Flowering phenology and height growth pattern are associated with maximum plant height, relative growth rate and stem tissue mass density in herbaceous grassland species


  • Shucun Sun,

    Corresponding author
    1. Department of Biology, Nanjing University, 22 Hankou Road, Nanjing 210093, China
    2. EcoRes Laboratory, Chengdu Institute of Biology, Chinese Academy of Sciences, Chengdu 610041, China
      Correspondence author. E-mail: shcs@nju.edu.cn
    Search for more papers by this author
  • Lee E. Frelich

    1. Department of Forest Resources, University of Minnesota, 1530 Cleveland Ave. N., St. Paul, MN 55108, USA
    Search for more papers by this author

Correspondence author. E-mail: shcs@nju.edu.cn


1. We hypothesize that flowering phenology correlates with plant height growth pattern and that the pattern is associated with functional traits including maximum plant height (Hmax), RGR, stem tissue mass density (SD), hollow ratio (proportion of central hollow of stem cross-sectional area) and leaf mass per area (LMA) in grassland herbaceous species.

2. We investigated plant height growth trajectories and flower phenology, and measured LMA, SD and hollow ratio for 25 herbaceous species including 20 dicot forb species and five monocot species in an old-field grassland of New England, USA. Hmax, RGR, T10 and T90 (Julian day when plant height was 10% and 90%Hmax respectively) were derived from a logistic function for each species and were analysed in relation to LMA and SD.

3.Hmax was positively correlated with T10, T90 and flowering onset time (Julian day when the first 10% of flowers were blossoming) across species and across evolutionary-correlated divergences. Early growing and flowering species were shorter than late ones, and species reaching Hmax earlier flowered earlier than their counterparts.

4. There was a positive relationship between T90 and RGR, in which early growing species were usually at mid-to-high levels of RGR, while late-growing ones had widely varied RGR. A similar relationship was found between flowering onset time and RGR. RGR was significantly negatively correlated with SD and LMA but positively with hollow ratio, as indicated by correlation analysis and phylogenetically independent comparative analysis.

5. Based on the above results, we propose that herbaceous species have two major dimensions of height growth strategies (early vs. late and fast vs. slow growth), collectively resulting in three extreme cases (early and fast, late and slow, and late and fast). Different height growth trajectories resulting from these strategies may reduce asymmetric competition among co-existing species in dense grasslands.

6.Synthesis. Flowering phenology and height growth patterns are significantly associated with functional traits such as RGR, LMA and hollow ratio in herbaceous grassland species. The difference in height growth trajectories and associated functional traits may allow species coexistence possibly at both plant and consumer trophic levels.


Plant phenology – the timing of life-history events – often varies remarkably among species within communities. Phenological variation has been suggested to be significant to plant growth and reproduction (Rathcke & Lacey 1985; Fenner 1998) and species coexistence in plant and pollinator communities (van Schaik, Terborgh & Wright 1993; Brody 1997). However, understanding the mechanisms underlying variation in plant phenology among species within physically and biotically similar habitats remains incomplete (Elzinga et al. 2007).

Flowering is a crucial event in plant life history; flowering phenology affects seed size, seed dispersal and abundance of pollinators (Bolmgren & Cowan 2008; Du & Qi 2010; Forrest, Inouye & Thomson 2010). Flowering phenology is known to be determined by many abiotic factors such as temperature, photoperiod and resource availability (e.g. nutrients and water; Rathcke & Lacey 1985) and biotic factors including pollinators, seed dispersers and floral pathogens (e.g. Elzinga et al. 2007). Current research reveals that plant functional traits may serve as determinants for flowering phenology. In particular, flowering onset time has been reported to be positively correlated with maximum plant height (Hmax). For example, among herbaceous grassland species, taller species often flower later than shorter ones (Dahlgren, von Zeipel & Ehrlén 2007; Du & Qi 2010). However, it is still ambiguous whether flowering phenology largely depends on height growth pattern in grassland species, because two critical questions have not been fully addressed. First, the relationship between Hmax and the time when plants reach Hmax is not clear. Although individuals and the whole community gain height as the growing season progresses, taller species may possibly reach Hmax earlier or later than short ones. Secondly, the relationship between the time when plants reach Hmax and flowering time is not clear; plants may flower before or after they reach Hmax.

Plant height is a leading dimension of ecological variation, is critical to species’ competitive ability and survival, and is related to many important functional traits (Westoby 1998; Westoby et al. 2002; Poorter et al. 2003). Although different height growth strategies or trajectories have been reported for woody species (Falster & Westoby 2003, 2005a,b), little is known about height growth patterns and associated mechanisms in herbaceous species.

