Biodiversity regulates ecosystem functions such as productivity, and experimental studies of species mixtures have revealed selection and complementarity effects driving these responses. However, the impacts of intraspecific genotypic diversity in these studies are unknown, despite it forming a substantial part of the biodiversity.
In a glasshouse experiment we constructed plant communities with different levels of barley (Hordeum vulgare) genotype and weed species diversity and assessed their relative biodiversity effects through additive partitioning into selection and complementarity effects.
Barley genotype diversity had weak positive effects on aboveground biomass through complementarity effects, whereas weed species diversity increased biomass predominantly through selection effects. When combined, increasing genotype diversity of barley tended to dilute the selection effect of weeds.
We interpret these different effects of barley genotype and weed species diversity as the consequence of small vs large trait variation associated with intraspecific barley diversity and interspecific weed diversity, respectively. The different effects of intra- vs interspecific diversity highlight the underestimated and overlooked role of genetic diversity for ecosystem functioning.
Understanding biodiversity's role in regulating key ecosystem functions such as productivity is a major challenge for ecology. Diverse communities can utilise resources effectively and extensively, and can be more productive, resulting in a positive relationship between producer (e.g. vascular plant) diversity and productivity (Tilman et al., 1997; Hector et al., 1999; Hooper et al., 2005; Cardinale et al., 2011). Positive biodiversity–productivity relationships are thought to be generated by two mechanisms: sampling or selection effects and complementarity effects (Loreau & Hector, 2001). Selection effects involve an increased probability of more diverse species mixtures harbouring high-yielding species that increase productivity and simultaneously dominate the mixture (Aarssen, 1997; Huston, 1997; Tilman et al., 1997). The productive species’ dominance will lead to over-yielding, but never to transgressive over-yielding: the mixture will perform better than the average monoculture, but will not outperform the best monoculture. By contrast, complementarity effects from niche differentiation or facilitation lead to enhanced performance of species in mixtures and increased net resource use (Trenbath, 1974), and potentially transgressive over-yielding (Loreau, 2004).
As well as interspecific diversity, intraspecific diversity may be important for the regulation of productivity (Reusch et al., 2005; Crutsinger et al., 2006; Hughes et al., 2008; Fridley & Grime, 2010; Kotowska et al., 2010; Cardinale et al., 2011; Cook-Patton et al., 2011; Crawford & Rudgers, 2012). Empirical and theoretical evidence also indicate the regulatory effects of intraspecific genetic diversity on species diversity (Booth & Grime, 2003; Vellend & Gerber, 2005; Whitlock et al., 2007; Fridley & Grime, 2010; Gibson et al., 2012): genetic diversity in populations has direct consequences for community species diversity through competition or niche differentiation. Consequently, besides direct effects of intraspecific diversity on productivity, intraspecific diversity could also indirectly affect the diversity–productivity relationship by influencing species diversity.
Key policy documents frequently promote the conservation of genetic diversity (United Nations Convention on Biological Diversity, 2010), but we have limited understanding of the relative role of genetic diversity in diversity–function relationships (Hughes et al., 2008). We are not aware of any study that quantifies selection vs complementarity effects in experiments that combine both genotype and species mixtures, and that compares the relative effects of intra- and interspecific diversity on productivity. This is likely due in part to the difficulty of regulating and keeping track of intraspecific diversity in mixtures. However, annual crop systems provide an excellent resource for exploring the role of genetic diversity. Although intensive agricultural crops are normally genetically uniform (i.e. comprising a single cultivar), genotype mixtures of multiple cultivars have been receiving more attention (Newton et al., 2009). Indeed, the range of available elite cultivars provides varied and genetically characterised material from which to construct mixtures with controlled levels of genotypic variation.
Diversity impacts on ecosystem functions may not be driven by species or genotype number per se. Recent interest has focussed on functional diversity and its impact on ecosystem processes such as nutrient turnover and system productivity (Hillebrand & Matthiessen, 2009; Newton et al., 2009). Critically, functional diversity may mediate the effects of intra- and interspecific diversity on ecosystem functions (Díaz et al., 2007), including productivity (Cadotte et al., 2009; Flynn et al., 2011; Roscher et al., 2012, 2013). Functional diversity can regulate resource exploitation and niche complementarity, and so the impacts of genotypic or species diversity might be determined by their relative influence on functional traits. Species diversity could contribute more than intraspecific genotype diversity to variation in functional traits, and therefore ecosystem functions (Auger & Shipley, 2013).
We hypothesised that increasing genotype or species diversity would lead to positive net biodiversity effects on productivity, but that the effect of species richness would be stronger than the effect of intraspecific genotype richness, in line with the expected higher trait variation associated with species diversity compared with genotype diversity. Besides direct effects of genotype and species diversity on productivity, we hypothesised that genotype richness would modulate the effects of species richness on community productivity.
