Seasonal plasticity is more important than population variability in effects on white clover architecture and productivity

Abstract

Background and Aims

The drivers of white clover (Trifolium repens) architecture and productivity are still imperfectly understood. Our aim here was to investigate the impact of genetic background, neighbourhood and season on different architectural traits, clover and total biomass yield, as well as the relationship between those traits and yield.

Methods

We grew eight white clover populations in pure stands and in mixed stands with contrasting mixture partners. Over four consecutive regrowth periods within 1 year, we measured trait sizes and determined clover and total yield amounts.

Key Results

The size of the architectural traits differed between populations and changed in response to neighbourhood and season. Population did not affect the sign or degree of those changes. Among the tested factors, season was by far the most important driver of white clover architecture, with the seasonal pattern notably differing between architectural traits. Clover and total yield were positively related to the architectural traits leaf area, petiole length, internode length and specific leaf area. Whereas the direction of the relationship was widely unaffected, its magnitude was clearly altered by neighbourhood and season.

Conclusions

Our results show that seasonal effects are the key for a deeper understanding of the architecture of white clover individuals and to improve the productivity of white clover communities.

Introduction

White clover (Trifolium repens), common in natural and sown temperate grasslands (Turkington and Burdon, 1983; Sackville Hamilton and Harper, 1989), is of considerable value for forage production in Central Europe, particularly as a component of mixtures with other species (Frame and Newbould, 1986; Abberton and Marshall, 2005; Annicchiarico et al., 2015). It shows a high level of genetic variability in diverse characteristics of its above-ground architecture (‘architectural traits’) among wild types as well as among commercial cultivars (Caradus, 1994; Caradus and Woodfield, 1997; Finne et al., 2000a). Beyond that, each of the architectural traits can plastically respond to the respective environmental conditions, thereby inducing strong architectural differences between genetically identical individuals (Bradshaw, 1965; Schlichting, 1986; Huber et al., 1999). Those architectural traits include the size of and the amount of biomass investment in above-ground plant parts (Dong, 1993; Tremmel and Bazzas, 1995; Leeflang et al., 1998; Liao et al., 2003), which have consequences for size and placement of resource-acquiring structures and biomechanical and hydraulic stability (Givnish, 1995; Huber, 1996).

In dense stands, typical for forage production, neighbours are probably the major force defining the environmental conditions (Burdon, 1980; Haynes, 1980), directly by their presence and indirectly by changing resource availability (Goldberg, 1990; Bennett et al., 2016). How any individual performs in a neighbourhood depends on both its reaction to its neighbours and their effects on resource availability, and how it in turn affects the resource availability and thus the performance of its neighbours (Goldberg, 1990; Tremmel and Bazzas, 1993). Thus, the size of architectural traits is critical in determining how successfully white clover and its neighbours perform in a neighbourhood (Turkington et al., 1991; Violle et al., 2007). Several authors have emphasized that the specific way in which architectural traits change in response to neighbours might reflect distinctive strategies depending on the relative height and density of those neighbours (Burdon, 1980; Novoplansky, 2009; Bennett et al., 2016; Gruntman et al., 2017; Bittebiere et al., 2019).