In dense temperate grasslands, most species are hemicryptophytes and therophytes under Raunkiær’s life-form classification system; species with these life forms die off above-ground during the winter and emerge from the ground in spring and accomplish vegetative growth, flowering, fruiting and seeding in a short growing season. Moreover, grassland herbaceous species are mostly light-demanding, as opposed to species from dense forests that may contain a large proportion of shade-tolerant species (Poorter et al. 2003; Poorter, Bongers & Bongers 2006). Therefore, competition for light (intra and interspecific) is intensive, so that grassland species require high carbon gain efficiency. Natural selection resulting from intensive light competition should favour shorter species that reach Hmax earlier than taller species in grasslands, so that these species have enough time in open canopy conditions to produce sufficient photosynthate for reproduction. There are potentially three strategies for a species to reach Hmax early to: have a small Hmax, have early seedling emergence, and grow rapidly in height. Because the timing of seedling emergence and RGR are limited by low temperature in early spring, these three strategies collectively lead to the prediction that species arriving at Hmax earlier should be shorter, emerge earlier and grow faster than their counterparts.

Variation in RGR of plant height may be explored in terms of LMA (leaf mass per area) and SD (stem tissue mass density), as suggested by studies on dry mass production (King et al. 2005, 2006a,b; Poorter et al. 2003; Poorter, Bongers & Bongers 2006; Poorter et al. 2008). As a component of RGR, LMA has been widely recognized to have a negative effect on RGR (e.g. Hunt & Cornelissen 1997; Poorter et al. 2003); high LMA is often associated with low nutrient concentration and low photosynthetic capacity of leaves, while low LMA enables plants to increase leaf area and maximize light interception (e.g. Hunt & Cornelissen 1997; Sun, Jin & Shi 2006). The stem density represents plant dry mass per unit fresh stem volume. It has recently been recognized as an important functional trait related to the trade-off between growth and defence in the stem economics spectrum (Chave et al. 2009). Low SD allows for high RGR (King et al. 2005; Poorter, Bongers & Bongers 2006; Poorter et al. 2008), while high SD contributes to resistance to breakage and pathogen attack (Chave et al. 2009). In addition, hollow stems may be an important trait influencing height growth rate. For any given biomass plants would grow taller with increasing hollow ratio (area ratio of hollow portion of the stem to stem cross-section) and hence decreasing SD, as shown by studies examining stem growth rate for crops such as broccoli, cauliflower and cabbage (Shelp & Shattuck 1987). Accordingly, provided that flowering phenology closely correlates with the pattern of plant height growth that primarily rests with Hmax and RGR, which is in turn associated with LMA, SD and hollow ratio, we may speculate that flowering phenology and height growth pattern should be associated with these functional traits.

In this study, we investigated flowering phenology and height growth trajectory, SD, LMA and stem hollow ratios for 25 herbaceous species (5 monocot and 20 dicot species) in a New England, USA, old-field grassland. We derived Hmax, timing of seedling emergence and the time when adults reached Hmax, and RGR from observed plant height versus time relationships for each species, and we calculated flowering onset and offset time based on flower demography. Our primary objective was to determine whether flowering phenology correlates with height growth pattern and whether the phenologies are associated with plant functional traits. Specifically, we predict that: (i) flowering onset time, Hmax and the timing of reaching Hmax are positively correlated, and (ii) flowering onset time and the timing of species reaching Hmax are negatively associated with RGR and SD, but positively with LMA and hollow ratio. Additionally, because across-species patterns might arise due to phylogenetic relationships (Martins 2004), we conducted phylogenetically independent comparative analyses (PICA) to test whether the relationships between the traits were evolutionarily constrained.

Materials and methods

Study site and species selection

The study was conducted in an old-field grassland at the Yale-Myers Research Forest, in north-eastern Connecticut, USA. The research site is a 3240-ha north-eastern hardwood ecosystem interspersed with old fields. Our research grassland was abandoned from subsistence farming in the late 1960s, and is currently covered with a variety of grass and forb species, with the most dominant being Phleum pratense, Solidago rugosa, Poa pratensis, Aster novae-angliae, Trifolium repens and Daucus carota. Total stem density is more than 200 m−2 during the growing season peak.

To facilitate the survey of flowering phenology and plant height growth, we randomly placed twenty 1-m2 circular cage enclosures (1.128 m in diameter and 1 m high) throughout the grassland. All the cages were sunk into soil about 10 cm to make them stable against the wind. The cages were made of aluminium screen with a mesh size of 5 mm. Such enclosures have been proved to have no significant adverse effect on microclimate for plant growth (Schmitz 2004).

