-December) and the last rain event before plants become dormant (late March to early April). (C) Max–Min temperature during the 8 weeks following the onset of the rainy season after the first effective rainfall. (D) Date of onset of the rainy season.

Phytotron experiment

Plants were grown in glass-enclosed growth rooms transmitting approx. 80 % of solar radiation, at 22/16 ± 0·5 °C (day/night) and 65/75 % (day/night) relative humidity; conditions favourable for P. bulbosa growth (Ofir and Kigel, 1999). Day temperature and humidity were maintained between 0700 and 1600 h. Plants were irrigated alternately with tap water and 50 % Hoagland's nutrient solution, once every 2 d. Two photoperiods were used: SDs of 9 h of natural daylight (0700–1600 h), and a daylength of 12 h consisting of 9 h of natural light (0700–1600 h) plus 3 h (0530–0700 and 1600–1730 h) of supplementary lighting [12 μmol m⁻² s⁻¹ photosynthetically active radiation (PAR) at plant level], supplied by 100 W krypton incandescent lamps (General Electric, Hungary) and 100 W white daylight fluorescent lamps (Omega, XCD, China). A 12 h daylength was chosen since it is just above the critical daylength for flowering induction in P. bulbosa, and marginally inductive for dormancy at 22/17 °C (Ofir and Kigel, 1999). Thus, it allowed a longer period of growth and flowering expression before dormancy imposition.

Twelve proliferating clones were randomly chosen from populations 4, 8 and 12 characterized by high flowering frequency in the net-house (Table 1), four clones per population. Resting bulbs from the selected clones were detached from dormant plants in the net-house on 24 April 2010 and dry-stored for 1 month in paper bags at 38 °C, to reduce bulb dormancy and enhance sprouting after planting (Ofir and Kigel, 2007). Bulbs were planted in 0·8 L pots in a substrate of vermiculite, sphagnum peat and volcanic tuff-gravel (1:1:2 v/v/v), three bulbs per pot. Pots were kept for 1 week at 10 °C until bulb sprouting and subsequently transferred to 22/16 °C and SDs, conditions that are non-inductive for flowering and dormancy. Daughter plants developing from the bulbs were thinned down to one plant per pot and allocated to different photoperiodic treatments when 2 weeks old.

In temperate grasses, exposure to shorter days promotes panicle proliferation (Heide, 1988, 1989). Thus, intraclonal variation in flowering and proliferation was studied by subjecting daughter plants taken from the individual clones to 12 h photoperiods that in P. bulbosa are marginal for flowering induction (Ofir and Kigel, 2006). Daughter plants were either kept continuously under non-inductive SDs or assigned to treatments differing in duration of exposure to 12 h daylength: one exposure to 15 or 30 d of 12 h photoperiods and returned to SDs; or subjected to an additional continued exposure to 12 h photoperiods after 30 or 45 d under SDs. Intercalation of non-inductive SDs between the marginally inductive photoperiods was designed to increase proliferation. Thus, treatments were: (a) continuous SDs; (b) 15 d SDs → 15 d of 12 h → continuous SDs; (c) 15 d SDs → 30 d of 12 h → continuous SDs; (d) 15 d SDs → 15 d of 12 h → 45 d of SDs → continuous 12 h; and (e) 15 d SDs → 30 d of 12 h → 30 d of SDs → continuous 12 h.

Each treatment included between seven and ten daughter plants from each clone. Plants were individually monitored until they became dormant, with most leaves yellowing or drying. Time to flowering was recorded at the emergence of the first panicle. Daughter plants were classified into three categories: (1) flowering, non-proliferated; (2) flowering-proliferated, and (3) non-flowering, until dormancy or termination of the experiment.

Statistical analyses

Experiments were arranged in a completely randomized design. Analysis of variance (ANOVA) and regression analyses were carried out with the JMP Software (SAS Institute Inc., Cary, NC, USA). Analyses of percentage data were carried out after arcsine transformation.