For white clover, numerous examples of neighbour-induced changes of architectural traits are available. The neighbours in these studies were almost exclusively either conspecifics or grasses (Marcuvitz and Turkington, 2000; Nassiri and Elgersma, 2002; Annicchiarico, 2003; Seker et al., 2003; Weijschedé et al., 2008; Huber et al., 2016). Some of the studies additionally tested the impact of white clover genetic background on architectural traits. To our knowledge, the effect of a forb on white clover architecture has never been investigated in detail (but see Mason et al., 2016), despite the growing interest in forbs for usage in forage production (Skinner et al., 2004; Høgh-Jensen et al., 2006; Cong et al., 2017). Moreover, the results achieved so far are at least partly contradictory. There is some evidence that these contradictions were caused by temporal dynamics in architecture. Not only do weather conditions change in the course of a year and affect the potential growth rates of all individuals (Barthram and Grant, 1994; Helgadottir et al., 2001; Anten, 2005), but white clover and its neighbours can also differ in their actual growth rates due to differences in phenological development (Barthram and Grant, 1994; Caradus et al., 1995; Elgersma et al., 1998). Therefore, white clover can be subjected to neighbours of temporally varying architecture as compared to its own (Turkington and Harper, 1979; Gruntman et al., 2017). Beyond that, agricultural management such as grazing or cutting might differently affect the architecture of white clover and its neighbours over time, for example by unequal proportions of removed plant biomass and due to varying regrowth rates (Woledge et al., 1992; Briske and Richards, 1995; Weijschedé et al., 2008).

Nonetheless, few studies have explicitly investigated the effect of season on the architecture of white clover growing together with neighbours under cutting, and those that did covered only single architectural traits (Marriott et al., 1997; Nassiri and Elgersma, 1998, 2002; Seker et al., 2003).

Further scaling-up from traits of individuals to productivity of communities is a prerequisite to ensure agronomic success, as traits that help to maximize individual performance might not act in concert with those contributing to improved community performance (Anten, 2005). Hence, several researchers have tried to directly relate various architectural traits of white clover to biomass yield, when grown with either con- or interspecifics (most commonly grasses) (Annicchiarico and Piano, 1994; Finne et al., 2000a; Annicchiarico, 2003), thereby revealing that those relationships are not universally valid. However, to date, we lack an integrated view of the (in)consistency across different neighbourhoods beyond legumes and grasses and across different seasons. This is surprising as it might help to explain and overcome variations in forage production especially for new combinations of forage plants.

Our work aimed at assessing: (1) if and how strongly important architectural traits of white clover individuals as well as the biomass yield of white clover communities are shaped by white clover genetic background and the environmental determinants neighbourhood and season; (2) which of the architectural traits are most closely related with biomass yield; and (3) if and how the environmental determinants neighbourhood and season alter the relationship between those traits and yield. To answer these questions, we used genetically distinct white clover populations and created different environments by growing them in pure or in mixed stands with a grass (Lolium perenne) or a forb (Cichorium intybus) over four consecutive regrowth periods within 1 year. The traits we measured reflect the size of and the amount of biomass investment in above-ground plant parts and have previously been shown to be linked to the performance of white clover or other clonal species elsewhere (see above).

MATERIALS AND METHODS

Study species

White clover is a clonal species that is composed of repeated modular units. Each unit consists of a root system, an internode, a petiole, a leaf and an axillary meristem that can form a flower or a lateral internode (Huber and During, 2000; Black et al., 2009). As each unit acquires the resources for its development mainly by its own (Bittebiere and Mony, 2015), it is the target level in our study. We chose six white clover populations originating from an ongoing breeding programme to improve mixture performance [Deutsche Saatveredelung AG (DSV), Asendorf, Germany] and two white clover varieties (‘Euromic’, ‘Bombus’) – hereafter also referred to as populations for reasons of simplicity – to cover a range of different genetic backgrounds. On a scale of 1 to 9 that is used to characterize white clover varieties registered for agronomic use in Germany (Bundessortenamt, 2020), leaf size of the populations range between 3 and 9 and phenology between 3 and 7. Further characteristics of the populations are provided in Supplementary Data Table S1.

Study site

The study was conducted on an experimental station of the University of Göttingen, Lower Saxony, Germany (51°29′N, 9°55′E, 160 m a.s.l.). The climate of the region is temperate with four distinct seasons. Long-time averages for annual air temperature and precipitation were 9.2 °C and 651 mm [1981– 2010, German Weather Service (DWD)]. Monthly means of air temperature and irradiance and monthly sums of precipitation during the experiment are given in Supplementary Data Table S2. The soil, classified as a Gleyic Fluvisol, had a ploughing layer of 0.3 m depth that contained 21 % clay, 68 % silt and 11 % sand. Extractable soil nutrient concentrations were 64 mg kg⁻¹ phosphorus, 139 mg kg⁻¹ potassium and 210 mg kg⁻¹ magnesium. Soil pH was 6.8. The extractions of phosphorus and potassium were done with calcium acetate lactate and of magnesium with CaCl₂. Soil pH was measured in a 0.1 m CaCl₂-solution.

