Folia Geobotanica

, Volume 51, Issue 2, pp 143–159

Plant functional traits – fixed facts or variable depending on the season?

  • Christine Römermann
  • Solveig Franziska Bucher
  • Melanie Hahn
  • Markus Bernhardt-Römermann

DOI: 10.1007/s12224-016-9250-3

Cite this article as:
Römermann, C., Bucher, S.F., Hahn, M. et al. Folia Geobot (2016) 51: 143. doi:10.1007/s12224-016-9250-3


Traits are widely used to detect and explain responses of ecosystem processes to environmental changes. Various studies use trait data from databases, often providing one value per trait and species, neglecting intraspecific trait variability along spatio-temporal gradients. Handbooks for standardized trait measurements claim that traits should be measured at an ‘optimal’ stage, which is typically defined to be reached when the plants are in full blossom. However, it is unclear whether this method is appropriate. The main aim of this study was to quantify the extent to which trait values vary with season and phenology, a type of variation that has been overlooked so far relative to other sources of intraspecific variation. Further, we aimed to investigate whether species rankings remain consistent throughout the year.

From April to November 2012, we monitored seven leaf traits [specific leaf area SLA, leaf dry matter content, chlorophyll fluorescence parameters (Fv/Fm, performance-index PI], stomatal density, stomatal size and the stomatal pore area index SPI) of 15 summer green woody species weekly under controlled conditions. In parallel, we recorded phenological stages.

The results showed that all traits varied significantly throughout the year in a species-specific manner. We detected trait relationships with vegetative but not with flowering phenology. Species rankings were inconsistent throughout the season in all traits.

We concluded that the seasonal timing of trait measurements is crucial. Most notably SLA, Fv/Fm and stomatal size were the most robust traits in terms of small intraspecific and large interspecific variation and showed largely consistent species rankings across seasons.


Fv/Fm intraspecific trait variability LDMC phenology SLA stomatal density 


Plant functional traits are frequently used to detect and explain responses of ecosystem processes to climate and land-use changes (Westoby 1998; Lavorel et al. 1999; Römermann et al. 2009; Bernhardt-Römermann et al. 2011a). Various studies use trait data from databases, such as LEDA (Kleyer et al. 2008) and TRY (Kattge et al. 2011), which represent excellent sources of information for different traits and species on European and global scales. For the most common species and traits, these databases incorporate many values per species. However, for many other species, only one value per trait is available. It is, therefore, difficult to directly include the effect of intraspecific trait variability when using database information. Nevertheless, when using information from trait databases to analyse changes in ecosystems, we often assume that traits are consistent in space and time (Garnier et al. 2001; Al Haj Khaled et al. 2005). This is the basic idea of the ‘species robustness assumption’ (sensu Garnier et al. 2001, also called the ‘stable species hierarchy hypothesis’; Kazakou et al. 2014), which postulates that as long as the intraspecific variability remains smaller than the interspecific variability, and species rankings do not substantially change along spatio-temporal gradients, the use of one (mean) trait value per species is justified. Garnier et al. (2001) illustrated the different cases in which we can assume stability in species ranking. Species rankings remain stable if (i) traits are not plastic or similarly plastic across all species, as species’ trait responses would therefore not differ along gradients. This assumption holds true for all categorical traits such as growth form or leaf distribution along the stem, but it has been shown to be not realistic for a variety of continuous traits. Species rankings also remain stable when (ii) traits show differential plasticity in response to changes in the environment, but only if either (a) the difference in plasticity between species remains lower than their initial difference when the direction of the trait response to the environment is identical across species or (b) the sum of changes in the trait values of the species remains lower than their initial difference when the direction of the trait response to the environment differs between species. As a consequence, species rankings may also remain consistent if the gradient of investigation is relatively short, though the term ‘short’ needs to be defined for each trait and type of gradient.

Up till now, several studies have focused on trait consistency along different spatial scales; Roche et al. (2004) and Hulshof et al. (2013) observed differences in trait values along large geographical gradients, while Albert et al. (2010) and de Bello et al. (2011) found strong dependencies between trait values and locally differing environmental factors such as water, light and nutrient availability. Furthermore, there is also evidence emerging that the time of the year when traits are measured affects trait values (Jurik 1986a,b; Garnier et al. 2001; Al Haj Khaled et al. 2005; Palacio et al. 2008; Dubey et al. 2011; Karavin and Kilinc 2011). Jurik (1986a) noted that due to the observed seasonal variability of traits, the definition of an entire multi-dimensional response surface of a trait to a parameter or gradient of interest would require extensive samplings over weeks or months. However, for most analyses focusing on ecosystem changes, this is neither feasible nor straightforward. Thus, we need to establish rules governing the use of trait data, considering intraspecific variability along spatio-temporal gradients on the one hand, but the feasibility of data collection on the other hand. It is crucial to further elucidate the effect of season on the intraspecific variability of traits to define such regulations.

The seasonal variation has only been investigated for few traits, for small groups of species or with a dataset limited to only a few collections in the year. Patterns over a species’ entire leaf lifespan have not yet been described. For selected tree species, Jurik (1986a,b) revealed that leaf mass per area (the inverse of specific leaf area, SLA) and photosynthetic capacity changed with season in northern hardwood tree species. In an extensive study, Garnier et al. (2001) showed that though SLA and leaf dry matter content (LDMC) varied with season, species rankings mostly remained consistent over time in herbaceous and woody species in a Mediterranean environment. In contrast, Palacio et al. (2008) found that seasonal intraspecific variations in LDMC exceeded seasonal interspecific variations in 12 woody Mediterranean species, leading to non-consistent species rankings; species rankings were clearly affected by the month of sampling. Al Haj Khaled et al. (2005) showed that LDMC and SLA were variable over the growing season but species rankings were conserved over different nutrient levels in grassland species in southern France. Dubey et al. (2011) demonstrated that different life forms exhibited different seasonal patterns in various leaf traits, possibly also leading to non-consistent rankings of species groups of dry tropical forest herbs. These partially contradictory findings imply that further research focusing on the seasonal variability of plant traits is urgently needed to accumulate sufficient evidence to formulate general rules on how to collect and use (mean) trait data.