The cages contained 37 species including three small shrub species, 26 forb and eight grass species. However, not all species could be found in any single cage. Our survey of flower phenology and height growth extended from mid-May to mid-September 2009, surveying each cage once every 5–15 days. We do not have complete records of flower phenology for four forb species or height growth for two species, because their flowering peaked before the survey period. Ultimately, we analysed 25 species, including 20 forb and five grass species (see Table 1). These species accounted for more than 85% of total above-ground biomass of the cages (data not shown).

Table 1.   Correlation matrix showing the correlation coefficients among studied variables of 25 herbaceous species in an old-field grassland. The lower-left and upper-right half matrices are for the relationships of across-species and across-evolutionary divergences respectively
 Hmax (cm)RGR (day−1)LMA (mg cm−2)SD (mg cm−3)Hollow ratioT10 (Julian day)T90 (Julian day)Flowering onset time (Julian day)Flowering offset time (Julian day)
  1. Hmax, maximum plant height; RGR, relative growth rate (β); α, the species-specific constant for the logistic function (H = Hmax/ (1 + Exp (α − β*T))); T10 and T90, the Julian day when plant height was 10% and 90%Hmax respectively; LMA, leaf mass per area; SD, stem tissue mass density; n.s., not significant.

  2. *< 0.05; **< 0.01; ***< 0.001.

Hmax (cm) 0.21ns0.38ns0.13ns−0.14ns0.69***0.43*0.41*0.41*
RGR (day−1)0.05ns −0.51*−0.73***0.49 *0.36ns−0.61**−0.62**−0.53**
LMA (mg cm−2)0.42*−0.5* 0.44*0.31ns−0.04ns0.35ns0.33ns0.25ns
SD (mg cm−3)0.22ns−0.69***0.44* −0.39ns−0.24ns0.64***0.67***0.51*
Hollow ratio−0.13ns0.46*−0.21ns−0.49* −0.27ns−0.6**−0.59**−0.62**
T10 (Julian day)0.69***0.37ns−0.06ns−0.02ns−0.23ns 0.45*0.4*0.38ns
T90 (Julian day)0.51**−0.6**0.25ns0.56**−0.61**0.33ns 0.94***0.9***
Flowering onset time (Julian day)0.48*−0.58**0.23ns0.6**−0.61**0.25ns0.95*** 0.81***
Flowering offset time (Julian day)0.38ns−0.58**0.38ns0.64***−0.67***0.42*0.9***0.84*** 

Height growth trajectories

We recorded plant height for the tallest individual of each cage for each species. Using the data on the maximum plant height at a specific time, we characterized intraspecific height growth trajectories by fitting a logistic function of the form:

image(eqn 1)

where observed height (H) was a function of survey time (T, Julian day of 2009), Hmax, relative growth rate (RGR; β) and a species-specific constant (α). The logistic function and its modified versions have been shown to be successful for fitting plant growth–time curves (Kaufmann 1981; Hara, Kimura & Kikuzawa 1991).

Values of Hmax, RGR and α were derived with the nonlinear least-square method (XLSTAT Win-2010) for each species, as shown by the examples of Rudbeckia hirta, S. rugosa and D. carota (Fig. 1; see also Table 1). All study species had highly significant fits to the logistic equation (r2 ranging between 0.76 and 0.93; all < 0.001). In addition, we derived T10 and T90 (the Julian day when plant height was 10% and 90% of Hmax respectively) for each species after inputting the estimated values of Hmax, α and β into eqn (1). T10 was used to characterize the timing of seedling emergence, and T90 as a proxy of timing when plants reached Hmax.

Figure 1.

 Example height growth trajectories for three typical herbaceous species in the study grassland including (a) Rudbeckia hirta, (b) Solidago rugosa and (c) Daucus carota. Curves indicate fitted logistic relationships (eqn 1) and parameters are given in Appendix S1 for all species.

Flowering phenology

With the same frequency as for the plant height survey, we recorded the number of flowering stems for clonal species and flowering individuals for non-clonal species within each cage. Flowering was defined as the presence of open flowers for forbs or exposed stamens and/or styles for grasses; flowering stems/individuals were defined as the stems/individuals bearing blossoming flowers. As flower (or inflorescence, the same hereafter) number was very large (even in individuals) for some species and it was very hard to tag each flower for those species, we used population density of flowering individuals (stems) to estimate flowering phenology. We employed a quadratic function to describe the number of flowering individuals (stems)–time (Julian day) relationship (eqn 2) for each cage and for each species.

image(eqn 2)

where number of observed flowering individuals (stems) (Y) was a function of time (X, in Julian day of 2009) and three species-specific parameters (A, B and C). Because A was found to be negative, we calculated maximum flower density for each species, and then we calculated the dates (two solutions) when 10% of the maximum flower density were blossoming. We defined the first and the latter dates as the flowering onset time and offset time respectively. Nevertheless, there were too few flowers within some cages to fit the relationship with the quadratic function, and not all species had the same sample size (number of cages). We calculated the mean flowering onset and offset time for each species (see Table 1). Because there were too few individuals and flowers in S. nemoralis and Lactuca canadensis, we pooled the data from all the cages and then fitted the quadratic function, and ultimately calculated the flowering onset and offset times.