Experimental design

Each of the eight white clover populations was grown in pure stands and in two-species mixed stands with perennial ryegrass (‘ELP 060687’) or chicory (‘Puna II’). Sowing took place in September 2017 with a seed rate of 1000 seeds m⁻² and a white clover seed proportion of 40 % in mixed stands. The stands were arranged randomly and replicated in three blocks. Plots measured 4.5 m × 5.0 m.

Data collection

Measurements were carried out over four consecutive regrowth periods from April to October 2019. Due to poor spring development of white clover in mixed stands with perennial ryegrass, sampling of these stands started in regrowth period two.

We conducted biomass harvests at the end of each regrowth period on 8 May, 27 June, 15 August and 8 October. For each harvest, whole-plot biomass was cut at a height of 5 cm and weighed using a combine harvester (Wintersteiger hd 1500, Wintersteiger AG, Ried im Innkreis, Austria). A representative subsample (250 g) was dried in a forced air oven at 60 °C for 48 h to determine the dry matter concentration of the harvested biomass, which was used to quantify total dry matter yield. Another representative subsample was sorted to species level. Species fractions were oven-dried analogously and weighed to estimate the clover percentage of the harvested biomass, which in turn was used to calculate clover dry matter yield.

In the week before each harvest, we sampled 15 randomly selected white clover plants per plot by collecting apex fragments with at least two mature modular units. Immediately after collection, we enclosed them in plastic bags and stored them at 5 °C before further processing. In the laboratory, we divided the first mature module of every fragment into leaf, petiole and internode section (i.e. the segment between the respective module and the preceding one counted from the apex). We scanned the sections of each plot and used the image analysis software programs WinFOLIA Pro 2016b (leaves) and WinRHIZO Reg 2016a (petioles, internodes) (Regent Instruments Inc., Québec, Canada) to measure mean leaf area, petiole length and internode length. Thereafter, sections were oven-dried at 60 °C for 48 h and weighed to determine their mean dry mass. Mean specific leaf area, specific petiole length and specific internode length were then computed as the ratio of mean area (specific leaf area) or length (specific petiole length, specific internode length) to mean dry mass of each section.

Data analysis

All statistical analyses were run in the R 3.5.1 software environment (R Core Team, 2018).

To assess the effect of genetic background (white clover population) and of environmental determinants (stand type, regrowth period) on each of the architectural traits as well as on clover and total dry matter yield, we applied linear mixed effects regression models using the package ‘nlme’ (Pinheiro et al., 2018). To compare all stand types across all regrowth periods despite missing data for the first regrowth of mixed stands with perennial ryegrass, we created a new combined factor (stand_regrowth). Full regression models thus contained population, stand_regrowth, their interaction and block as fixed effects and plot as a random effect. All full models were visually checked for normality and homogeneity of variance of residuals. To ensure these, we square-root transformed clover and total yield and fitted appropriate variance structures where necessary. We utilized the second-order Akaike information criterion corrected for small sample sizes (AICc) to choose each final model, i.e. the simplest model with an AICc-difference to the best model of <2. The significance of fixed effects remaining in the final models was determined with sequential Wald chi-square tests. Differences in effect levels were tested for significance using Tukey-based post-hoc tests as implemented in the ‘emmeans’ package (Lenth et al., 2019).