RESULTS

Variation in flowering and proliferation among populations and across years

Flowering

Poa bulbosa populations collected along the rainfall gradient differed widely in the flowering potential of their clones, as revealed by continued monitoring for 11 years in the net-house. Regression analysis showed that the proportion of flowering clones (i.e. flowered at least once during 11 years), as well as the frequency of flowering of individual clones across years (i.e. interannual flowering stability) decreased with increasing rainfall at the site of population origin (R² = 0·311, P = 0·02 and R² = 0·351, P = 0·02, respectively, d.f. = 15) (Fig. 2). Furthermore, populations with a higher proportion of flowering clones also had clones with a higher frequency of flowering (R² = 0·567, P = 0·002, d.f. = 13). Of the 17 studied populations, only the Montfort population (810 mm rain year⁻¹) did not flower in the net-house.

Fig. 2.

Interannual flowering stability: variation in the frequency with which clones of the different populations flowered across years (2001–2011) as a function of the amount of annual rainfall at the site of population origin.

Large interannual variation was found among populations in the proportion of clones reaching flowering (Fig. 3). The average proportion of flowering clones gradually decreased between 2001 and 2006, from 40–50 % to 10 % (P = 0·01). Pulses of soil fertilization at the onset of the 2006/07 and 2009/10 growing seasons greatly increased the proportion of flowering clones, and this effect was maintained for at least 2 years (Fig. 3). Populations in which clones did not flower in the year preceding fertilization – eight populations in 2006 and six populations in 2009 – had flowering clones after soil fertilization. The number of panicles produced by clones differed greatly among populations, with a range of 0–14·3 panicles per clone. As for flowering, panicles per clone also decreased with increasing rainfall at the site of population origin, but the relationship was only marginally significant (R² = 0·194, P = 0·08, d.f. = 15). Production of panicles across years decreased gradually from 3–4 panicles per clone to >1 in 2005/06 (P = 0·05) (Fig. 3). Soil fertilization in 2007 strongly increased panicle production, but this was also a transient effect that decayed after 2 years. Subsequent soil fertilization in 2009 again increased panicle production.

Fig. 3.

Interannual variation in the proportion of clones attaining flowering and in the number of panicles per clone. Values are averages for 15 populations that flowered at least once during 11 years in the net-house. Soil fertilization was applied before the onset of the rainy season in 2006 and 2009. Means ± s.e.

Analysis of the relationships between climatic conditions of the common garden in different years and the proportion of flowering clones and panicles per clone showed no consistent trends with seasonal precipitation (range: 276–735 mm) or with the duration of the growing season (range: 75–135 d). In contrast, the proportion of flowering clones among populations decreased with higher temperatures during the period following sprouting in the net-house (R² = 0·435, P = 0·05) (Fig. 4). In fact, the timing of the first effective rain event determined the temperature and daylength conditions encountered by the plants during subsequent early growth. Onset of the rainy season gradually shifted to earlier dates between 2000/01 and 2010/11 (Fig. 1D). Consequently, the average daily maximum–minimum temperature during the 2 months following sprouting steadily increased during these years (Fig. 1C). As a result, bulb sprouting advanced from late December to mid-November, and panicle emergence occurred 2–3 months later, from early March to mid-January. Accordingly, daylength during the period between rainfall onset and panicle emergence varied from 11·5 to 10 h in different years, and the average maximum–minimum temperature ranged between 23–12 °C and 18–8 °C, with annual deviations of up to 2–3 °C from the average (Fig. 1). Since flowering induction and initial panicle development in P. bulbosa occur shortly after sprouting (M. Ofir and J. Kigel, unpubl res.), the higher temperature probably impaired these processes, even leading to panicle proliferation. Conversely, soil fertilization strongly increased the proportion of flowering clones and the number of panicles per clone. Nevertheless, enhancement of flowering by soil fertilization was lower in the 2009/10 season (P = 0·05), during which temperatures after sprouting were higher than in the 2006/07 season (Fig. 4).

Fig. 4.

Relationship between the percentage of flowering clones and average daily temperature during the first 2 months after onset of the rainy season. Values are averages for flowering populations. The filled symbols indicate the percentage of flowering clones in the seasons in which soil fertilizer was applied (2006/07 and 2009/10). These values were not included in the regression. Means ± s.e.

