Complementarity of Gymnosperms and Angiosperms Along an Altitudinal Temperature Gradient

Home , Evergreen

doi: 10.1111/oik.05360 127 1787–1799 OIKOS Research Complementarity of gymnosperms and angiosperms along an altitudinal temperature gradient

John P. Caspersen, Esther Thürig, Andreas Rigling and Niklaus E. Zimmermann

J. P. Caspersen (http://orcid.org/0000-0003-2931-9755) ([email protected]), Faculty of Forestry, Univ. of Toronto, 33 Willcocks Street, Toronto, ON, M4X1A4, Canada. – E. Thürig, A. Rigling and N. E. Zimmermann, Swiss Federal Res. Inst. WSL, Birmensdorf, Switzerland.

Oikos In seasonal tropical forests, evergreen–deciduous mixtures are more productive than 127: 1787–1799, 2018 monocultures because they intercept more light throughout the year, reflecting com- doi: 10.1111/oik.05360 plementary resource use by functional groups possessing different traits. This suggests that temperate and boreal forests may also exhibit overyielding, due to the difference in Subject Editor: Robin Pakeman phenology between gymnosperms and angiosperms. However, complementarity could Editor-in-Chief: Gerlinde De Deyn also arise from differences in morphology between needle leaves and broad leaves, Accepted 13 June 2018 or facilitation by N-fixing species. Alternatively, mixtures may be more productive simply because interspecific competition is less intense than intraspecific competition. We used forest inventory data to assess the complementarity of the main functional groups in Switzerland. We employed a trait-based analysis of competition to determine whether: 1) trait differences reduce the intensity of competition between complementary functional groups, 2) complementarity is observed along a broad altitudinal temperature gradient. N-fixing species facilitated the growth of non-fixing species, such that 50/50 mixtures were 50% more productive than monocultures, though half the overyielding was due to the alleviation of intraspecific competition. In contrast, we found no evidence of complementarity between evergreen and deciduous species. For example, in stands where larch was mixed with other gymnosperms, there was no reduction in heterospecific competition among evergreen species, even though evergreens cast shade on one another throughout the year. In cold montane forests, broadleaf species reduced the suppression of needleleaf species, and vice versa. Thus, 50/50 mixtures of needleleaf and broadleaf species were 15% more productive than needleleaf monocultures. However, in warm lowland forests, broadleaf species exac- erbated the suppression of needleleaf species, completely offsetting the positive effect that needleleaf species had on broadleaf species. In summary, we found no evidence of complementarity between evergreen and deciduous species, but needleleaf–broadleaf mixtures exhibited overyielding in cold montane forests, which is consistent with the stress gradient hypothesis, though the underlying mechanisms remain uncertain.

Keywords: competition, leaf traits, mixed forests, overyielding, plant–plant interactions, stress gradient hypothesis