We further investigated the relationship between architectural traits and clover or total dry matter yield in two successive steps. To avoid multicollinearity and to derive a summarized response of the major trait pattern, we first used the package ‘vegan’ (Oksanen et al., 2019) to perform a principal component analysis (PCA) on the centred and standardized architectural traits. As a second step, we carried out linear mixed effects regressions of clover or total yield against stand_regrowth and its interaction with the scores obtained from the first and second PCA-axis, with block as a fixed and plot as a random effect (package ‘nlme’). Clover yield and total yield were square-root transformed and variance structures were applied where needed. Model examination and selection procedures as well as tests of significance for fixed effects remaining in the final models were done as described above. The significance of yield changes in response to variations in effect levels was determined using Monte Carlo-based post-hoc tests.

To estimate how much of the trait and yield variation can be explained by white clover genetic background and environmental determinants we followed the approach of Legendre (2008). By means of sequential redundancy analysis within the ‘vegan’ package, followed by subtractions, we partioned trait variations into sole and joint effects of white clover population, stand type and regrowth period. Yield variations were partioned via sequential linear regression analysis followed by subtractions.

RESULTS

Effect of genetic background and environmental determinants on traits and yield

The studied white clover populations differed in the size of the measured architectural traits except for internode length and specific leaf area (Table 1). Relative differences between the highest and lowest population means were largest for specific internode length (9.0 %), followed by leaf area (8.8 %), specific petiole length (7.3 %) and petiole length (5.6 %; Fig. 1). Stand type and regrowth period jointly affected the size of all architectural traits. However, the interaction effect between population and environmental determinants was not retained in any of the final models, i.e. the sign and degree of the joint effect of stand type and regrowth period on traits did not vary between populations (Table 1).

Table 1.

Effect of white clover population as well as joint effect of stand type and regrowth period on architectural traits and on clover and total dry matter yield.

Effect	Leaf area	Petiole length	Internode length	Specific leaf area	Specific petiole length	Specific internode length	Clover yield	Total yield
Stand_regrowth	134.4***	180.6***	58.0***	46.9***	85.3***	48.2***	122.0***	295.1***
Population	30.8***	20.0***	–	–	33.0***	34.4***	9.3***	6.2***

F-values and P-value levels of sequential Wald chi-square significance tests are given. Significance levels were as follows: * = P < 0.05, ** = P < 0.01, *** = P < 0.001. Dashes signify effects not retained in the final models.

Fig. 1.

Effect of white clover population as well as joint effect of stand type and regrowth period on (A–F) architectural traits and on (G) clover and (H) total dry matter yield. Values are means (±s.e.) across white clover populations (left part of each graph; n = 24) or across stand types and regrowth periods (right part of each graph; n = 33). Different letters indicate statistically significant differences between means (α = 0.05). T1–T8, different white clover populations; T, white clover pure stands; TL, mixed stands with perennial ryegrass; TC, mixed stands with chicory; R1–R4, regrowth one to four.

Leaves and petioles were generally largest in pure stands and smallest in mixed stands with perennial ryegrass, whereas long internodes in mixed stands with perennial ryegrass were restricted to regrowth two and three. Independent of stand type, we found the highest mean values of leaf area, petiole length and internode length in the second regrowth and the lowest in the fourth regrowth. For specific leaf area, specific petiole length and specific internode length, no consistent differences between stands existed. However, all stands had the highest mean values of specific leaf area in the second regrowth, while specific petiole length was largest in regrowth three or four, and specific internode length in regrowth two or three (Fig. 1).

Clover and total dry matter yield were dependent on both the effect of population and the joint effect of stand type and regrowth period (Table 1), but not on their interaction. Clover yield (9.3 % of mean value) varied stronger between populations than total yield (3.6 %). While white clover pure stands generally accounted for the largest mean clover yields, mixed stands with chicory had the largest mean total yields in all regrowths. Among regrowth periods, the second regrowth yielded the highest yield and the fourth the lowest yield (Fig. 1).