Some previous studies have defined the times when the first individual started and the last individual ended flowering as flowering onset and offset time respectively (e.g. Sherry et al. 2007; Hovenden et al. 2008). In this study, because there was large variation in flowering onset/offset time among flowers and among individuals, and because flowering phenology sometimes appeared sporadic in some species (e.g. possibly due to herbivore damage or physical stress), we used percentages of blossoming flowers to estimate the phenology, following Dunne, Harte & Taylor (2003).

Stem tissue mass density and leaf mass per area

Stem tissue mass density (dry mass per fresh volume; mg mm−3) was measured for at least four flowering individuals of each species. One 1-cm-long segment was cut from stems at 1 cm above the ground surface for each individual. The long and short axes were measured for each segment assuming that stem cross section was elliptical in shape. For the species with hollow stems, we also measured the long and short axes for the hollow part of each segment. The cross-sectional area or hollow area was calculated as πab/4, where a and b were the long and short axes, respectively; and hollow ratio was calculated as the area ratio of the hollow part to the whole cross-sectional area. Next, every stem segment was dried at 65 °C for 48 h and then weighed. We calculated stem tissue density as the dry mass divided by fresh volume that was calculated based on whole sectional area whether the stem was hollow or not.

Leaf mass per area was measured as the leaf dry mass divided by fresh leaf area. We collected mature and fully developed leaves from the least-shaded, middle part of the plant canopy, because the basal leaves were often shaded and largely different from upper leaves in both shape and size. This sampling strategy permits valid interspecific comparison, although it was slightly different from the standard method that required selecting leaves in full sunlight (Cornelissen et al. 2003). The collected leaves were scanned and digitized images were used to calculate the projection area of the fresh leaves. These leaves were further oven-dried for 48 h at 65 °C and then weighed to 0.1 mg.

Data analysis

Data on the study traits were averaged for each species (Table 1). Relationships among the traits studied were determined with correlation analysis and regression curves and the above-mentioned parameters were fitted using the method of least squares. These analyses were carried out using STATISTICA software (Statsoft Inc. 2001).

Phylogenetic Comparative Methods of COMPARE (Martins 2004; http://compare.bio.indiana.edu/) were employed to conduct PICA. The calculation method of phylogenetically independent contrasts followed Martins (2004). The phylogenetic tree was constructed following Phylomatic, version 4.01 (http://www.Phylodiversity.Net/phylocom/; Webb, Ackerly & Kembel 2008). The branch length on the tree does not indicate the time of evolutionary divergence between species or taxa, and therefore it was not used in calculations. Regression of evolutionary divergence was conducted using standard model I techniques. In this way, we could determine whether the correlation between different functional traits varied with evolutionary divergence.

In order to detect whether there is a particular structural organization among the study variables, we also conducted a path analysis using structural equation modelling, as a complement to the regression analyses. We set Hmax and hollow ratio as the driving variables, T90 as a dependent variable, and SD and GR as causal variables (see Shipley 2004). Flowering phenology was not necessarily included as a dependent variable because it was highly dependent on T90 (see Appendix S1 in Supporting Information). We aimed to test to what extent hollow ratio, SD and RGR directly and indirectly affected T90. The path analysis was performed with structural equation modelling using the SEPATH module of STATISTICA (Shipley 2000). The parameters were estimated using generalized least squares followed by maximum likelihood. Comparative fit index (CFI) and root mean square error of approximation (RMSEA) were used to assess closeness of fit. Good models usually have a RMSEA<0.05 and a CFI > 0.95 (e.g. Vile, Shipley & Garnier 2006).


Relationship between flowering phenology and height growth pattern

Plant height increased progressively during the growing season for all the species, but their Hmax varied several fold, ranging from 25 to 150 cm among species (Appendix S1; see also Fig. 1). Among the species, Hmax was positively correlated with both T10 (Table 1) and T90 (Fig. 2a), i.e. seedlings of taller species emerged and reached Hmax later than those of shorter species. Moreover, T10 and T90 were positively correlated (Table 1), indicating that the species of seedlings emerging earlier tended to reach Hmax earlier.