Proliferation

Variation in pseudovivipary was analysed in clones from four populations with a high proportion of proliferating clones – populations 1 (19 %), 4 (83 %), 12 (43 %) and 13 (71 %) – representing a total of 23 proliferating clones. No relationship was found between the proportion of proliferating clones in these populations and rainfall at the site of population origin. Therefore, proliferating clones were pooled together in the following analysis and compared with non-proliferating clones.

Usually only one type of panicle – normal or proliferated – was produced by a flowering clone in a given growing season. In contrast, most clones were able to produce either normal or proliferated panicles, or remained vegetative in different years (Fig. 5). Furthermore, proliferating clones did not show a distinct time sequence of flowering and proliferation across years. Of the 23 proliferating clones, only one clone produced proliferated panicles every year, while the other clones alternated irregularly between years, producing normal panicles, proliferated panicles or not flowering at all.

Fig. 5.

Variation across years in the incidence of flowering and proliferation in proliferating clones from four populations inhabiting sites with different annual precipitation and grown in the net-house. The same clone is able to produce normal or proliferated panicles, or remain vegetative in different years.

Comparison of flowering patterns in proliferating and non-proliferating clones showed similar interannual variation in the proportion of flowering clones and similar responses to soil fertilization in the 2006/07 and 2009/10 seasons (results not shown). However, the frequency of flowering years was marginally higher in the proliferating clones (60·1 ± 5·0 % vs. 47·6 ± 2·4 %, P = 0·06).

Intraclonal variation in flowering and proliferation

Intraclonal variation in flowering and proliferation was studied under controlled conditions in the phytotron experiment. Families of daughter plants raised from bulbs taken from 12 proliferating clones belonging to three populations were exposed to marginal induction of flowering, consisting of 15 or 30 d under 12 h days and transferred back to non-inductive SDs, or subjected to an additional continued exposure to 12 h days until plant dormancy. Populations did not differ in their responses to the photoperiodic treatments, and the 12 clones were pooled together in the following analysis.

Individual clones produced mixed families of daughter plants: some of them flowered with normal or proliferated panicles, while others remained vegetative, even though they grew under the same conditions. The proportion of daughter plants that reached flowering in each clone family was not affected by the photoperiodic treatments and was similar to the non-inductive SD treatment (average 41%, Fig. 6A). However, under SDs, most of the flowering daughter plants produced proliferated panicles. Increasing the exposure to 12 h days gradually reduced the proportion of plants with proliferated panicles (P = 0·05) (Fig. 6A). The number of normal and proliferated panicles produced by flowering daughter plants increased after a transient exposure to 15 and 30 d of 12 h daylength compared with SDs (P = 0·05) (Fig. 6B). Contrary to expectations, this enhancement was cancelled by intercalation of 15 or 30 SDs before further exposure to 12 h days (P = 0·05) (Fig. 6B). Despite the contrasting responses to the 12 h treatments, the proportion of proliferating panicles in these treatments was similar, but lower than in the SD treatment (55–65 % vs. 92 %, P = 0·05) (Fig. 6C). This may indicate that the intercalated SDs reduced panicle formation (Fig. 6B), probably by rendering the plants insensitive to further exposure to marginal 12 h photoperiods that also induced earlier dormancy (Fig. 7A).

Fig. 6.

Effects of photoperiodic treatments on the reproductive development of clone families of daughter plants. (A) Percentage of non-flowering plants, or of those flowering with normal or proliferated panicles. (B) Number of normal or proliferated panicles per plant. (C) Percentage of proliferated panicles in flowering plants. Bars indicate the s.e.

Fig. 7.

Effects of photoperiodic treatment on (A) time to dormancy imposition and (B) time to flowering, in flowering, proliferating and non-flowering daughter plants. Bars indicate the s.e.