Relationship between traits and yield

The first two axes of the PCA performed on the six architectural traits explained 87 % of the trait variation (Fig. 2). The first of the two axes (PC 1) was negatively correlated with all traits linked to plant part sizes, i.e. leaf area (r = −0.91), petiole length (r = −0.96) and internode length (r = −0.78) as well as to specific leaf area (r = −0.71), whereas the second axis (PC 2) mainly captured the variation of traits quantifying plant part biomass investment amounts, i.e. specific leaf area (r = −0.60), specific petiole length (r = −0.87) and specific internode length (r = −0.93).

Fig. 2.

Loading plot of principal component analysis (PCA) on architectural traits. Arrow lengths indicate importance of architectural traits for PCA-axes (PC). The percentage of trait variation explained by each PC is given in parentheses.

For the relationship with clover or total dry matter yield, solely the interaction effect between environmental determinants and PC 1 scores was retained in the final models. Accordingly, clover and total yield were completely unrelated to PC 2 scores, whereas the direction and magnitude of their relationship with PC 1 scores differed between stand types and regrowth periods (Fig. 3). The relationship between yield and PC 1 was predominantly negative, and that with architectural traits linked to size and with specific leaf area was thus predominantly positive. Usually, no such significant relationships were found for the second regrowth of all stands. Neither was this the case for the first (clover yield) or the third and fourth regrowth (total yield) of mixed stands with chicory (Fig. 3; Supplementary Data Table S3). Thus, the weakest yield changes in response to variations in PC 1 scores were usually detected in regrowth two. While total yields reacted most strongly to PC 1 in regrowth one, no consistent seasonal pattern existed for clover yields. In addition, the reaction of clover yield was also not consistent across stands. By contrast, total yields changed always least in response to variations in PC 1 scores in mixed stands with chicory (Fig. 3; Table S3).

Fig. 3.

Relationship between architectural trait pattern (first principal component axis) and (A) clover or (B) total dry matter yield for different stand types and regrowth periods across white clover populations (n = 24). Lines are model estimates (with 95 % confidence bands). Marginal pseudo-R², F-values and P-value levels of sequential Wald chi-square significance tests are given. Significance levels were as follows: * = P < 0.05, ** = P < 0.01, *** = P < 0.001. T, white clover pure stands; TL, mixed stands with perennial ryegrass; TC, mixed stands with chicory; R1–R4, regrowth one to four.

Proportion of trait and yield variation explained by genetic background and environmental determinants

Most of the variation in architectural traits as well as in clover and total dry matter yield, namely 70 %, 72 % and 86 %, was caused by the experimental factors, though partitioned unevenly among them (Table 2). About half of the trait and clover yield variation and two-thirds of the total yield variation were explained solely by regrowth period and an additional 2–6 % jointly with stand type. The sole stand type effect accounted for much less of trait and yield variation, namely 5 % and < 15 %. By contrast, the population effect explained a larger percentage of trait variation (13 %) than of yield variation (<5 %).

Table 2.

Percentage of variation (%) of architectural traits as well as clover and total dry matter yield explained solely or jointly (\) by the white clover population, stand type and regrowth period effect.

Effect	Traits	Clover yield	Total yield
Stand	5.1	14.8	13.0
Regrowth	50.5	46.5	66.6
Population	13.0	4.5	1.6
Stand\regrowth	1.6	6.3	4.7
Stand\population	0	0	0
Regrowth\population	0	0	0
Explained	70.2	72.1	85.9

Discussion

While genetic and environmental drivers of white clover architecture and productivity have been widely studied, their relative importance remains unclear, and the relationship between white clover architecture and productivity is still not fully understood. In our 1-year study, the genetic background (white clover population) and the environmental determinants neighbourhood (stand type) and season (regrowth period) explained a large percentage of the variation in architecture and yield, confirming that we investigated key drivers, though they were not equally important.