In this study, we aimed to reveal the importance of the seasonal variability for leaf traits in 15 woody species and to test the species robustness assumption (sensu Garnier et al. 2001; Kazakou et al. 2014) in matters of seasonal trait variability. More specifically, we tested whether species rankings in selected leaf traits remain consistent throughout the year. We went one step further by also investigating the effect of plant phenology on trait values, since most handbooks for standardized trait measurements claim that traits should be measured at an ‘optimal’ stage, which is typically defined to be reached when the plants are in full blossom (Cornelissen et al. 2003; Knevel et al. 2003; Perez-Harguindeguy et al. 2013). However, for practical reasons, traits are often measured in parallel to vegetation samplings, including species that are not flowering. But it is unclear whether this method is justified and whether leaf traits are related to phenology at all. To overcome the limitations of previous studies on this topic, our measurements were conducted weekly and included the full life cycle of the leaves.

We focused on seven leaf traits, namely, SLA and LDMC as well as parameters describing the photosynthetic efficiency as deduced from chlorophyll fluorescence measurements (Fv/Fm, performance index PI) and stomatal traits (stomatal density, stomatal size and the derived stomatal pore area index, SPI). SLA positively relates to relative growth rates across species and scales positively with photosynthesis rates (Perez-Harguindeguy et al. 2013). LDMC correlates positively with leaf lifespan and negatively with relative growth rates. High LDMC species tend to be more resistant to hazards whereas low LDMC species tend to occur in productive, often highly disturbed environments (Perez-Harguindeguy et al. 2013). Consequently, SLA and LDMC as prominent leaf economic traits are frequently used to understand plant species and ecosystem responses to changes in environmental conditions (e.g. Craine et al. 2001; Kahmen and Poschlod 2004; Wright et al. 2005; Römermann et al. 2011a; Bernhardt-Römermann et al. 2011b; Lauterbach et al. 2013). Fv/Fm is related to the efficiency of Photosystem II (PSII) electron transport under changing environmental conditions (White and Critchley 1999; Griffin et al. 2013) and is widely used by plant physiologists as a sensitive measure of plant photosynthetic activity. PI is essentially an indicator of plant vitality and indicates a plant’s ability to resist constraints from the outside (Strasser et al. 1999; Clark et al. 2000; Strasser et al. 2000). Chlorophyll fluorescence is increasingly used in ecology to assess the vitality and performance of plants since this non-destructive method is fast as well as cost effective and requires a minimum of expertise in contrast to the traditional gas exchange measurements (e.g. Schreiber et al. 1995; Strasser et al. 1999; Clark et al. 2000; Dias and Brüggemann 2010). Stomatal traits are important hydraulic traits and mediate the gas exchange between the plants photosynthetic layer and its atmosphere. Stomatal traits were shown to be related to the water use efficiency in plants on inter- as well as intraspecific levels (Bucher et al. 2016). The density of stomata and their diffusive conductance are positively correlated with net-photosynthesis rates and biomass production (Woodward et al. 2002). They are more and more used to understand responses of plants to changing environmental conditions (Franks et al. 2012; Stojnic et al. 2015; Bucher et al. 2016; Carlson et al. 2016). By quantifying how these traits change across the season and with phenology, we aim to contribute to a better understanding of this source of intraspecific trait variability.

Material and methods

Study site, species selection and trait measurements

Encompassing the whole growing season of the year 2012 (from 25 April to 2 November 2012), we monitored the plant traits presented in Table 1 weekly under controlled field conditions in the Botanical Garden of the University of Regensburg (south Germany). Measurements started when leaves were completely unfolded and continued until leaf fall and as long as green leaves for measurements were available. Once leaf colouring started, we selected green, healthy leaves for trait measurements.
Table 1

Overview of the selected plant species, families, species abbreviations used in the figures and tables, information on whether species started flowering before the first leaves emerged as well as value ranges [min, max] of measured leaf traits. Specific leaf area (SLA) is given in mm2·mg−1, leaf dry matter content (LDMC) is given in mg·g−1, the chlorophyll parameters Fv/Fm and PI are both dimensionless. Stomatal density is given in numbers per mm2 and stomatal size in μm2. The stomatal pore area index SPI is dimensionless. The last three rows give the factors by which traits vary among species (calculated as the global maximum divided by the global minimum of the respective trait) and within species (min and max indicate the variation for the least and the most variable of the species, respectively)

Species name


Species abbreviation

Flowers before leaves





Stomatal density

Stomatal size


Acer pseudoplatanus L.




[10.72; 31.66]

[226.05; 372.44]

[0.77; 0.83]


[211.36; 449.45]

[129.47; 220.89]

[5.40; 12.30]

Alnus glutinosa (L.) P. Gaertn.




[17.16; 31.13]

[260.13, 414.34]

[0.76; 0.85]


[100.59 432.90]

[227.64; 555.30]

[4.79; 19.10]

Betula pendula Roth




[9.03; 26.95]

[254.45; 507.59]

[0.58; 0.85]


[84.03; 171.89]

[304.34; 759.01]

[5.79; 19.00]

Cornus mas L.