In order to test these hypotheses, we constructed communities of annual plants and manipulated both intraspecific genotype (barley cultivars) diversity and (weed) species diversity. To assess whether net biodiversity effects resulted from selection or complementarity effects, and how these were influenced by intra- vs interspecific diversity, we used additive partitioning sensu Loreau & Hector (2001).
Materials and Methods
We constructed mesocosm communities in a glasshouse at the James Hutton Institute, Aberdeen (UK). The communities had varying weed species and barley (Hordeum vulgare L.) cultivar diversity and were sown in wooden, soil-filled boxes (0.7 × 0.7 × 0.3 m; surface area 0.49 m2), lined with plastic and filled with homogenised loam topsoil over c. 5 cm of c. 20 mm granite gravel. Mesocosms were kept under natural daylight and received no fertiliser, being watered uniformly with 1–2 l water per mesocosm 2–3 times per wk. Watering volume and frequency increased as the plants grew, and depended also on the weather. Air temperature and relative humidity in the glasshouse varied temporally between 5 and 25°C and beween 40 and 100%, respectively.
All plants were planted between 6 and 14 June 2012, except for Oxbridge barley (19 June 2012) and Euphorbia helioscopia (18 July 2012) which had delayed planting dates due to low seed viability. Before planting, all seeds were soaked overnight in 1 μM gibberellic acid (GA; Sigma-Aldrich); after GA treatment, seeds were spread onto a layer of commercial seedling compost and maintained at 5°C in the dark for 2 d to synchronise germination. Germinated weeds were transferred to the glasshouse under ambient temperature and daylight conditions for three additional weeks. Barley seedlings were grown at 12 h 20°C : 12 h 18°C, day : night regime for only 1 wk before transplanting. To ensure a standardised arrangement of plants within mesocosms, seedlings were planted in a grid pattern defined by a square wooden frame, with stretched strings dividing the mesocosm surface into 64 equal-sized squares whose centres marked the planting points. For each mesocosm, 60 seedlings were transplanted individually into the square units, with equal proportions of each barley cultivar or weed species in mixtures (detailed below). This ensured a planting density similar to barley crop fields. Plants of each cultivar or species were randomly allocated to grid squares, leaving four empty squares per mesocosm. In mesocosms with all five barley cultivars, different cultivars were identified using colour-coded wire rings secured loosely around the stem base of each barley plant.
Mesocosms were monitored twice weekly for plant survival, health and development; here we report final biomass data. To prevent lodging, the tops of all barley plants were brushed gently with a wooden stick five times a week for 10 wk after planting (a technique commonly used when growing cereals in glasshouse conditions). Biomass of volunteer weeds was extremely low (final harvest mean ± SE = 0.15 ± 0.02 g dry mass mesocosm) and is unlikely to have influenced the much greater biomass (on average 584 times larger) of the barley and experimental weed species considered here. To limit heavy infestations of powdery mildew and aphids, plants were treated with a fungicide (1 ml l−1; ALTO 100 SL, Syngenta Crop Protection Inc., Greensboro, NC, USA) and soap solution (10 ml l−1; SAVONA, Koppert B.V., AD Berkel en Rodenrijs, the Netherlands) at 30 and 80 d after planting (dap) to ensure consistent survival of plants across mesocosms and treatments. All plants in each mesocosm were harvested between 100 and 104 dap, except Oxbridge and E. helioscopia (harvested at 91–95 and 62–66 dap, respectively). For each mesocosm, aboveground plant material was removed at the stem base, and separated into species and barley cultivars. Dry mass of each species and cultivar was recorded after at least 72 h in an 80°C oven.
Diversity treatments
Intraspecific genotype diversity was manipulated using combinations of five barley cultivars (Optic, Oxbridge, Sebastian, Tipple and Westminster, hereafter called ‘genotypes’) showing a genetic similarity of 61–68% based on single nucleotide DNA polymorphisms (see Supporting Information Notes S1) and species diversity with combinations of five annual weeds common in barley crops (Capsella bursa-pastoris (L.) Medik. (Brassicaceae), Euphorbia helioscopia L. (Euphorbiaceae), Poa annua L. (Poaceae), Senecio vulgaris L. (Asteraceae) and Stellaria media (L.) Vill. (Caryophyllaceae)). It is important to note that the weed seeds – obtained from a commercial supplier (Herbiseed, Twyford, UK) – included an unknown number of genotypes. In this experiment we compared, therefore, intraspecific genotype diversity (of barley) with species diversity (of weeds) where the latter also includes an unknown element of intraspecific genotype diversity (as is the case in similar species diversity experiments). Consequently, we interpret the differences in the effects of varying barley and weed diversity as predominantly a consequence of differences in interspecific diversity, whereas responses in the barley mixture treatments alone represent intraspecific diversity effects.