1787 Introduction stress gradient hypothesis (Bertness and Callaway 1994, Jucker et al. 2016). Biodiversity–productivity experiments have shown that In this paper, we use forest inventory data to examine species mixtures containing multiple functional groups are the wealth of species interactions observed in pre-existing more productive than species mixtures containing only one plots spanning a large altitudinal temperature gradient in functional group, even if the number of species is held con- Switzerland (Brändli et al. 2010). We conducted a trait-based stant across all treatments (Reich et al. 2004, Mouillot et al. analysis of competition to assess the complementarity of the 2011). The enhanced productivity of these mixtures is called main functional groups, including a deciduous gymnosperm overyielding, and is the result of complementary resource (for the reason given above) as well as three N-fixing angio- use by functional groups that possess different traits. For sperms, in order to demonstrate the utility of our trait-based example, biodiversity–productivity experiments conducted methods using a functional group (N-fixing angiosperms) in seasonal tropical forests have also documented overyield- that is known to promote overyielding through facilitation. ing in mixtures of evergreen and deciduous tree species, However, since overyielding also occurs in the absence of indicating that these two functional groups exhibit temporal facilitation, our method focuses on competition as a com- niche complementarity due to their difference in phenology mon mode of interaction between complementary species. (Sapijanskas et al. 2014). In particular, species differences in Our analysis builds on the method developed by phenology allow mixtures to intercept more light throughout Kunstler et al. (2016), which was designed to quantify how the year than stands comprised of only one of the functional trait differences reduce the intensity of competition between groups (Forrester and Pretzsch 2015, Forrester and Bauhus two or more complementary species, thereby promoting 2016), a phenomenon known as temporal light partitioning. overyielding. In contrast, many biodiversity–productivity This suggests that overyielding may also occur in tem- studies do not directly assess the impact of one species on perate and boreal forests (Cannell et al. 1992, Scherer- another, because they do not explicitly quantify the intensity Lorenzen et al. 2005), due to the difference in phenology of competition, or how it varies with trait differences. We between gymnosperms and angiosperms (Ishii and Asano modified the method of Kunstler et al. (2016) to estimate a 2010, Lu et al. 2016), provided that the shoulder growing set of ‘pairwise dissimilarity effects’ that quantify how binary seasons for evergreen species (before emergence and after trait differences affect the intensity of competition between abscission of deciduous leaves) are long enough to allow for complementary functional groups (as opposed to species), temporal niche complementarity (Chabot and Hicks 1982, while allowing for asymmetric interactions in which only one Kikuzawa 1991, Givnish 2002, van Ommen Kloeke et al. species benefits from the interaction. 2012). Yet, complementarity could arise from other trait dif- Our study was designed to address the following ques- ferences between these two functional groups, such as the tions regarding mixtures that contain two or more functional difference in morphology between needle leaves and broad groups: 1) Do differences in phenology reduce the intensity leaves (Augusto et al. 2015). For example, leaf litter qual- of competition between evergreen and deciduous gymno- ity has a profound influence on decomposition and nutrient sperms? 2) Do differences in phenology or leaf morphology cycling (Cornwell et al. 2008), so mixing these very different reduce the intensity of competition between angiosperms and litters may allow for complementary use of soil resources by gymnosperms? 3) Do N-fixing species facilitate the growth of needleleaf and broadleaf species (Gessner et al. 2010). non-fixing species, indicating that the supply of extra nitro- Numerous studies have examined the productivity of gen completely offsets any competitive effects? 4) Do these monocultures and mixed stands comprised of two common dissimilarity effects cause overyielding in any of the following temperate species such as Norway spruce and European beech mixtures: evergreen and deciduous gymnosperms, needleleaf (Pretzsch 2005, Pretzsch et al. 2017). However, such pairwise gymnosperms and broadleaf angiosperms, N-fixing and non- replacement studies generally do not attempt to disentangle fixing angiosperms? 5) Do these dissimilarity effects depend the effects of phenology and leaf morphology. Nevertheless, on temperature, such that the magnitude of overyielding they do have considerable practical value because coni- varies across the altitudinal temperature gradient? fer stands are being widely converted to mixed stands that include a variety of species (Knoke et al. 2008). Thus, the results of pairwise replacement studies must be generalized Material and methods by assessing complementarity across a broad variety of species and functional groups (Mina et al. 2017), including We used data from the National Forest Inventory (NFI) of mixtures of evergreen and deciduous gymnosperms, in order Switzerland, which is inventoried on a 1 × 1 km grid covering to assess the complementarity of species differing in phenol- the entire country (Brassel and Lischke 2001). The inventory ogy separately from that of species differing in leaf morphol- plots span a broad gradient in altitude, from 281 m a.s.l. in ogy. Furthermore, complementarity must be assessed across the lowlands to 2218 m a.s.l. in the Alps, and a broad gra- a broad range of climates, in order to determine whether dient in mean annual temperature, from 12.2°C to 0.6°C, overyielding increases as temperature decreases (Paquette and respectively (Supplementary material Appendix 1). Montane Messier 2011, Potter and Woodall 2014, Wu et al. 2015, forests with mor soils are dominated by gymnosperms (e.g. Jucker et al. 2016, Mina et al. 2017), as predicted by the spruce, fir and larch), whereas lowland forests with mull