[17.84; 36.08]

[292.41; 422.36]

[0.40; 0.85]

[0.05; 12.75]

[67.48; 151.52 ]

[387.00; 937.57]

[5.96; 21.70]

Corylus avellana L.




[12.13; 31.21]

[293.30; 471.22]

[0.43; 0.85]

[0.06; 9.96]

[72.57; 154.06]

[269.98; 983.59]

[3.47; 14.20]

Crataegus laevigata (Poir)DC.




[12.35; 25.65]

[325.55; 566.51]

[0.70; 0.85]

[0.39; 12.34]

[58.57; 152.79]

[382.88; 647.95]

[6.79; 16.80]

Forsythia europaea




[12.45; 25.65]

[240.58; 389.13]

[0.49; 0.85]

[0.16; 12.86]

[117.14; 216.45]

[254.03; 392.09]

[7.85; 17.10]

Fraxinus excelsior L.




[14.24; 40.53]

[196.73; 420.54]

[0.63; 0.84]

[0.77; 12.57]

[87.85; 292.85]

[217.21; 416.63]

[6.35; 17.10]

Prunus spinosa s.l. L.




[12.41; 29.51]

[279.37; 443.34]

[0.64; 0.85]

[0.50; 13.17]




Rosa canina s.l. L.




[10.17; 21.29]

[318.54; 456.59]

[0.68; 0.85]

[0.78; 15.88]

[73.85; 166.79]

[159.53; 509.28]

[6.46; 18.20]

Salix caprea L.




[10.80; 33.05]


[0.08; 0.86]

[0.00; 17.34]

[96.77; 606.06]

[73.02; 195.74]

[3.05; 11.40]

Sambucus nigra L.




[10.26; 26.63]

[204.16; 309.36]

[0.48; 0.83]

[0.04; 6.77]

[67.48; 136.24]

[751.04; 1295.91]

[12.60; 29.20]

Sorbus aucuparia L.




[21.18; 32.59]

[305.62; 489.50]

[0.30; 0.85]

[0.09; 9.31]

[57.30; 92.95]

[277.96; 450.38]

[4.00; 6.23]

Syringa vulgaris L.




[6.74; 17.64]

[313.26; 420.26]

[0.77; 0.86]

[1.57; 29.28]

[75.12; 350.14]

[187.15; 484.74]

[5.30; 21.40]

Tilia platyphyllos Scop.




[22.70; 32.53]

[196.42; 341.75]

[0.81; 0.86]

[1.17; 13.27]

[68.75; 201.17]

[247.28; 385.34]

[3.24; 14.00]

Variation among species (factor):








Variation within species [min, max] (factor):

[1.43; 3.06]

[1.34; 2.14]

[1.06; 11.25]

[7.02; 17336]

[1.62; 6.26]

[1.26; 3.64]

[1.56; 4.32]

Measurements were carried out on 15 summer green, woody species (Table 1). We decided to focus on individuals growing under optimal conditions in the botanical garden with sufficient nutrient and water supply to assess the seasonal variability. We performed all measurements on the same individual adult plant throughout the year. This approach directly controlled for the potential effect of genotypic variation on seasonal trait variability. It was not possible to add replicate individuals as the selected species were represented only once in the Garden. Hence, analyses are based on intra-individual variation, but we assume that there is negligible intraspecific variation in how traits change along the growing season within individuals. Therefore we refer to intraspecific variation in the following.

Once a week, the seven plant traits were measured according to well-established methods described in handbooks for standardized measurements (Cornelissen et al. 2003; Knevel et al. 2003; Grant and Vatnick 2004; Perez-Harguindeguy et al. 2013). Two replicate leaves exposed to full sunlight were chosen for trait measurements on each sampling event. We restricted the weekly samplings to only two leaves to ensure that the trees could support the destructive sampling. We controlled for leaf age by always choosing leaves showing the same size as 90 % of the leaves present per sampling event.

We measured SLA as the ratio of fresh leaf area to dry mass, expressed in mm2·mg−1. Leaf area was measured on scanned fresh leaves using the LeafTraits package for R (Bernhardt-Römermann, unpubl.). Leaf dry matter content (LDMC), the ratio of dry mass to fresh mass, is a measure of tissue density and was measured in parallel to SLA by weighing the fresh leaves immediately after collection.

Additionally, two parameters (Fv/Fm and the performance index PI) related to photosynthetic efficiency were recorded using chlorophyll fluorescence measurements with a portable continuous excitation time-resolved chlorophyll fluorimeter (PocketPEA from Hansatec). Fv/Fm, the proportion of variable fluorescence (Fv) to maximum fluorescence (Fm), is an indicator for the potential photosynthetic activity and the quantum yield of a leaf. Maximum values of Fv/Fm for unstressed plants are usually in the range of 0.8 to 0.85; this value decreases as the plant’s photosynthetic ability declines due to any form of stress. PI is a measure for photosynthetic performance and gives the probability that an absorbed photon is trapped by a reaction centre in Photosystem II (for details on the exact algorithms implemented to calculate PI, see Strasser et al. 2004). Pre-dawn chlorophyll fluorescence measurements were carried out on two replicate, fully dark-adapted leaves.

Three traits that describe stomatal characteristics were measured: stomatal density and size were measured on the lower surface of the leaves at approximately the same position using the technique of clear nail polish impression of leaf stomata (Hilu and Randall 1984). Stomatal counts were conducted for each of two different fields of view per leaf sample. In each field of view, guard cell length and width were measured on two different stomata, from which we calculated the approximate stomatal size (A = ¼ × π × length × width). Measurements were conducted using a 400-fold magnification (objective 40×, ocular 10×) with a Zeiss microscope; units were converted to millimetres. As stomatal conductance depends on both stomatal density and size of the stomatal aperture, we also calculated the dimensionless stomatal pore area index (SPI) as described by Sack et al. (2003) SPI integrates the information on density and size following the formula SPI = (guard cell length)2 × stomata density. It was not possible to perform the stomatal counts and measurements on Prunus spinosa due to their leaf glands.