Barley genotypes were chosen to include contrasting weed resistance abilities, based on experimental field plot data (A. C. Newton & S. Hoad, pers. comm.). To manipulate species diversity we followed a conservative approach by not including N-fixing weeds because their impact can mask diversity effects (Cadotte et al., 2009). Genotypes and species were grown in monoculture (one barley (B) or one weed (W) – 10 treatments; all 60 plants in each mesocosm being the same genotype or weed species), and mixtures of either all five genotypes (5B – one treatment, 12 plants per barley genotype), all five weed species (5W – one treatment, 12 plants per weed species), one barley genotype and five weed species (1B5W – five treatments, 10 plants per genotype or weed species), and five barley genotypes and five weed species (5B5W – one treatment, six plants per genotype or weed species). Therefore, barley genotype and weed species richness both varied between zero, one and five (see Table S1). Four replicates of each treatment and 18 different species × genotype combinations resulted in 72 mesocosms in total. Importantly, due to space constraints, this design is not fully-factorial: no mixtures contained only a single weed species (i.e. there are no 1B1W and 5B1W combinations). Hereafter we refer to combinations with more than one barley genotype and/or weed species as mixtures, and those with one barley genotype and no weeds, or one weed species and no barley, as monocultures.
Trait measurements
In order to assess trait variation of the mixture components (i.e. the barley cultivars and weed species), we randomly selected five individuals of each weed species or barley cultivar within each of the four 5B5W mesocosms and measured plant height, plant width, biomass, leaf area, specific leaf area (SLA), leaf dry matter content (LDMC) and relative water content (RWC), whereas leaf carbon and leaf nitrogen concentrations were determined for all individuals of the same cultivar or species and mesocosm pooled. These traits were expected to reflect various aspects of the niche of each cultivar and species (sensu Violle & Jiang, 2009), such as growth rate, competitive ability and resource use (light, water and nutrients). Trait measurements except biomass were performed 56–63 dap for all cultivars and species, except for Oxbridge (51–58 dap) and E. helioscopia (21–28 dap). Biomass was determined at final harvest between 100 and 104 dap, except Oxbridge and E. helioscopia (harvested at 91–95 and 62–66 dap, respectively).
For most traits we followed the protocol of Cornelissen et al. (2003) for trait measurements. Plant height (cm) was measured as the vertical distance between the uppermost tip of photosynthetic tissue and the ground level. Plant width (cm) was measured as the horizontal distance between the two outermost tips of photosynthetic tissue. For plant biomass (g) aboveground plant material of each individual was removed at the stem base, and dry mass recorded after at least 72 h in an 80°C oven. For leaf area (mm2), SLA, LDMC and RWC we carefully removed a fully developed and healthy leaf of each plant individual, determined fresh mass at field turgor, fully rehydrated the leaf overnight and measured fresh mass at full turgor and leaf area. After drying the leaf at 80°C for 48 h we determined dry mass. SLA was then calculated as the ratio between leaf area of the fresh leaf and leaf dry mass (mm2 mg−1), LDMC as the ratio between leaf dry mass and fully rehydrated fresh mass (mg g−1), and RWC as the ratio between leaf water content in the glasshouse and water content after rehydration at full turgor (%). Leaf water content in the glasshouse was calculated as the difference between field fresh mass and dry mass, and water content at full turgor was calculated as the difference between saturated mass and dry mass. For leaf C and N concentrations (% of dry mass) the dried leaves of each cultivar or species of each mesocosm were pooled, ball milled to fine powder and total C and N contents of a 2 mg subsample were determined using an Exeter Analytical CE440 Elemental Analyzer (EAI, Coventry, UK). For further analysis we used mean trait values per cultivar or species and mesocosm.
Additive partitioning of biodiversity effects
Following Loreau & Hector (2001), the net biodiversity effect was determined as the observed absolute performance (i.e. aboveground biomass (g) per mesocosm) of mixtures compared with their expected performance based on the relevant monocultures. Net biodiversity effects are the sum of two groups of underlying effects with different causes – the selection effect and the complementarity effect.
Selection effects are quantified as the covariance between the biomass of a species or genotype in monoculture and the change in relative yield of each species or genotype grown in mixture (see Notes S2). Positive selection effects indicate dominance in the mixture by species or genotypes that perform best in monoculture. However, they do not lead to increased net performance of the mixture above that expected from the best performing monoculture. Negative selection effects occur if species that perform poorly in monoculture dominate a mixture (Loreau & Hector, 2001).
Complementarity effects are indicated by changes in the performance of a species or genotype in mixture that are not compensated by similar-scale but opposing changes in other mixture components. They quantify the change in mixture yield relative to that expected from the monocultures (see Notes S2). As with selection effects, complementarity effects can be positive or negative (Loreau & Hector, 2001). Increasing performance of one component in the mixture without a balancing decrease in performance in another component results in a positive complementarity effect, and enhanced net performance of the mixture based on expectations from weighted monoculture means (Turnbull et al., 2013).
Statistical analyses
A linear mixed-effects model was applied to test for differences in aboveground biomass per mesocosm between monocultures and mixtures, including the treatments as a random variable.