Effect of genetic background on traits and yield

Chosen with the prior expectation of being distinguishable in architecture, the studied populations differed in four of the six measured architectural traits. However, while population was a greater source of trait variation than neighbourhood, it was notably less relevant for architecture than season. Even though these results are generally supported by earlier investigations that emphasized the importance of the environment for architecture, the proportion of the genetically caused variation was surprisingly low. For example, Finne et al. (2000a) and Huber et al. (2016) found the genetic variances to be in the same order or to even exceed that of the environmental variances for various architectural traits.

Beyond that, the architectural differences between populations remained stable in all neighbourhoods and seasons, indicating that the selected populations did not respond differently to altered environmental conditions (De Kroon et al., 1994). This contrasts with most of the findings for architecturally distinct white clover populations subjected to contrasting environmental conditions (Caradus and Chapman, 1991; Finne et al., 2000a; Annicchiarico, 2003), but agrees with the results of a few other studies (Weijschedé et al., 2008; Bittebiere et al., 2012; Huber et al., 2016).

Differences in clover and total yield between white clover populations were also present but of lesser magnitude and they only accounted for a very small percentage of yield variation as compared to neighbourhood and season. In addition, those differences remained unchanged by neighbourhood and season. Similar results were achieved by several researchers (Elgersma and Schlepers, 1997; Nassiri and Elgersma, 2002; Bittebiere et al., 2012; Huber et al., 2016), and might be linked to the environmental stability of architectural differences between populations.

Effect of neighbourhood on traits

The size of the measured architectural traits differed between intra- and interspecific as well as between both interspecific neighbourhoods. These results are in line with previous studies. First, those studies found grasses to affect the architecture of several white clover genotypes in a different way compared with conspecifics (Weijschedé et al., 2008; Bittebiere et al., 2012). Second, they revealed that white clover architecture further depends on the grass species or variety with which it was grown (Turkington et al., 1991; Barthram, 1997; Bitterbiere et al., 2012). Those neighbourhood-specific plastic changes of architecture were assumed to follow from divergent growth strategies depending on the height and density of neighbours (Novoplansky, 2009; Bittebiere et al., 2012; Gruntman et al., 2017). In this regard, the three potential neighbours in our study were notably different: chicory was the tallest and white clover was the densest among neighbours (visually estimated). However, our experimental design implied that the contacts of white clover in its neighbourhood could have also been intraspecific (in mixed stands) or even intra-individual.

Nonetheless, as observed for tall and sparse neighbours by Gruntman et al. (2017) and supported by the findings of Thompson (1993), internodes of white clover were usually longest when grown in mixed stands with perennial ryegrass, thereby enabling its modules to encounter microenvironments with more preferable conditions and in doing so acting as an avoidance mechanism (Novoplansky, 2009). In contrast, as observed for dense and especially tall neighbours, white clover grown in mixed stands with chicory displayed longer petioles than white clover grown in mixed stands with perennial ryegrass. This might be interpreted as an early response to crowding, allowing its modules to position the leaves in higher zones with more favourable light conditions (Thompson, 1993; Gruntman et al., 2017). In addition, white clover also tended to have larger leaf areas and specific leaf areas in those stands, both of which are regarded as mechanisms to ensure a higher light interception under shaded conditions and thus promoting tolerance in situations where neighbours are too tall to be outgrown (Novoplansky, 2009; Gruntman et al., 2017). White clover grown in pure stands appeared to follow a similar strategy to white clover grown in mixed stands with chicory (tolerance strategy), although the shorter neighbours seemed to have resulted in further petiole elongation. This is in agreement with the theoretical considerations that plants are less likely to aggressively confront and outgrow conspecifics (Tremmel and Bazzas, 1993; Bennett et al., 2016; Bittebiere et al., 2019).