To analyse whether trait values are associated with plant phenology, we recorded the phenological phases of the species weekly during the period starting on 15 March 2012 (day 74 of the year) until 2 November 2012 (day 307). We monitored the different growth stages and reproductive phenological events following the methods described by Dierschke (1972). Table S1 gives an overview on the monitored phases, which are divided into vegetative and generative phases.

Statistical analyses

To investigate whether plant trait values depend on season and phenology and whether trait responses were species-specific, we used linear models with the plant trait as the dependent variable and day of the year, its quadratic term (day of the year2), the monitored generative and vegetative phenology and species identity as explanatory variables. Additionally, we included the interactions between species identity and day of the year (day:species, day2:species) as well as species identity and vegetative or generative phenology (vegetative phenology : species, generative phenology : species) as explanatory variables. Models were simplified using stepwise backwards selection, where we always excluded the least significant variables until the final minimal adequate model was reached (following the procedure described in Crawley 2007). After each simplification step, we checked whether the model simplification was justified by comparing the models using analyses of variance. Because we dealt with time series data, we considered temporal autocorrelation and possibly heteroskedasticity by correcting the covariance structure of the linear models using heteroskedasticity and autocorrelation consistent (HAC) covariance estimators (Zeileis 2004). To meet model assumptions (normal distribution and homogeneous variance in residuals), SLA, stomatal density, stomatal size and SPI were log-transformed prior to analyses.

To examine whether the consistent species rankings hypothesis holds true, i.e. whether species rankings varied with season, we performed Spearman rank correlation tests on species ranks following the approach described in Garnier et al. (2001). For each trait and sampling event, species ranks were deduced from the dataset after ordering the species by their mean trait value.

To support the interpretations of the seasonal variations in plant traits, we analysed whether trait-trait correlations exist and whether these differ between species or change in the course of the year. In a first step, Pearson correlations were performed on the full dataset, i.e. including data from all species and all sampling events. As described above, some traits were log-transformed to achieve linearity. In a second step, we included species as covariate both alone and in interaction with the trait of interest in these models to investigate whether detected trait-trait relations hold true when focussing on within-species patterns. When the effect of species was significant in these models, we tested per species whether the slope of the regression line was significantly different from zero. In a third step we run the same linear models as described above but included day of the year as covariate instead of species to investigate whether trait-trait correlations hold true across the year. In this final model, we tested for break-points in the trait-trait relationships that may occur in the course of the year using the segmented function from the R package ‘segmented’ (Muggeo 2003; Muggeo 2008). Segmented relationships are defined by the slope parameters and the break-points where the linear relation changes. Initial values for the break-points were set to those days in the year when measurements were carried out (N = 24).

All calculations were performed using R 3.1.1 (R Foundation for Statistical Computing 2014), with the additional packages ‘lmtest’ (Zeileis and Hothorn 2002), ‘sandwich’ (Zeileis 2004) and ‘ltm’ (Rizopoulos 2006) besides the ‘segmented’ package (Muggeo 2003, 2008).


Effects of season and phenology on trait values

The time series of trait data showed that all investigated traits varied significantly throughout the year and with phenology (except for SLA, which was not affected by phenology). Species identity had a significant influence on trait values and their variation throughout the year. Model results are summarized in Table 2, and trait variation with season and phenology are shown in Fig. 1, Fig. 2 and Fig. 3 for selected species. Responses of all species are shown in in Fig. S1 to Fig. S7 of the Electronic Supplementary Material.
Table 2

Results of the linear models explaining the plant traits specific leaf area (SLA; log-scaled), leaf dry matter content (LDMC), Fv/Fm, performance index (PI), stomatal pore area index (SPI, log-scaled), stomatal density (log-scaled) and stomatal size (log-scaled) with day of the year, day of the year2, species, species phenology (generative phases and vegetative phases) as well as the interactions with species. The F-statistics and the R2 coefficient are given in the first row. Other rows indicate the effects of the different included explanatory variables, with *** P < 0.001, ** P < 0.01, * P < 0.05, + P < 0.1, n.s. – not significant. ‘(n.s.)’ indicates that this variable was not included in the final minimal adequate model






Stomatal density

Stomatal size


Model results

R2 = 0.81***

F44, 291 = 27.95

R2 = 0.83***

F84, 252 = 15.10

R2 = 0.66***

F85, 253 = 5.79

R2 = 0.79***

F85, 253 = 10.97

R2 = 0.85*** F48,249 = 29.12

R2 = 0.91*** F81, 212 = 27.24

R2 = 0.69*** F48,248 = 11.32









Day of the year








Day of the year2








Generative phases








Gener. phase:Species








Veget. phase:Species








Vegetative phases
























Fig. 1

Seasonal changes in the leaf morphological traits SLA (upper row) and LDMC (lower row) for three selected plant species. The lines represent the results of the linear models (see Table 2) explaining the variations in SLA and LDMC, including the data of all investigated species. For simplicity, only aggregated phenological phases are drawn. The grey bar indicates the time of flowering (from the start of phase 4 to the end of phase 8; for a definition of phenological phases please see Table S1 in the Electronic Supplementary Material); the shaded area (loosely striped) shows the period when leaves were green (from the start of phase 2 to the end of phase 8); and the intensely shaded area indicates the period when leaves started changing colour (from the start of phase 7 until the end of phase 10). Phenological observations started on 15 March 2012 (day 74) and ended on 2 November 2012 (day 307). Responses of all species are shown in Fig. S1 and Fig. S2 of the Electronic Supplementary Material