For each mixture significant negative or positive selection, complementarity and net biodiversity effects (i.e. significant deviations of these scores away from zero) were assessed using two-tailed Student's one-sample t-tests. Comparisons of biodiversity effects between 5B, 5W and 5B5W mixtures were conducted using two-tailed Welch two-sample t-tests.
Linear mixed-effects models were applied to test relationships between genotype and species richness and selection, complementarity and net biodiversity effects on aboveground mixture biomass. Barley genotype composition was included as a random effect consisting of seven levels (none, Optic, Oxbridge, Sebastian, Tipple, Westminster, mixture). Fixed effects of ‘genotype richness’ and ‘species richness’ were assessed using type-II analyses of variance and Wald χ2-test statistics. Fixed effect importance was calculated as the product of the slope (i.e. factor coefficient) and maximum range of the factor covered by the experiment (Connolly et al., 2011). The lack of a fully factorial design prevented exploration of interactive effects of genotype and species richness. The random effect ‘genotype composition’ was assessed using likelihood ratio tests comparing the models with and without this random effect (Hector et al., 2011). Weed species composition effects could not be assessed as either none or five weed species were present in mixtures.
We performed a principal component analysis of the nine traits monitored for the five barley cultivars and the five weed species grown in the 5B5W mixture (Table S2). Before analysis trait values were scaled to unit variance and zero centred because traits were in different units and had different variances. We also applied a Varimax rotation to reduce the number and multi-collinearity of variables.
All statistical analyses were performed with R v2.15.3 (R Core Team, 2013) and nlme (Pinheiro et al., 2013).
Results
Diversity effects on aboveground biomass
Overall, mixtures (110.3 ± 5.4 g mesocosm) had 41% higher average biomass than monocultures (78.2 ± 8.2 g mesocosm; F(1,70) = 9.28, P =0.003). However, no mixture produced as much biomass as the most productive monoculture Stellaria (Fig. 1).
Mean (± 1 SE) aboveground biomass per mesocosm for monocultures (diamonds) and mixtures (circles). Monocultures were grown for each of five barley (Hordeum vulgare) cultivars (open diamonds) and five weed species (closed diamonds), whereas mixtures either comprised all five barley cultivars (5B), all five weed species (5W), one barley cultivar (Op, Optic; Ox, Oxbridge; Se, Sebastian; Ti, Tipple; We, Westminster) and five weed species (+ 5W), or five barley cultivars and five weed species (5B5W). Dashed line, average aboveground biomass per mesocosm for monocultures; solid line, average aboveground biomass per mesocosm for mixtures. For each treatment n =4.
Partitioning of biodiversity effects
Barley genotype diversity was associated with a marginally significant net biodiversity effect (Fig. 2), with a 7.9% increase in biomass in the 5B mixture compared with that predicted from monocultures. Although neither selection nor complementarity effects were significant, this tendency towards a positive net biodiversity effect appears to be driven by weak positive complementarity effects (Fig. 2).
Net biodiversity, selection and complementarity effects for the mixtures of barley (Hordeum vulgare) genotypes and weed species. B, barley genotype (Op, Optic; Ox, Oxbridge; Se, Sebastian; Ti, Tipple; We, Westminster); W, weed species. Mixtures were either five barley genotypes (5B), five weed species (5W), one barley genotype combined with five weed species (barley cultivar + 5W), and five barley genotypes combined with five weed species (5B5W). Values shown (g 0.5 m−2) are treatment means ± 1 SE. Differences of means from 0 were tested with Student's one sample t-test, with significance levels indicated: **, P <0.01; *, P <0.05; †, P <0.1; ns, P >0.1.
Weed species diversity was associated with a significant positive net biodiversity effect driven by a strong selection effect (Fig. 2), resulting in a 68% biomass increase in the 5W mixture compared with that expected from monocultures. This selection effect was caused by dominance of Stellaria (+ 184%) and Senecio (+ 23%) and declines in Euphorbia (−52%), Poa (−73%) and Capsella (−80%).
For 1B5W mixtures overall (i.e. pooling data from across mixtures containing different barley genotypes) we found a highly significant net biodiversity effect (t =5.94, df = 19, P <0.001). Again, this was driven by a strong selection effect (t =9.82, df = 19, P <0.001), with no significant complementarity effect (t =0.59, df = 19, P =0.56). Small differences occurred between 1B5W mixture types, indicating that weed species diversity effects depended on the barley genotype present (Fig. 2; Table S2). Nevertheless, in nearly all 1B5W mixtures there was a significant or marginally significant net biodiversity effect driven by a significant selection effect (Fig. 2). The exception to this was the mixture containing the genotype Optic, which showed only a marginally significant selection effect. Overall, the selection effect was driven by increased biomass of Stellaria (176 ± 19%) and barley (27 ± 7%) at the expense of Senecio (−3 ± 12%), Euphorbia (−42 ± 8%), Capsella (−73 ± 4%) and Poa (−75 ± 4%). Of the barley genotypes, highest relative mass gain in 1B5W mixtures compared with monocultures was observed for Sebastian (+ 54%), followed by Oxbridge (+ 28%), Tipple (+ 21%), Westminster (+ 19%) and Optic (+ 12%).