The differences in leaf area and petiole length between neighbourhoods were accompanied by widely inversed differences in specific petiole length. These in turn might reflect higher investments in support structures that were biomechanically necessary to avoid petiole buckling of larger leaves and petioles (Huber et al., 1998; Leeflang et al., 1998). Additionally, a lower positioned and smaller leaf area might have resulted in a lower carbohydrate gain, which might have limited the construction of robust petioles (Weijschedé et al., 2006). This might have also applied to internodes and was possibly strengthened by the large internode length in mixed stands with perennial ryegrass. Further, due to the function of internodes as hydraulic rather than as biomechanical support (Huber, 1996), the differences in specific internode length between neighbourhoods might have partly arisen from varying hydraulic requirements of different-sized petioles and leaves.

Although we could attribute many of the architectural differences between neighbourhoods to neighbour-specific strategies to improve resource capture or their consequences, we found discrepancies in some seasons.

Effect of season on traits

Instead of being temporally static, the size of architectural traits was highly dynamic and the impact of season on architecture notably exceeded that of neighbourhood. Previous results for white clover grown with neighbours over several seasons generally agree with these findings (Barthram and Grant, 1994; Marriott et al., 1997; Helgadottir et al., 2001; Nassiri and Elgersma, 2002), but the seasonal effect we observed was still surprisingly high. One possible reason is that our experiment took place about 1.5–2 years after sowing and, hence, probably in the phase where white clover shifts from a transition period with both tap-root and clonal growth to exclusive clonal growth (Brock et al., 2000). This shift can be accompanied by changes in architecture (Brock and Tilbrook, 2000; Bonesmo and Bakken, 2005).

In earlier investigations, architectural differences between seasons were normally attributed to changes in weather conditions, primarily in light regime and temperature (Brougham, 1962; Boller and Nösberger, 1983; Sackville Hamilton and Harper, 1990). Less often, they were also explained by agricultural grazing or cutting management if applied (Davidson et al., 1990; Barthram and Grant, 1994). It is widely accepted that plants under grazing or cutting can adapt their architecture to avoid or withstand a later removal of the same extent, for example by reducing leaf area, petiole length and internode length and by increasing specific leaf area (Culvenor et al., 1989; Davidson et al., 1990; Barthram and Grant, 1994; Weijschedé et al., 2008).

In our experiment, the seasonal differences in the size of those architectural traits that we assigned to neighbour-specific strategies to improve resource capture, i.e. leaf area, petiole length, internode length and specific leaf area, appeared to be primarily driven by changes in light regime and temperature. Like light intensity and temperature, these architectural traits initially increased in size and peaked in the second regrowth followed by declines until the last regrowth. By contrast, the seasonal differences in specific petiole length and specific internode length were more probably the consequence of the losses of photosynthetic surface and of resources in the previous harvest. Increments in the size of both architectural traits until the third regrowth might reflect their cheap reconstruction for rapid light foraging to compensate for the preceding losses or for reducing the losses in future harvests (Culvenor et al., 1989; Huber, 1996; Díaz et al., 2016; Wang et al., 2018). The increasing specific internode length could further be attributed to mobilization of carbohydrates from internodes, as rapidly regrowing petioles and leaves are strong carbohydrate sinks (Culvenor et al., 1989; Leeflang et al., 1998). Declines of specific petiole length and specific internode length in the last regrowth might mirror smaller losses in the former harvest and replenished internode carbohydrate stores to survive over winter (Boller and Nösberger, 1983; Turner and Pollok, 1998). Diverging results for the second regrowth of mixed stands with perennial ryegrass could have been caused by their large internode lengths leaving even less resources for robust constructions.

Apart from architectural differences between seasons, we found seasonal differences in architectural changes in response to neighbourhood. Such differences were also detected by Marriott et al. (1997) and Nassiri and Elgersma (2002). These might have been the result of an asymmetry in phenological development between white clover and its neighbours, temporally unequal biomass amounts that were lost by cutting white clover and its neighbours, and temporally varying regrowth rates (Woledge et al., 1992; Barthram and Grant, 1994; Briske and Richards, 1995).