Fig. 2

Seasonal changes in the photosynthetic traits Fv/Fm (upper row) and PI (lower row) for three selected plant species. The lines represent the results of the linear models (see Table 2) explaining the variations in Fv/Fm and PI, including the data of all investigated species. For simplicity, only aggregated phenological phases are drawn. The grey bar indicates the time of flowering (from the start of phase 4 to the end of phase 8; for a definition of phenological phases please see Table S1 of the Electronic Supplementary Material); the shaded area (loosely striped) shows the period when leaves were green (from the start of phase 2 to the end of phase 8); and the intensely shaded area indicates the period when leaves started changing colour (from the start of phase 7 until the end of phase 10). Phenological observations started on 15 March 2012 (day 74) and ended on 2 November 2012 (day 307). Responses of all species are shown in Fig. S3 and Fig. S4 in the Electronic Supplementary Material

Fig. 3

Seasonal changes in the stomatal traits (first row: stomatal density per mm2, second row: stomatal size in μm2 and last row: SPI) for three selected plant species. Lines represent the results of the linear models (see Table 2) explaining the variations in each of these traits including the data of all investigated species. For simplicity only aggregated phenological phases are drawn. The grey bar indicates the time of flowering (from the start of phase 4 to the end of phase 8; see Table A.1); the shaded area (loosely striped) shows the period when leaves were green (from the start of phase 2 to the end of phase 8); and the intensely shaded area indicates the period when leaves started changing colour (from the start of phase 7 until the end of phase 10). Phenological observations started on 15 March 2012 (day 74) and ended on 2 November 2012 (day 307). Responses of all species are shown in Fig. S5, Fig. S6 and Fig. S7 in the Electronic Supplementary Material

SLA varied over time in a species-specific manner. Neither the vegetative nor the generative phenological phases had an impact on SLA (P > 0.1; Fig. 1, Fig. S1, Table 2). For all species but Acer pseudoplatanus, SLA maxima were reached in early spring; A. pseudoplatanus showed a midseason peak.

LDMC varied over time and with vegetative phenological phase. In all cases, species had an effect alone and in interaction with these terms. Fig. 1 and Fig. S2 illustrate the seasonal changes and underline the effect of phenology. There was no significant effect of generative phase on LDMC (P > 0.05; see Table 2), but the vegetative phase determined trait values (P < 0.001). While a number of species had a midseason peak in LDMC (e.g. A. pseudoplatanus, Cornus mas, Fraxinus excelsior, Rosa canina; Fig. S2), in others, LDMC increased (e.g. Crataegus laevigata) or followed the shape of a saturation curve over time (e.g. A. glutinosa). The time at which LDMC started to decrease or stabilize coincided with the time that leaves began changing colour; this was particularly apparent for A. glutinosa, Betula pendula, F. excelsior and R. canina (Fig. S2).

Fv/Fm varied over time and with vegetative phenological phase. The responses were highly species-specific (Table 2, Fig. 2, Fig. S3). Depending on the species, the maxima were either reached in the middle or at the end of the vegetation period before leaves started to change colour (Fig. 2 and Fig. S3a). From Fig. 2 and Fig. S3a, we can deduce that the effects of season and phenology were largely determined by values measured towards the end of the year; however, when these values were excluded from the models, the general results did not change (Fig. S3b).

PI was strongly determined by the day of the year and the vegetative phenology, and there was a clear species effect on the slope and the intercept on the hump-shaped relationships (Table 2; Fig. 2, Fig. S4).

Stomatal density varied with season and vegetative phenology in a species-specific manner (Table 2). While some species showed an almost linear decrease in stomatal density over time (e.g. Syringa officinalis), others showed a hump-shaped relationship (e.g. A. glutinosa). For most species, only marginal changes were observed (e.g. R. canina, Fig. 3 and Fig. S5).

Stomatal size was related to season and generative phenology (Table 2, Fig. 3 and Fig. S6). While most of the species showed an increase in stomatal size (e.g. R. canina, C. laevigata, Forsythia. europaea) or a hump-shaped reaction (e.g. Syringa vulgaris, Sambucus nigra), in others, stomatal size decreased over time (e.g. A. glutinosa, Salix caprea).

The species-specific changes of SPI with season are shown in Fig. 3 and Fig. S7. Vegetative phenology impacted trait values, but this effect was not species-specific (Table 2, Fig. S7).

Intraspecific vs interspecific variation

Table 1 gives an overview of the ranges of trait values as well as the magnitude of variation on both the inter- and intraspecific levels. The interspecific variation was much higher than the intraspecific variation for SLA (a factor of 6 compared with a factor of 3 for the most variable species, S. caprea), stomatal density (a factor of 19.6 compared with a factor of 6.3 for S. caprea), stomatal size (a factor of 17.8 vs a factor of 3.6 for the most variable species, Corylus avellana) and SPI (a factor of 9.6 vs a factor of 4.3 for Tilia platyphyllos). The intraspecific variation was comparable to the interspecific variation in LDMC and Fv/Fm. LDMC varied by a factor of 2.88 among species and by a factor of 2.14 for the most variable of these species (F. excelsior), whereas Fv/Fm varied with factors of 11.3 and 11.25, respectively (S. caprea). PI varied across multiple orders of magnitude on both inter- and intraspecific levels. When leaf senescence began, these values approached zero, leading to these high factors.