Figure 2.

 Bivariate relationships (a) between maximum plant height (Hmax) and T90 and (b) between Hmax and flowering onset time. *< 0.05; **< 0.01.

There were also large variations in flowering onset and offset time among species. The earliest species was Lychnis flos-cuculi; its flowering onset was about 3 months earlier than that of S. rugosa and S. graminifolia, which flowered during late August (Appendix S1). Flowering onset time was positively correlated with the offset time (Table 1), suggesting that early flowering species tended to shed flowers earlier.

Flowering onset time, but not offset time, was positively correlated with Hmax (Fig. 2b), i.e. the taller species flowered later than the short species. More importantly, T90 and flowering onset time were highly and positively correlated (Table 1), suggesting that different species tended to reach Hmax and flower in a temporally consistent way.

Relationship between plant phenology and functional traits

Both T90 and flowering onset time were negatively correlated with RGR (Fig. 3a,b). The early growing and early flowering species tended to have a relatively greater RGR, but the species with large T90 and flowering onset time had variable RGR. In contrast, positive relationships were found between T90 and SD and between flowering onset time and SD (Table 1). Furthermore, hollow ratio was negatively correlated with T90, flowering onset time and offset time (Table 1), and the late species were more likely to be characterized by solid stems. However, LMA was not significantly associated with flowering phenology and height growth pattern (Table 1).

Figure 3.

 Bivariate relationships of relative growth rate versus (a) T90 and (b) flowering onset time in 25 herbaceous grassland species. **< 0.01.

Hmax was positively correlated with LMA and the taller species tended to be characterized by a high LMA (Table 1). The stem density was negatively correlated with RGR and hollow ratio, but positively correlated with LMA (Table 1). The species with low-SD stems tended to grow faster and have a higher LMA and hollow ratio than the high-SD species.

In addition, the above across-species correlations were generally consistent with the results of PICA (Table 1). However, some marginally significant cross-species correlations were not significant in PICA (e.g. the relationship between SD and hollow ratio), possibly due to the small sample size in this study.

Direct and indirect effects of functional traits on height growth pattern

Path analysis provided a good fit to the hypothesized structural equation model (λ = 4.642, d.f. = 2, = 0.098, CFI = 0.991, RMSEA = 0.102). The results were generally consistent with the regression analyses. The variables including hollow ratio, SD, RGR and Hmax accounted for 72.2% of the variation of T90, which explained more than 90% of the variation in flowering onset time (= 25, R2 = 0.911, < 0.001). Hollow ratio directly affected SD that directly influenced RGR, which directly impacted T90; its direct effects on T90 and RGR were also significant (Fig. 4). Hmax had both significant direct and indirect effects on T90 via direct effects on growth rate and SD (Fig. 4). However, the direct effect of SD on T90 was insignificant since the path from SD to T90 was not significantly different from zero (> 0.05; Fig 4).

Figure 4.

 Diagrammatic rendition of the structural equation model for the effect of maximum plant height (Hmax), hollow ratio, stem density and growth rate on T90 among the 25 grassland species. Path coefficients between variables are unstandardized partial regression coefficients. The path coefficients between variables are shown if they are significantly different from zero. Arrow widths are proportional to the standardized path coefficient. Variance explained by the model (r2) is given under each variable name. The abbreviations same as in Table 1.


We have shown that flowering onset time is positively correlated with maximum plant height and time when the species reaches Hmax, as indicated by the positive correlations among Hmax, T90 and flowering onset time. This clearly suggests that plant height growth pattern could be a major determinant for flowering phenology among the herbaceous species studied. We also showed that the height growth pattern (as characterized by T10 and T90) was significantly correlated with RGR of plant height, which was associated with LMA, SD and hollow ratio. In addition, the correlations among the study parameters were generally consistent between the results of across-species analyses and PICA. This suggests that the relationships are not simply a consequence of common ancestry, but have resulted from ecological forces. Consequently, it could be concluded that flowering phenology and height growth pattern were associated with the functional traits including RGR, SD, LMA and hollow ratio in the study species, which might be of significance for understanding species coexistence in grasslands.