The 5B5W mixture showed a significant positive selection effect and a marginally significant positive net biodiversity effect (Fig. 2). The selection effect was due to dominance by Stellaria (+ 211%) and Sebastian (+ 20%) at the expense of Optic (−4%), Oxbridge (−17%), Tipple (−18%), Westminster (−27%), Senecio (−33%), Euphorbia (−73%), Poa (−77%) and Capsella (−86%). The only marginally significant net biodiversity effect may have resulted from the negative, but nonsignificant, complementarity effect counteracting the positive selection effect.
Effects of changes in barley genotype or weed species richness on biodiversity effects
Although net biodiversity effects and their components (i.e. selection and complementarity effects) for individual barley genotype mixture treatments were not or only weakly significant (as reported above; Fig. 2), combining data from across mixture treatments showed that net biodiversity effects in mixtures tended to decrease with increasing barley genotype richness (Fig. 3a; Table 1), with smallest net biodiversity effects in 5B mixtures driven by declining trends in both selection (Fig. 3b) and complementarity effects (Fig. 3c). Barley genotype diversity impacts on selection and complementarity effects were of similar importance (Table 1).
Table 1. The effects of barley (Hordeum vulgare) genotype and weed species richness (fixed effects) and barley genotype composition (i.e. the identity of the barley cultivar, random effect) on net biodiversity, selection and complementarity effects calculated from aboveground biomass
Effect variable
Model effects
df
χ2
P-value
Coef/logLikelihood
Importance
The table shows for each model variable the degrees of freedom (df), χ2-value, P-value, factor coefficient (Coef) or likelihood ratio test score (logLikelihood) for fixed and random effects, respectively, and estimate of importance for the fixed effects calculated as the division of the slope by the factor range (see the Materials and Methods section for details).
The partial effects of changing barley (Hordeum vulgare) genotype richness (a–c) and weed species richness (d–f) in crop–weed mixtures on the net biodiversity effect (a, d), selection effect (b, e) and complementarity effect (c, f). See Table 1 for the corresponding statistical results. The model fit (solid lines) and associated 95% confidence intervals (dashed lines) shows the change in the biodiversity effects (g 0.5 m−2) with changing genotype (green) and species richness (red). Symbols show the observed value of each mixture mesocosm. (a–c) Closed triangles and circles, mixture mesocosms with 0 and 5 weed species, respectively. (d–f) Open triangles, squares and circles, mixture mesocosms with 0, 1 and 5 barley cultivars, respectively. Symbols at number of genotypes or number of species equal to 5 were jittered to enhance visibility.
Net biodiversity effects did not change with changes in weed species richness from 0 to 5 (Fig. 3d; Table 1), but this hides underlying but counteracting changes in selection and complementarity effects. Selection effects increased with weed species richness (Fig. 3e), but complementarity effects tended to decrease (Fig. 3f; Table 1). Species diversity was more important for selection than complementarity effects. Compared with the effects of barley genotype diversity, weed species diversity was more important for the selection effect and of similar importance for the complementarity effect (Table 1).
The net biodiversity effect (t =4.03, df = 5.54, P =0.008) and the selection effect (t =8.87, df = 3.02, P =0.003), but not the complementarity effect (t =−0.80, df = 5.98, P =0.45), were both significantly greater for 5W compared with 5B mixtures, indicating stronger biodiversity effects by weed species than barley genotypes through stronger selection effects. The 5B5W mixture had significantly larger selection effects compared with the 5B mixture (t =3.28, df = 3.01, P =0.046), and marginally significantly smaller selection effects than the 5W mixture (t =−2.17, df = 5.21, P =0.080). Net biodiversity and complementarity effects did not differ significantly between 5B or 5W and 5B5W mixtures. This illustrates overall dilution by 5B mixtures of selection effects, for example when combined with five weed species in the 5B5W combination.
The random effect ‘genotype composition’ significantly influenced the response of the net biodiversity, selection, and complementarity effects to changes in species and genotype diversity (Table 1), indicating that these effects differed between barley genotypes (Fig. 2).
Trait variation in barley and weed mixtures
The trait values summarised in Table 2 indicate lower trait variation among the five barley genotypes compared with the five weed species. This was confirmed with the species scores plotted on the first to third principal components (Fig. 4), which showed substantially more trait variation among weed species than among barley cultivars. The first principal component (explaining 27% of the trait variation among species and cultivars) had high positive variable loadings of plant height, leaf area, LDMC and leaf C and high negative loadings of SLA and leaf N (Table S2). The second principal component (23%) had high positive loadings of leaf N and high negative loadings of plant width and biomass. The third principal component (17%) had a high positive variable loading of RWC.