Consistent species rankings over time

Species rankings were not entirely consistent for the investigated traits (Table 3). For SLA, LDMC, stomatal density and stomatal size, the number of significant comparisons exceeded the number of non-significant comparisons. For Fv/Fm, PI and SPI, the pattern was reversed, and negative correlations were even detected. Most of the non-significant correlations were found for the comparisons that included the early and late season days (Table S2).
Table 3

Species rankings for the different traits. Summary of the Spearman correlation coefficients (ρ) comparing the rankings of species according to their traits from different sampling days. Given are the minimum and maximum ρ that were obtained in the pairwise correlations between the rankings based on the traits per sampling day as well as the number of significant and non-significant results. The full correlation matrices can be found in Table S2 of the Electronic Supplementary Material.






Stomatal density

Stomatal size


Min ρ








Max ρ








Number of significant ρ’s








Number non-significant ρ’s








Trait- trait correlations

Table 4 gives an overview of the results of the trait-trait correlations. The left panels in Fig. 4a and Fig. 4b show the results of considering species-specific trait-trait correlations (‘species models’). The right panels of Fig. 4a and Fig. 4b show the output of the segmented regression analysing the break-points where the trait-trait relations changed in the course of the year (‘year models’).
Table 4

Trait-trait correlations of all investigated species and sampling days; given are the Pearson correlation coefficients (R2) along with the P-values resulting from pairwise correlations. Italic letters indicate log-relationships between the two given parameters. *** P < 0.001, ** P < 0.01, * P < 0.05, n.s. – not significant






Stomatal density

Stomatal size

Leaf dry matter content (LDMC)



Photosynthetic capacity (Fv/Fm)




Performance index (PI)





Stomatal density






Stomatal size







Stomatal pore area index (SPI)







Fig. 4

a – Trait-trait correlation of the leaf morphological and chlorophyll fluorescence traits separated according to species (left panels; ‘species models’) and day of the year (right panel, ‘year models’). All species and year models revealed highly significant results. In the species models, only the regression lines for those species showing slopes that were significantly different from zero are shown. In addition, the 95 % confidence intervals are drawn. Different time periods in the year model were determined by the segmented regression analyses. b – Trait-trait correlation of the stomatal parameters stomatal size (in μm), stomatal density (per mm2) and stomatal pore area index SPI (dimensionless) separated according to species (left panels; ‘species models’) and day of the year (right panel, ‘year models’). All species and year models revealed highly significant results. In the species models, only the regression lines for those species showing slopes that were significantly different from zero are shown. In addition, the 95 % confidence intervals are drawn. Different time periods in the year model were determined by the segmented regression analyses

SLA and LDMC were significantly negatively correlated (R2 = 0.19, d.f. = 332, P < 0.001). This negative relationship could be confirmed for 11 out of the 15 species by the species models (model result testing for the relation of SLA and LDMC with species alone and in interaction with LDMC: R2 = 0.79, F29,304 = 40.21, P < 0.001). These 11 species had species-specific slopes and intercepts (Fig. 4a top row). The year model revealed that the SLA-LDMC relation changed several times in the course of the year: it was weakest in spring (< 9 May) and midsummer (> 3 July). In the period in-between, the relationship was largely consistent, though also divided into two periods by the model, but with only minor differences in the slope (Fig. 4a top row).

SLA and PI were significantly negatively correlated on the logarithmic scale (R2 = 0.22, d.f. = 334, P < 0.001). Less than half of the species (N = 7) followed this overall negative and logarithmic relationship with species-specific slopes and intercepts (Fig. 4a second row; model result testing for the relation of SLA and PI with species alone and in interaction with PI: R2 = 0.71, F29,306 = 25.23, P < 0.001). The strength in the PI-SLA association changed in the course of the year; before 11 July, stronger associations (i.e. steeper slopes) were observed compared to later days in the year (Fig. 4a second row).

PI and Fv/Fm were positively correlated on a logarithmic scale (R2 = 0.75, d.f. = 337, P < 0.001), which could be confirmed for all species investigated, but with species-specific slopes and intercepts (Fig. 4a third row; model result testing for the relation of PI and SLA with species alone and in interaction with SLA: R2 = 0.84, F29,309 = 56.15, P < 0.001). The general relationship between these two parameters mainly changed in springtime and remained consistent after 24 June (Fig. 4a third row).

Stomatal size was correlated with stomatal density on a log-log scale (R2 = 0.46, d.f. = 192, P < 0.001) with individuals having either many small or few large stomata. The model testing for the relation of stomatal size and density with species alone and in interaction with stomatal density was highly significant (R2 = 0.79, F29,264 = 33.18, P < 0.001). Only three species showed this logarithmic negative association between stomatal size and density, while T. platyphyllos showed a positive relation (Fig. 4b first row). The overall trait-trait correlation remained largely consistent in the course of the year (Fig. 4b first row).

SPI and stomatal size were positively related (R2 = 0.31, d.f. = 334, P < 0.001). This positive relationship was confirmed for six out of the 15 species by the species model (model result testing for the relation of SPI and stomatal size with species as covariate: R2 = 0.59, F15,278 = 26.64, P < 0.001. All species had species-specific slopes and intercepts (Fig. 4b last row). The year model revealed that the general relationship only slightly changed in the course of the year (Fig. 4b last row).

The remaining trait-trait correlations were either not significant or the R2 was extremely low (R2 < 0.07) and therefore ecologically irrelevant (Table 4). Including species or year did not improve the trait-trait correlation.


The results clearly demonstrate that all investigated traits varied substantially throughout the year and partly with phenology and that all variations were highly species-specific. For most traits, species rankings did not remain consistent, neither over time nor with phenology. The main direction of trait-trait relationships remained consistent across species and in the course of the year. The intensity and magnitude of these relationships, however, differed in a species-specific manner and changed during the year. Overall, these findings may have important consequences for the applicability of mean trait values, either deduced from large trait databases or a researcher’s own trait measurements.