Relationship between phenologies of height growth and flowering

The positive correlation between Hmax and flowering onset time is consistent with several previous studies of herbaceous grassland species. For example, Vile, Shipley & Garnier (2006) showed that the flowering onset time of 34 herbaceous species primarily depended on maximum plant height in a study of Mediterranean old-field succession. This was further confirmed by Bolmgren & Cowan (2008) who showed that taller species flowered later than shorter ones for north-temperate perennial herbs, and Du & Qi (2010) who revealed that maximum plant height was positively correlated with flowering onset time for herbaceous species in a large survey of 11 plant communities representing a Qinghai-Tibetan flora. The current study goes beyond this positive correlation between flowering onset time and Hmax to link flowering phenology with plant height growth pattern, because we also showed positive correlations among Hmax, T90 and flowering onset time, which excluded the possibility of inconsistency or mismatch between flowering phenology and plant height growth pattern for reasons explained in the following two paragraphs.

First, the positive correlations among Hmax, T10 and T90 excluded the possibility that the shorter species tended to flower earlier but reach Hmax later than the taller species, given the existing knowledge that taller species often flower later than shorter species (Vile, Shipley & Garnier 2006; Bolmgren & Cowan 2008). Such relationships may differentiate the timing for exploitation of resources among species of dense grasslands, where plant–plant competition could be extremely high. The herbaceous species studied – despite growing in the same habitats and sharing similar life forms – had widely varied Hmax. For example, the tallest species (S. nemoralis) has individuals close to 1.5 m tall, while the shortest one (Cerastium vulgatum) has Hmax of not more than 30 cm. These extreme species both flowered and fruited, accomplishing all their critical life-history events during the growing season. Thus, differing T10 (≈ the time when seedlings emerge) and T90 (≈ the time when species reach Hmax) among species is one of the most effective ways to alleviate competitive intensity, particularly for the shorter species. The relatively early growing short species in this study are similar to spring ephemerals that finish above-ground portions of their life cycles before the forest canopy closes and similar to understorey woody species that leaf out earlier than canopy trees (Sun, Jin & Shi 2006). Such phenological avoidance of shade stress mediated by taller species can be an important strategy for carbon gain and species persistence (Seiwa 1998; Walters & Reich 1999). In contrast, the late-growing species with a greater T90 and Hmax might have an advantage in light interception and growth temperature during the middle to late growing season. Although light intensity during the late growing season (e.g. August to mid-September) is about the same as the early growing season (April–May) in the northern hemisphere, mean (daily or monthly) temperatures are usually higher during the late growing season on the study site (http://usclimatedata.com/climate.php?location=USCT0044) and could be more effective for leaf photosynthesis.

Secondly, the highly positive correlation between T90 and flowering onset time excluded the possibility that early flowering species reached Hmax later than late ones and vice versa, thus further excluding a causal relationship between the two variables. We observed that the species mostly flowered several days before (but seldom after) they reached Hmax (Appendix S1), although the intervals between T90 and flowering onset time were different among species. Furthermore, flowering height, which was calculated from eqn 1 using flowering onset time and the three variables (Hmax, α and β), was highly significantly correlated with Hmax (R2 = 0.942, < 0.001). This suggested that the size of flowering plants was critical to the timing of flowering and hence reproduction. A potential physiological mechanism underlying the vegetative growth–flowering phenology relationship is that plants cannot reproduce much earlier than the time at which they have accumulated sufficient material resources (reflected by plant height) and they also cannot delay reproduction much beyond Hmax, in which case they may run out of time to produce seeds before the end of the growing season. Therefore, early flowering plants have a relatively small resource availability but a long time to develop seeds, while the taller, late-flowering species have a short time for seed development but might be able to support a high reproductive capacity. Thus, seed size (as determined by developmental time) has been related to the variation in flowering onset time among perennial herb species (Bolmgren & Cowan 2008; Du & Qi 2010). We did not record seed size for the study species, although we did notice that the taller, late-flowering L. canadensis bore smaller seeds than the shorter, early flowering C. vulgatum. A potential evolutionary mechanism underlying the vegetative growth–flowering phenology relationship is that the shorter species may gain a short-term overtopping advantage while they are flowering, as suggested by Vile, Shipley & Garnier (2006). However, not all the shorter species always flowered and reached Hmax earlier than the taller species, partly explaining why the determinant coefficients of the relationships between Hmax and T90 and between Hmax and flowering onset time were at a relatively low level (R2 < 0.5 for both; Fig. 2). For example, most individuals of Prunella vulgaris were less than 40 cm tall, and the Hmax was 47 cm, but flowering for this species peaked in mid-August, later than many taller species such as Solidago spp. and Chrysanthemum spp. (Table 1).