Time to flowering and time to dormancy were shortened by increasing exposure of the daughter plants to 12 h days (F = 18·10, P = 0·001 and F = 25·13, P = 0·001, respectively; Fig. 7). Daughter plants with proliferated panicles flowered earlier under SDs and after one exposure to 15 or 30 d of 12 h daylength than plants producing normal panicles. Again, this differential response disappeared after intercalation of non-inductive SDs (photoperiod × panicle type interaction: F = 6·921, P = 0·001; Fig. 7). In contrast to flowering, time to dormancy did not differ between normally flowering, proliferating and non-flowering daughter plants under each photoperiodic treatment (Fig. 7).

DISCUSSION

Flowering and pseudovivipary among populations and climatic conditions

Populations of P. bulbosa occurring along the rainfall gradient showed high interannual variation in the proportion of clones that flowered in the common garden experiment. Yet, a trend for a higher proportion of flowering clones with decreasing rainfall was maintained across 11 growing seasons, despite wide interseasonal differences in climatic conditions. This finding provides further support for our assumption (Ofir and Kigel, 2003) that keeping dispersal potential by production of sexual and/or asexual summer-dormant propagules (i.e. seeds and bulbils) is advantageous under arid conditions with low vegetation density, but with high death risk due to drought. In contrast, in mesic sites with higher plant density, allocation of resources to in situ persistency and spreading based on basal bulb production is probably of greater advantage under stronger competition from neighbour plants than propagule dispersal (Bullock et al., 1995; Eriksson, 2011).

Populations and clones of P. bulbosa were highly versatile in their reproductive behaviour across years. A given clone flowers in some years but not in others. Similarly, some clones can produce normal or proliferated panicles in different years, or became dormant without flowering. In P. bulbosa, both flowering and dormancy are promoted by longer days, but flowering is inhibited by high temperature and drought since they induce and advance dormancy onset (Ofir and Kigel, 2006, 2007). Thus, the non-flowering phenotype probably results from early induction of summer dormancy by the combined effect of longer days, higher temperatures and drought towards the end of the growing season (Ofir and Dorenfeld, 1992; Ofir and Kigel, 2006). It can be argued that the interannual variation in flowering and proliferation observed in the net-house was due to differences in the timing of the first effective rain event and, therefore, in the climatic and daylength conditions encountered by the plants during their early growth. Moreover, the observed interannual variation in flowering and proliferation in clones of P. bulbosa in the net-house can be attributed to differences in the induction of two counteracting processes – flowering and dormancy – in response to interannual variation in daylength and temperature during the early growing season. Late onset of the rainy season will lead to exposure of plants to shorter days and lower temperatures, compared with earlier onset of the season. In P. bulbosa, SDs and low temperature enhance flowering (Ofir and Kigel, 2006), as found in other temperate grasses (Heide, 1994). Furthermore, the fact that in temperate grasses the critical daylength for flowering is shorter under lower temperatures (Heide, 1994) may explain the early flowering of P. bulbosa under SDs in winter, well in advance of dormancy induction later in the season. Conversely, since dormancy imposition is enhanced by early drought spells occurring during March (Ofir and Dorenfeld, 1992), late onset of the rainy season may prevent flowering due to early plant dormancy.

Flexibility in the production of proliferated inflorescences has been reported in temperate grasses and usually occurs under conditions sub-optimal for flowering – shorter days and lower temperatures (Wycherley, 1952; Evans, 1964; Junttila, 1985; Heide, 1988; Ofir and Kigel, 2003). It is noteworthy that clones from the Montfort population (mesic site, 810 mm rain year⁻¹) never flowered in the net-house, but 70 % of the plants of this population reached flowering in the phytotron under 9 h daylength and 16/10 °C, conditions that postpone or prevent dormancy induction (Ofir and Kigel, 2006). Thus, sexual reproduction is not necessarily suppressed in non-flowering clones of P. bulbosa. Its expression, however, can be limited to very specific and less frequent environmental conditions. In these clones, the threshold conditions for flowering and pseudovivipary are probably rarely exceeded in the field, so seed or bulbil production occur only occasionally.