Seasonal variation in leaf morphological traits (SLA and LDMC)

There was little variation in SLA after an early season change that was not related to the monitored leaf phenological phases. LDMC increased at the beginning of the season and then decreased as leaves started colouring in most of the evaluated species; vegetative phenology had a significant impact. The early season changes in SLA and LDMC can be attributed to the initial period of leaf growth or laminar expansion when relative rates of cell division and cell expansion change (Steer 1971). Once cell expansion has ceased, dry matter, especially in the form of non-structural carbohydrates (Palacio et al. 2008), accumulates in the cells, leading to the observed decrease in SLA and increase in LDMC. After these initial changes, SLA and LDMC and also their relation varied little over time, reflecting the relative constancy of environment and constancy of leaf developmental stage in the summer and early autumn (compare also Jurik 1986b). During the course of leaf senescence, proteins, nucleic acids and other nitrogenous compounds are degraded and re-translocated into other organs (Dubey et al. 2011). However, we did not find a clear change in SLA during leaf senescence. In contrast, LDMC decreased in most species by the end of the season, most likely due to the described re-translocations.

We suggest measuring LDMC and SLA at a time not too early in the year but rather when leaves have certainly reached maturity, i.e. by the end of May or in June for Central European species, as both were quite consistent during the summer months. As SLA was not determined by phenology and as it was relatively constant after an initial early season decrease, it may be considered a preferred trait for functional ecological studies aiming to characterize species by their average trait values and having only a restricted time frame for trait measurements during the year. Accordingly, SLA values as deduced from databases should be less affected by the time when measurements were carried out.

Also species rankings for SLA remained relatively constant throughout the year for about 80 % of the species investigated and the intraspecific variation was much smaller compared with the interspecific variation (a factor of 3.1 compared with 6.0, respectively). Compared with other traits (Table 4), SLA was the most robust and constant trait throughout the year.

Even though we analysed only one individual per species (and, hence, the intra-individual variability as one component of the intraspecific variability) and focussed on tree species only, we could reveal clear patters. Moreover, our results are in accordance to the findings of other studies: Garnier et al. (2001) also showed that classifications based on SLA were more stable across seasons compared with LDMC in Mediterranean herbaceous and woody species. Dubey et al. (2011) analysed forest herbs in dry tropical environments and showed that the seasonal pattern in SLA was consistent across four different life forms (annual and perennial herbs and grasses). In contrast, along a climatic gradient in Mediterranean environments, SLA was variable, but LDMC fairly consistent (Roche et al. 2004).

Seasonal variation in Fv/Fm and PI

The chlorophyll fluorescence measurements revealed that Fv/Fm and PI were strongly influenced by season and vegetative phenology, with a clear species effect. Both traits showed a distinct early season increase, which could be related to higher light availability and increased temperatures in spring, assuming that soil moisture and nutrients did not limit plant growth in our study. Afterwards, at a time when leaves were likely to have reached maturity, Fv/Fm was quite steady until leaf fall. This pattern can be related to the fairly constant light conditions over the growing season, as described by Jurik (1986a) in successional northern hardwood tree species. At the end of the growing season, Fv/Fm quickly declined, which might represent diminished function resulting from age-related changes and senescence of foliar tissue. The same pattern was also observed for deciduous but not for evergreen oak in an experimental setting (Holland et al. 2014). Even if the microenvironment is stable over time, cumulative damages and reduced internal conductance of CO2 can lead to lower potential photosynthetic capacities in older leaves (see also Kikuzawa and Lechowicz 2011).

In contrast, PI did not show the same constant period during the growing season, but appeared to follow the same seasonal pattern as previously described for Amax (see Kikuzawa and Lechowicz 2011 and references therein). Amax shows a clear midseason peak followed by a gradual decline, likely linked to age-related processes. Hence, PI as an indicator of plant vitality and a plant’s ability to resist constraints from the outside, seems to be much more temperature-dependent than Fv/Fm (see also Holland et al. 2014). The effect of day length and irradiation might also be important, though this has not explicitly been tested. Our findings emphasize that Fv/Fm seems to represent the potential photosynthetic capacity of an individual plant under optimal environmental conditions (i.e. the potential performance of photosystem II; Strasser et al. 2004), whereas PI instead indicates the realized photosynthetic performance under prevailing conditions. However, further studies are needed to more clearly unravel the driving environmental factors of Fv/Fm and PI and to include more than one individual tree per species to test whether all individuals of the same species show the same responses to changing conditions as assumed in this study.

The intraspecific variation was approximately in the same order of magnitude as the interspecific variation in Fv/Fm and PI. The extremely high intra- and interspecific variation in PI values can most likely be attributed to the fact that we also included the measurements on senescent (though still greenish) leaves where the PI was close to zero. Species rankings for neither Fv/Fm nor PI were consistent, as the number of non-significant correlations of species ranks exceeded by far the number of significant correlations. However, we demonstrated that Fv/Fm and PI are highly correlated with one another and that this correlation holds true across the species and largely over the season. As Fv/Fm is relatively constant across the season, we might still use measurements from only one sampling event as a crude proxy for plant photosynthetic performance in studies aiming to describe species based on their mean trait values.