Relationship between height growth phenology and functional traits

Although temperature was usually low in early spring, RGR of the shorter, early growing species (with small T10 and T90) were mostly at middle or high levels relative to the taller, late-growing species (Fig. 3a). This is consistent with the prediction that natural selection should favour early growing species with shorter stature and fast growth. However, the taller species had large variation in RGR and associated growth trajectories, forming a seemingly triangular relationship between Hmax and RGR. For example, goldenrod and wild carrot (Fig. 1b,c) reached Hmax and flowered at similar times, but they had contrasting growth trajectories. Goldenrod mostly emerged very early in the growing season; in early May, their height was over 40 cm, taller than wild carrot that became one of the tallest species late in the growing season. However, goldenrod plants had small RGR and gradually reached their Hmax and flowered. In contrast, wild carrot remained short for a long time until late July when it grew very quickly, reached a Hmax higher than, and flowered later than, goldenrod (Appendix S1; Fig. 1c). Similar patterns were found for the monocot species. For example, P. pratense grew slowly with a high density until late July when it flowered and then reached its Hmax, whereas Agrostis perennans remained shorter than the former until the middle of July when it grew very quickly so as to exceed the height of P. pratense and then flowered. A similar pattern has also been observed by Grime & Hunt (1975), who indicated that slow-growing species mostly attained large stature at maturity, surprisingly consistent with our results. It seems that such a triangular relationship between Hmax and RGR could be widely applicable to grassland species.

It is apparent that for any given biomass, plants with low SD would grow taller than those with high SD, and less apparent that the low-density species are often more productive, as suggested by the positive relationship between LMA and SD (Table 1). The shorter species were more likely to have a lower SD than the taller ones; although the taller species had large variation in SD, corresponding to the triangular relationship between RGR and Hmax. Similarly, the species with a low RGR and a large T90 (slow-growing species reaching Hmax late; e.g. S. rugosa and A. novae-angliae) were mostly high in SD but small in T10, consistent with the negative relationships between SD and RGR and between SD and T90, while the species with high RGR and large T90 (e.g. wild carrot) were characterized by low SD and relatively larger T10. These observations are in accordance with the results of Castro-Díez et al. (1998), who analysed laboratory-grown seedlings of 80 European woody and semi-woody species and found that fast-growers were characterized by cheaply constructed stems (low SD), and King et al. (2005) who found that tree diameter growth rate was negatively correlated with wood density among 21 species in a south Asia tropical forest. There is no doubt that SD at least partly accounts for the variation in RGR and hence T90, as suggested by the results of the correlation analysis and the path analysis (Appendix S1, Fig. 4).

Leaf mass per area generally increased progressively among the species in the plant community, i.e. the shorter species generally were characterized by a low LMA but the taller species varied largely in LMA, and the fast-growing species were smaller in LMA than the slow-growing ones. As suggested in previous studies (e.g. Hunt & Cornelissen 1997; Sun, Jin & Shi 2006; Poorter et al. 2009), low LMA could allow for a large photosynthetic area in the rapid-growing species, especially in the shorter species; high LMA might be helpful for the species reaching Hmax later to alleviate possible UV-B damage and extend leaf longevity in high light conditions (e.g. Poorter et al. 2009).

Corresponding to the negative effect of SD on RGR, hollow ratio was negatively correlated with SD but positively with RGR across the study species. The fastest growing species often have hollows stems, e.g. wild carrot (Appendix S1). This is consistent with the well-known phenomenon that rapid height growth of woody bamboos is primarily because of the trait of hollow stems. Different from rapidly growing bamboos, hollow stems of herbs are often short-lived and therefore they may invest less in mechanical support tissues; hollow stems in herbaceous species may contribute to height growth to a greater extent than in woody species. Indeed, hollow stems are more often observed in herbs than in woody species (Niklas 1995). In our study, all the grass species had hollow stems (Appendix S1), but more than 80% of all species had solid stems. More importantly, hollow ratio not only directly affected SD, but also indirectly T90 and hence flowering phenology, as indicated by the results of the path analysis. Several studies have also clarified the importance of hollow ratio to plant mechanical stability in wild woody species (e.g. Niklas 1995; Brouat & McKey 2001), gas exchange for wild herbs living in wetlands such as reeds and cordgrass (Spatz et al. 1997), and mechanical support and vulnerability to fungi and animal attack (Niklas 1992, 1995), but the ecological importance of hollow stems in height growth has seldom been explored in wild herbs.

It is worthwhile to note that the early growing species were rarely high in both RGR and SD, although it is theoretically possible for those species having a large storage organ. For example, the wild carrot had a relatively large tuber with considerable underground nutrient storage, but it was large in both T90 and flowering onset time and characterized by a low SD and a high hollow ratio (Appendix S1). The answer to the question why such a species did not reach Hmax early in the growing season, with rapid growth and high SD, perhaps lies in the trade-off between quick growth in favourable environments and slow growth in poor habitats (e.g. Grime & Hunt 1975). In the study site, the herbaceous species often suffer adverse disturbances such as frost and low-temperature damage in early spring but the climate is mild and stable in the middle and late growing season with abundant rainfall. Thus, fast growth with dense stems in early spring would mean a high risk of investing resources with possibly low payback, which should be selected against.