Table 2. Trait values (mean ± 1 SE) of nine traits measured for the five barley (Hordeum vulgare) cultivars and five weed species in the 5B5W mixture mesocosms
Cultivar/species
Plant height (cm)
Plant width (cm)
Biomass (g)
Leaf area (mm2)
SLA (mm2 mg−1)
LDMC (mg g−1)
RWC (%)
C (%)
N (%)
SLA, specific leaf area; LDMC, leaf dry matter content; RWC, relative water content.
n =4 (except Poa and Euphorbia, where n =2 and n =3, respectively, due to mortality).
Principal component (PC) analysis of nine traits (plant height, plant width, biomass, leaf area, specific leaf area, leaf dry matter content, relative water content, leaf carbon and leaf nitrogen concentration) of the five weed species (red) and five barley (Hordeum vulgare) cultivars (green) in the four mesocosms consisting of the 5B5W mixture. Principal components 1, 2 and 3 explained 27%, 23% and 15% of the variance, respectively. Weed species: Cb, Capsella bursa-pastoris; Eh, Euphorbia helioscopia; Pa, Poa annua; Sv, Senecio vulgaris; Sm, Stellaria media. Barley cultivars: Opt, Optic; Oxb, Oxbridge; Seb, Sebastian; Tip, Tipple; Wes, Westminster.
Discussion
In our crop–weed system we observed over-yielding of mixtures but no transgressive over-yielding. A positive net biodiversity effect was observed for the weed species mixtures (5W), but this became weaker when five barley genotypes were included (5B5W). The higher net biodiversity effect for weed species mixtures compared with barley genotype mixtures, or combined barley and weed mixtures, was attributable to large selection effects in weed species mixtures. These were absent in exclusively intraspecific barley genotype mixtures (5B) and much reduced under high barley genotype richness in weed species mixtures (i.e. the 5B5W treatment). Net biodiversity effects in intraspecific barley genotype mixtures operated mainly through complementarity effects, which were generally weak and influenced to a similar extent also by weed species diversity. Overall, increases in both barley genotype and weed species diversity led to weak complementarity effects, whereas increased species diversity led to strong positive selection effects; intraspecific barley genotype diversity was crucial in modulating the direction and strength of the weed species mixtures’ selection effects.
Effects of intra- and interspecific diversity
Barley genotype (5B) mixtures showed relatively weak positive net biodiversity effects due to weak positive complementarity effects; no single genotype dominated the barley mixtures. Weak effects of intraspecific genotype diversity on productivity have been found previously in agricultural and natural vegetation systems (Booth & Grime, 2003; Newton & Guy, 2009; Fridley & Grime, 2010). In these studies, detection of net biodiversity effects was perhaps limited by the relatively small variability in performance encompassed by genotypic diversity. The same may have been true here: total aboveground biomass of the best-performing monoculture genotype, Westminster, was 1.5 times that of the worst performing, Oxbridge (Fig. 1). Studies finding substantial effects of intraspecific diversity used more variable genotypes, with monoculture performance varying between genotypes by a factor exceeding two (Reusch et al., 2005; Crutsinger et al., 2006; Kotowska et al., 2010).
The variability in potential performance encompassed by the weed species (5W) mixtures was much higher, with monoculture performance varying by a factor of 17 between Stellaria and Euphorbia. We found a strong positive net biodiversity effect: weed species mixtures produced 78% greater biomass than expected from monoculture yields. This scale of effect is within the average range of species richness effects on productivity found in other studies in natural systems (Cardinale et al., 2007). The positive biodiversity effect in weed species mixtures can be attributed almost exclusively to a selection effect; the most productive weed species Stellaria dominated the mixture, compensating for productivity declines in less productive species.
The proposal that lower trait variability is associated with intraspecific genotype diversity compared with species diversity is supported by examination of the trait space occupied by barley genotypes and weed species (Fig. 4). They illustrate the much larger trait space volume occupied by the weed species compared with the barley genotypes in 5B5W mixtures. This analysis demonstrates Stellaria's distinct combination of traits (Fig. 4), including rapid growth and large biomass (Table 2), making Stellaria an obvious potential driver of significant positive selection effects. It also demonstrates the considerable trait variation between the five weed species. We selected five phylogenetically distantly related species and this may have enhanced functional trait variability. Previous studies showed that phylogenetic distance and corresponding trait differences in species mixtures are very strong predictors of species diversity effects on productivity (Cadotte et al., 2009; Flynn et al., 2011). Our results (comparing the functional variability associated with either species or intraspecific genotype diversity) confirm these previous findings. Strong selection effects are likely in species mixtures where one species shows extreme trait values (Loreau, 2000), particularly in short-term studies of rapidly growing annual plants (Cardinale et al., 2007; Fargione et al., 2007; Jiang et al., 2009).