Seasonal variation in stomatal traits

All stomatal traits strongly changed with season and with phenology, and the seasonal trait responses were species-specific. In some species, stomatal density decreased, in some it increased, and for others it remained fairly constant during the primary growing season. In contrast, stomatal size increased over time in most of the species. The observed decrease in stomatal density in some species and the increase in stomatal size in most species could be attributed to leaf growth: an increase in leaf area and the associated cell growth inevitably leads to a decrease in stomatal density per unit leaf area and an increase in stomatal size. The observed increase in stomatal density in other species may be attributed to the fact that stomata are not only initiated during the early stages of leaf expansion but sometimes also during leaf growth (Gay and Hurd 1975). For example, Casson and Gray (2008) noted that stomata develop late in leaf development. However, it is likely that the ability to initiate stomata through the later part of the year depends on the species, leading to the observed species-specific responses. Environmental conditions such as light levels, water availability and CO2 concentration also determine stomatal density (see Casson and Gray 2008 and references therein; Bucher et al. 2016), but, as discussed above, these factors are unlikely to have driven our observed species-specific responses. In addition, stomatal density is influenced by leaf position on the tree (sun leaf vs shade leaf) and by the sampling position on the leaf (Poole et al. 1996). However, we minimized the sampling error by carefully selecting only sun leaves and by taking the stomata samples from approximately the same area of the leaf throughout the year. We assume that both leaf position on the tree and sample position on the leaf have not largely affected our dataset, as in this case, the single data points would have been more scattered across the year without providing any clear trend.

SPI showed a midseason peak for most of the species, after which the SPI decreased. Our data reflect the findings of Dubey et al. (2011), who also found a midseason peak in (measured) stomatal conductance in forest herbs in the tropics. In a previous study focusing on tree species, it was demonstrated that (measured) leaf conductance was constant throughout the season, with a slight decline in autumn (Jurik 1986a). SPI appears to be a useful parameter to be included in functional ecology, as it integrates stomatal size with stomatal density, thereby representing a driving factor of photosynthetic efficiency (Holland and Richardson 2009; Bucher et al. 2016). However, though the intra-individual variability was much smaller than the interspecific variability, species rankings were not consistent over time because the number of non-significant comparisons exceeded the significant correlations by far. We therefore propose to continue focusing on the two traits, stomatal density and stomatal size, separately as crude proxies for the potential efficiency of photosynthesis, while bearing in mind that stomatal density in particular changes with season and phenology. Thus, single measurements in the year only represent one small snap shot.


We conclude that the seasonal timing of trait measurements clearly influences the outcome of trait analyses and should therefore be taken into account in upcoming studies. We detected relationships with vegetative phenology but not flowering time (except for stomatal size). In our investigated trees, SLA, Fv/Fm and stomatal size were the most robust traits in terms of low intraspecific and high interspecific variation as well as consistent species rankings across seasons. We further confirm that measurements should be performed in midsummer, i.e. after leaves have matured and before the start of leaf senescence.

Even though our results were based on leaf trait measurements of one individual tree from each of 15 woody species growing under optimal (though artificial) conditions only, this dataset enabled us to present clear seasonal variations and to unravel some of the main processes determining trait values across seasons. Further research is needed that explicitly quantifies the relative contribution and the difference of the intra-individual and the intra-specific variation for a range of plant traits and groups of species (trees, herbs and grasses; see also Hulshof and Swenson 2010, who compared within-individual variation with intra- and interspecific variation for ten co-existing tree species, though seasonal variation was not included here). Further studies should also assess the importance of seasonal variation compared to other sources of intraspecific variation as e.g. differences in environmental conditions for a larger set of traits and species, also from outside Europe. Only with such extended approaches, profound generalizations on the need and the possibility of considering seasonal variability in trait studies will be possible. Our study should be seen only as a first step into this direction.

Depending on the relative influence of the seasonal variation on the total intraspecific variation, it may have a considerable effect on the outcome of functional analyses, as outlined in the introduction (cf. species robustness assumption). Accordingly, additional analyses should focus also on the consequences for plant functional trait studies analysing entire communities instead of single species. The ecological consequences of incorporating aspects of seasonal variability in functional analyses of plant communities cannot be evaluated from our approach and should be elucidated in further studies, though this will be extremely time consuming and hardly feasible in species rich communities.

Aspects of seasonal variation in plant traits may be explicitly addressed in the development of trait databases. By recording and providing information on the phenological state and the day of the year when measurements were carried out, more profound assessments of the effect of seasonal variation will be possible for a wide variety of traits and species in the near future. This research may inform further theoretical work and modelling approaches aiming to estimate variation based on ontogenetic processes and enable their explicit consideration in upcoming functional trait studies.


We would like to thank Patrizia König for helpful comments on the manuscript and Jennifer McCaw for assistance in the field. The research of CR was kindly supported by the German Research Foundation (DFG) project RO 3842/3-1, and the research of MBR was partly supported by the DFG project BE4143/2-1.

Supplementary material

12224_2016_9250_MOESM1_ESM.docx (930 kb)
ESM 1(DOCX 930 kb)

Funding information

Funder NameGrant NumberFunding Note
Deutsche Forschungsgemeinschaft
  • RO 3842/3-1
  • BE4143/2-1

Copyright information

© Institute of Botany, Academy of Sciences of the Czech Republic 2016

Authors and Affiliations

  • Christine Römermann
    • 1
    • 2
  • Solveig Franziska Bucher
    • 1
  • Melanie Hahn
    • 3
  • Markus Bernhardt-Römermann
    • 4
  1. 1.Plant Biodiversity Group, Institute of Systematic BotanyFriedrich-Schiller University JenaJenaGermany
  2. 2.German Centre for Integrative Biodiversity Research (iDiv)Halle-Jena-LeipzigLeipzigGermany
  3. 3.Theoretical Ecology, Institute of BotanyRegensburg UniversityRegensburgGermany
  4. 4.Institute of EcologyFriedrich-Schiller University JenaJenaGermany

Personalised recommendations