Pulses of soil fertilization strongly increased the proportion of flowering clones, panicle production and the proportion of proliferated panicles. Effects of mineral availability on pseudovivipary have been little studied. In P. alpina var. vivipara, higher P and N levels increased the number of proliferating tillers from 28 to 70 % (Pierce et al., 2003). These effects can be attributed to enhanced tillering or to a higher ratio of tillers reaching flowering before dormancy imposition. In P. bulbosa we could not differentiate between these complementary alternatives, since this grass produces large numbers of small tillers that are difficult to separate and count in the dense tufts growing in the pots. In a previous study we found low ratios of flowering tillers: 1–7 % in field conditions (Ofir and Kerem, 1982) and 2–24 % under controlled conditions (9 h daylength, 16/10 °C; Ofir and Kigel, 2006). These ratios are within the range of flowering tillers reported in demographic studies of tiller dynamics in other temperate perennial grasses (Jonsdottir, 1991; Herben et al., 1993; Janisova, 2006). It is plausible that under field conditions flowering and pseudovivipary are controlled by both soil resources and climatic factors inducing flowering and dormancy. In grasses, the developmental fate of tillers is determined by tiller age and by tiller interactions that regulate tiller density, according to soil resource availability (Derner and Briske, 1999; Janisova, 2006). In this context, non-flowering tillers represent the stored regenerative potential for the following growing season. In P. bulbosa, poor soil conditions probably lead to a smaller proportion of tillers reaching flowering and, therefore, to a larger reserve of dormant basal bulbs for the next season.

Intraclonal variability

High intraclonal tiller heterogeneity occurs in P. bulbosa, as shown by the divergent developmental fates of daughter plants arising from bulbs taken from the same parent clone and grown under the same conditions, namely normal flowering, proliferating or non-flowering daughter plants. This suggests that the physiological state of the planted bulbs, as determined by age and growth conditions in the previous season, modulates flowering response in the following growth cycle. The fact that about 41 % of the daughter plants remained vegetative in all photoperiodic treatments – non-inductive SDs, as well as after short and prolonged exposure to 12 h days – suggests a juvenile stage preventing flowering in young P. bulbosa tillers. Demographic studies in several perennial grasses have shown that flowering by tillers younger than 1–2 years is rare (Jonsdottir, 1991; Herben et al., 1993; Janisova, 2006). This age requirement for flowering may involve a critical tiller size for flowering (Herben et al., 1993; Laterra et al., 1997; Janisova, 2006), which in P. bulbosa could be related to a critical bulb size, as in other geophytes (Le Nard and De Hertogh, 1993).

Ecological significance

Poa bulbosa thrives in semi-arid Mediterranean grasslands and rangelands under heavy grazing pressure (Noy-Meir, 1990; Peco et al., 2005). Its small size and prostrate habit, clonal reproduction by basal bulbs and the phenological traits of early flowering and summer dormancy onset are convergent strategies allowing escape from grazing and from drought. High versatility in the reproductive traits of P. bulbosa perhaps adds further flexibility in its adaptation to arid conditions and grazing. A life history strategy combining sexual and asexual reproduction is probably stable in highly variable environments, with differential success of the various forms of reproduction under different conditions (Bengtsson and Ceplitis, 2000). Clonal species are considered to be more tolerant to extreme abiotic environments, and pseudovivipary is common in arctic and alpine species (Lee and Harmer, 1980). Plantlets and bulbils possess more reserves than seeds, thus improving chances for establishment compared with seedlings. On the other hand, occasional flowering allows genetic recombinations, thus circumventing constraints to genetic diversity due to clonal reproduction (Vallejo-Marin et al., 2010). Indeed, genetic variation in populations of clonal species is not necessarily poor (Silvertown, 2008). Genetic variation in the studied populations of P. bulbosa certainly occurs, as shown by the differences in flowering and proliferation among coexisting clones grown in the common garden experiment. Clones probably differ genetically in inductive thresholds for flowering, pseudovivipary and dormancy in response to the relevant environmental factors. Further work is needed in P. bulbosa to assess the fitness costs involved in almost exclusive clonal reproduction and of the importance of genetic diversity for the maintenance of viable populations with increasing aridity.