Interaction effects between intra- and interspecific diversity
Positive net biodiversity or selection effects occurred in all weed species mixtures irrespective of the presence and diversity of barley genotypes (i.e. Stellaria dominated in 5W, 1B5W and 5B5W mixtures). However, intraspecific diversity of barley influenced this in two ways.
First, the direction and strength of the net biodiversity, selection and complementarity effects were influenced by barley genotype identity in 1B5W mixtures: different barley genotypes created different dominance patterns (see Table S3). Productivity increases in Stellaria and barley (irrespective of genotype) in mixtures generally compensated for declines in productivity by other weed species, but relative mass gains and losses of all mixture components varied considerably. We interpret this as variation in interaction intensities depending on the barley genotype present. Genotype dependency of interspecific competition has been widely reported (Taylor & Aarssen, 1990; Lankau & Strauss, 2008; Fridley & Grime, 2010), particularly for spring barley (Hansen et al., 2008). Fridley et al. (2007) demonstrated that genotype-specific effects can alter dominance patterns and compensatory growth of co-existing species. Differences in barley genotype competitive ability might therefore drive variation in compensatory growth patterns among the 1B5W treatments, and the corresponding variation in selection and net biodiversity effects.
Second, although increasing barley genotypic diversity did not in itself lead to strong biodiversity effects on community productivity, overall increases in barley genotypic richness tended to dilute selection effects. As barley genotypic diversity increased from 0 to 1 to 5, selection effects – driven mainly by weed species diversity – tended to decline (Fig. 3b). Previous studies have indicated that intraspecific genotype mixtures can alter biomass compensation patterns of co-occurring species. Crawford & Rudgers (2012) showed that when genetic diversity of the dominant Ammophila breviligulata in a dune system was low, the relationship between species richness and community-level productivity was negative or neutral. However, under high Ammophila genetic diversity the relationship became positive. By contrast, increasing barley genotype diversity in our communities had negative effects on relationships between species richness and productivity. We hypothesise that the differences in responses may be related to the dominance of the particular species within which genetic diversity is changing. High genotypic diversity of dominant species (e.g. A. breviligulata) might allow greater niche space occupancy, reduce intraspecific competition and increase dominance. Therefore, increasing genetic diversity within dominants enhances the selection effect in species mixtures. By contrast, high genotypic diversity of subordinate species (e.g. barley) might reduce community-level productivity by reducing the dominance of the most productive species. Therefore, the enhanced overall competitive impact of barley due to increasing genotypic diversity from 0 to 1 to 5 increasingly constrains the selection effect. In summary, enhanced intraspecific diversity promotes positive complementarity effects through niche differentiation, in turn increasing or reducing selection effects in species mixtures depending on the dominance of the species within which genetic diversity increases. Few studies are available to test this tentative mechanism, but it seems plausible and represents an interesting future research target.
Enhanced weed species richness had the opposite effect to barley genotype richness, significantly increasing selection effects (Fig. 3e). This contrast is crude, as weed species richness was 0 or 5 (meaning that weed species were either all absent or all present in mixtures), but demonstrates the limited potential for selection effects when the mixture contains only intraspecific diversity (5B mixtures). Both barley genotype and weed species richness were associated with trends toward declining complementarity effects (Fig. 3c,f). Net biodiversity effects are the sum of selection and complementarity effects. The consequence is that with increasing barley genotype richness there is a trend toward a decline in net biodiversity effects (Fig. 3a), whereas with increasing weed species richness there is no change in net biodiversity effects as the positive effects of weed species richness on selection effects, and its negative impact on complementarity effects, cancel each other out.
In conclusion, even though intraspecific genotype diversity had weak direct effects on community-level productivity, it had significant indirect effects by modulating the species diversity–productivity relationship. This indirect effect mainly causes a change in the dominance patterns of the co-existing species, observable as a change in the selection effect of species diversity on productivity.
Trait-based approach to biodiversity effects
The net effect of interspecific diversity on productivity was about eight times higher than that of intraspecific genotype diversity, the main difference being strong, positive selection effects in species mixtures. Intraspecific diversity had weak effects on productivity, acting mainly through complementarity effects (see also Cook-Patton et al., 2011). Here, we propose a functional trait-based framework that explains different effects of intra- and interspecific diversity, and between-study differences in the magnitude of biodiversity effects, as well as temporal changes in the size of biodiversity effects. Trait-based approaches successfully explain biodiversity effects in species mixtures (Loreau, 2000; Cadotte et al., 2009; Flynn et al., 2011; Roscher et al., 2012, 2013). Our framework extends to the effects of intraspecific diversity on productivity, and we propose that genotypes and species may have complementary effects on productivity through complementarity and selection effects, respectively.
Intraspecific trait variability is generally smaller than interspecific variation (Fig. 4; Auger & Shipley, 2013). Limited intraspecific trait variation likely excludes the occurrence of extreme trait values needed for selection effects (Loreau, 2000), but is sufficient for niche differentiation (Jung et al., 2010) and hence complementarity effects. Complementarity effect size may then depend on trait variation, with more niche differentiation if trait variation is relatively high (but still below a threshold for selection effects). Our five barley genotypes showed a relatively small trait range despite being selected for distinctive weed-competitive traits (Fig. 4). This may explain their weak complementarity effects.