Effect of neighbourhood and season on yield and its relationship with traits

In line with earlier studies, clover and total dry matter yield depended on neighbourhood and season, with the effect of season accounting for a substantially higher percentage of yield variation than that of neighbourhood (Hill and Michaelson-Yeates, 1987; Evans et al., 1989; Elgersma et al., 1998, Gilliland et al., 2009).

Using suitable varieties, white clover mixed stands are regularly found to be higher yielding than the corresponding pure stands (Annicchiarico and Piano, 1994; Nassiri and Elgersma, 2002; Nyfeler et al., 2009). Surprisingly, although we utilized white clover populations bred to perform in mixtures, only mixed stands with chicory outperformed white clover pure stands. This was due to higher yields of white clover and of the mixture partner in mixed stands with chicory as compared to mixed stands with perennial ryegrass. Both might be explained by mean size and spatial distribution of the mixture partner. While chicory individuals were taller and denser than perennial ryegrass individuals, resulting in higher mixture partner yields, their spatial distance was greater, thereby leaving larger patches for white clover and causing higher clover yields. In agreement, Hutchings and Wijesinghe (2008) reported that clonal species were higher yielding in heterogeneous environments with large-scale patches than in heterogeneous environments with small-scale patches or in homogeneous environments.

Yield turned out to be related to the architectural traits leaf area, petiole length, internode length and specific leaf area. Those architectural traits are not only linked to neighbour-specific strategies to enhance resource capture, but they also directly influence height and density of biomass above the cutting height (i.e. harvestable biomass amount). Accordingly, their relationship with yield is in line with previous findings for clover (Annicchiarico et al., 1999; Finne et al., 2000b; Annicchiarico, 2003; Hoekstra et al., 2018). However, none of those studies included as many traits and performed such a comprehensive analysis across different neighbourhoods and seasons. Beyond that, in our study, the relationship also existed for total yield, though it was of lesser magnitude than for clover yield.

While being overall positive in its direction, the magnitude of the relationship of architectural traits with clover or total dry matter yield clearly differed between neighbourhoods and seasons, as proposed for leaf size by Mackay (1991). Interestingly, the regrowth period with the largest trait sizes and total yield amounts (second regrowth) coincided with that with the weakest relationship between traits and yield. In addition, the stand type with the highest total yields (mixed stands with chicory) was also that with the smallest yield changes in response to variations in trait sizes. Thus, when the white clover architecture ensured a high amount of harvestable clover biomass, further increments in the size of architectural traits did not notably improve yield. Considerable increases in yield were also absent when the harvestable biomass of the mixture partner was high.

Conclusions

Our results provide evidence that the architecture of white clover is to a large extent shaped by its genetic background, neighbourhood and season, with season being by far the most important driver. The seasonal pattern differed between architectural traits and was further modified by the identity of neighbours, indicating different underlying causes for each trait.

Despite their variability in architecture, the white clover populations did not differ in their architectural changes in response to neighbourhood and season. This suggests that the same plastic responses occur across a wide range of genetic backgrounds.

Moreover, our results show that the biomass yield of clover communities is positively related to a trait syndrome consisting of large values for leaf area, petiole length, internode length and specific leaf area. Large values for these traits might thus be targeted in future breeding efforts and selection of white clover varieties. Yet, the impact of these traits on biomass yield varied between seasons and additionally depended on the identity of the neighbourhood. Increasing the resilience of white clover architecture against seasonal influences could thus help to improve its productivity. Based on our results, several above-ground architectural traits should be taken into account.

Further investigations should address the consistency of seasonal differences in white clover architecture over multiple years.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Table S1: Characteristics of utilized white clover populations. Table S2: Weather conditions during the experiment. Table S3: Model estimates and significance of the relationship between architectural trait pattern and clover and total dry matter yield.

Funding

This work was supported by the German Federal Ministry of Education and Research as part of the IMPAC³-project (grant number FKZ031A351A).