In summary, because flowering phenology and plant height growth pattern closely correlated with each other and because the height growth pattern was significantly associated with RGR, which in turn correlated with LMA, SD and hollow ratio, these functional traits might serve as determinants for the difference in the plant phenologies among species within communities.

Implication for species coexistence in grasslands

Many mechanisms have been proposed to explain species coexistence in grasslands, including the trade-offs between survival and reproduction, shade tolerance and survival rate, seed size and seed output (e.g. Zobel 1992; Westoby et al. 2002; Poorter et al. 2008). Recent studies have recognized the importance of the difference in height growth trajectories to species coexistence in woody species of Australian forests (Falster & Westoby 2003, 2005a,b). Taller plants often have an advantage over shorter ones in light interception and in plant growth and reproductive success; if the advantage becomes overwhelming, asymmetric competitive exclusion will occur. Accordingly, different life-history strategies of height growth are important for species coexistence in plant communities such as the grasslands in this study, where competition for light and other resources is intense because herbaceous grassland species mostly share similar life forms, and plant density is usually high. In this study, we further propose that differences in height growth trajectory, flowering phenology and associated functional traits (LMA, stem density and hollow ratio) may collectively contribute to species coexistence in grasslands.

Because height growth trajectories are presumably determined by Hmax, timing of height growth (associated with α) and RGR (β; see eqn 1), there are two potential mechanisms for species to avoid asymmetric competition resulting from the same height growth patterns and hence to facilitate species coexistence. One is to differ in the timings of seedling emergence, reaching Hmax and flowering among species, and the other is to differ in RGR and hence in the growth trajectory among species, but to reach height maxima and flower simultaneously. The slow-growing, high-SD species tended to have the advantage over the rapid-growing ones during the middle of the growing season (curve B in Fig. 5), while the rapid-growing, low-SD species had the advantage over slow-growing species during the late growing season, because they were usually taller than their counterparts. This is possible because species differ widely in height growth rate depending on several functional traits including LMA, SD and hollow ratio. These two mechanisms (i.e. early vs. late and slow vs. fast growth) may have four potential combinations. However, there are usually only three species groups, as indicated by a, b and c in Fig. 1, because natural selection should not favour a species with a high stem density to grow rapidly and reach Hmax early in the growing season as discussed above. Putting these mechanisms together, it can be predicted that taller species tend to reach their Hmax and flower later than shorter species and that species with high SD and high LMA are more likely to reach their Hmax later in the growing season than their counterparts. Our results were generally consistent with these predictions.

Figure 5.

 Schematic diagram showing three extreme cases for herbaceous species in dense grassland. Early, fast-growing species (solid line), late slow-growing (dashed line) and late fast-growing species (shade line). These cases resulted from two dimensions of alternative strategies of height growth, early versus late and fast versus slow-growing (note that early and fast-growing species do not exist).

Pollinators are also important for maintaining plant species diversity in grasslands. The relationships among flowering phenology and Hmax and T90 create a high level of flower apparency for pollinators throughout the growing season because taller flowers are more likely to be visited by insects (Galen & Stanton 1989; Galen & Cuba 2001). Thus, the progression in plant flowering height perhaps contributes in a symbiotic fashion to the secondary production of pollinators and species diversity at the primary and secondary trophic levels. Although flowering onset time appeared disconnected among the study species, flowers were continuously available to pollinators because flowering duration was long enough to bridge different flowering flushes (see flowering onset/offset times in Appendix S1).

Finally, we want to point out that plant stem density (stems per area) might have also contributed to species coexistence in the grassland studied. We observed that the tallest species such as L. canadensis and S. nemoralis became taller than many of the other species after late July, thereby possibly resulting in asymmetric competition. However, population density of these two species is too low (less than 0.1 m−2 for both) to competitively exclude their opponents. In the future, the density effect, as well as some other factors affecting the outcome of competition, should be integrated into studies on height growth strategies in order to further improve the understanding of species coexistence in dense grasslands.


We thank Peter B. Reich for insightful comments on early versions of this manuscript, Yale School Forest and Oswald Schmitz for providing research facilities, and Dror Hawlena and Kathy Huges for their assistance in the field. This research was supported by Nanjing University and the ‘100-Talent Program’ of the Chinese Academy of Sciences.