As trait variation increases, for example switching from genotype to species mixtures, selection effects are more likely (Loreau, 2000). Our dominant species, Stellaria, has relatively unique traits (Fig. 4), appearing to promote its dominance in mixtures. This leads to compensatory effects on co-occurring species and ultimately a selection effect. The selection effect is therefore likely important for net biodiversity effects in multi-species communities. As with complementarity effects, the magnitude of selection effects will co-vary positively with trait variation (Cadotte et al., 2009; Roscher et al., 2012).
We propose, therefore, that the relatively small trait variation associated with intraspecific diversity is the main driving force behind complementarity effects (in both genotype mixtures and in species mixtures that include intraspecific genetic diversity), whereas the relatively large trait variation associated with interspecific diversity leads to selection effects. As noted, we did not control for intraspecific genotypic diversity within our weed species, but analysis of the within-species trait variation indicates that intraspecific trait variation is similar between barley and most of the weed species (Table S4). This confirms that the increased trait variation we observed in weed species mixtures as opposed to barley genotype mixtures predominantly results from enhanced inter- rather than intraspecific diversity.
Studies following biodiversity effects through time in species-richness experiments with defined communities have indicated that selection effects are generally relatively constant (van Ruijven & Berendse, 2005; Cardinale et al., 2007; Fargione et al., 2007). If selection effects are driven mainly by between-species differences of species mixtures, this will be stable if species diversity is maintained. By contrast, the same studies demonstrate temporal increases in complementarity effects. As intraspecific genetic diversity and composition of these experimental assemblages is very likely to change over time due to genetic drift and selection (Nestmann et al., 2011), we propose that these changes are responsible for temporal increases in complementarity effects in natural systems.
The weaker complementarity compared with selection effects in our study may therefore be partially due to its short duration preventing genetic diversification or drift in the experimental communities. This could also explain the absence of transgressive over-yielding, a phenomenon thought to arise due to temporally increasing complementarity effects (van Ruijven & Berendse, 2005; Cardinale et al., 2007; Fargione et al., 2007). Our experimental duration was close to the field situation for annual crops. Consequently, complementarity effects may be limited in annual crop mixtures and transgressive over-yielding likely rare unless crop mixtures have rather high initial intraspecific diversity and rather high trait variation without extreme trait values (see Kiær et al., 2012). To increase complementarity effects, the design of genotypic mixtures of single species annual crops might therefore focus particularly on promoting genotypic diversity, but without extreme trait values. By contrast, mixtures involving combinations of species with very distinct trait values – for example classic intercrops – may result in positive selection effects. Thus their productivity might be improved further by increasing genetic diversity in the dominant species. Consequently, intercrops of genotypes or species, or both combined, may increase yield over average monocultures and particularly help to increase yield stability over time, as species and genotype mixtures are less prone to environmental fluctuations than a corresponding monoculture (Booth & Grime, 2003; Kiær et al., 2012; Finn et al., 2013).
Conclusion
Combining intraspecific genotype mixtures and interspecific species mixtures revealed that intraspecific genotype diversity may affect productivity through complementarity effects, whereas between-species differences promote selection effects on productivity. We propose, using a trait-based framework, that limited trait variation associated with within-species genetic diversity is responsible for the complementarity effects in diversity–productivity studies. By contrast, wider trait variation associated with between-species diversity drives selection effects whereby some species dominate and suppress subordinate species. We also propose that the direction and strength of selection effects are likely driven by genetic and phenotypic variation within dominant vs subordinate species and their corresponding complementarity effects. Strong complementarity effects within dominant species will increase selection effects, and complementarity effects among subordinate species will reduce them. These findings indicate the crucial importance of intraspecific genetic diversity for ecosystem functioning and the necessity to account for genetic diversity in studies of biodiversity effects on ecosystem functions.
Acknowledgements
We thank Gill Banks, Linda Ford, Marie Gallon, Carol Kyle, Laura Lopez del Egido and Allan Sim for their help with the experiment, Paul Shaw and Bill Thomas for the description of the barley cultivars, David Gibson for comments on the experimental design, Yann Hautier for advice on statistical analyses, and two anonymous referees for helpful comments on a previous version of the manuscript. C.S. was supported by the Swiss National Science Foundation (PA00P3_136474 and PZ00P3_148261), S.K. by the Leonardo da Vinci programme from the European Commission, and L.M. by the Spanish Ministerio de Economía y Competitividad (AGL2008-05532-C02-01). The study was funded by the Scottish Government's Rural and Environment Science and Analysis Services (RESAS) through Workpackages 1.1 and 3.4 of its Strategic Research Programme.
1469-8137/images/journal_branding.png)