1788 soils are dominated by angiosperms (e.g. beech, maple and where ns is the number of individuals of species s, BAi is the 2 ash). However, both groups co-occur across the gradient, and basal area (m ) of tree i in the first inventory, iA is the area angiosperms are locally abundant in montane forests, par- (m2) of the sample plot for tree i (trees greater than 36 cm ticularly in young stands on recently disturbed sites, while DBH were sampled in larger plots), 10 000 is the number gymnosperms are locally abundant in lowland forests, par- of square meters in a hectare, BAIi is the annual basal area ticularly in managed stands. In these cases, local abundance increment (calculated using the amount of time that elapsed reflects transient dynamics driven by natural disturbance between the first and second inventory), and sG is the basal and/or management. area increment (m2 ha–1 year–1) of a species cohort. Together, these two variables were used to estimate the Data collection relative growth rate of each species, or the growth rate per unit basal area, as explained below. Note that this growth met- We used data from the first and second census (1983-1995), ric does not account for differences in wood density among and excluded plots that were harvested between the two cen- species or functional groups. Wood density of angiosperms sus periods. The resulting dataset included 2891 plots con- is generally higher than that of gymnosperms, so angio- sisting of two concentric circles with the following horizontal sperms accrue more mass per unit growth in basal area. This 2 projection area: a 200 m plot for sampling trees 12-36 cm difference between angiosperms and gymnosperms should be 2 DBH, and a 500 m plot for sampling trees >= 36 cm DBH. kept in mind when interpreting the rate of basal area growth. The diameter and species identity of each tree was measured in both censuses, except trees which died or were cut by Trait-based model of asymmetric interactions the second census, in which case the status of the tree was recorded. The dataset includes 50 species from 4 functional Similar to Kunstler et al. (2016), we estimate a set of ‘pair- groups possessing a different combination of 3 binary traits: wise dissimilarity effects’ that quantify how binary trait dif- needleleaf/broadleaf, evergreen/deciduous, N-fixer/non- ferences alter the intensity of competition between pairs fixer (Supplementary material Appendix 1). The functional of functional groups. For example, our dissimilarity effects groups include evergreen gymnosperms, deciduous gymno- would allow deciduous species to reduce the suppression sperms, deciduous angiosperms, and N-fixing angiosperms of evergreen species, compared to the suppression that two (also deciduous). Larch Larix decidua was the only deciduous evergreen species exert on one another in a purely evergreen gymnosperm, while the N-fixing group includes two native stand (given that shade is cast throughout the year). However, alder species (Alnus incana and Alnus glutinosa) as well as one our model differs in two specific ways. First, our dissimilarity invasive species from North America (Robinia pseudoacacia). effects allow for asymmetric interactions in which only one For each plot, we also assembled a set of environmental species benefits from the interaction. For example, our model variables, including slope, aspect, mean annual temperature, allows for the possibility that evergreen species do not have an annual precipitation, and a drought index. Slope and aspect equivalent effect on deciduous species (the interaction is non- were measured in each plot using a clinometer and com- reciprocal). Second, we focus on binary traits because they can pass, respectively. The climate variables were calculated using be used to assign species to discrete functional groups, unlike daily data from MeteoSwiss weather stations, which were continuous traits such as specific leaf area (Kunstler et al. first interpolated to a 100 m resolution using a digital eleva- 2016). This approach allows us to determine whether function model and the Daymet software from Thornton et al. tional group diversity enhances productivity independently (1997). The mean annual temperature was then calculated by of species diversity (Reich et al. 2004, Mouillot et al. 2011, averaging the daily values observed between 1983 and 1995. Sapijanskas et al. 2014). Furthermore, this allows us to draw The drought index was calculated by taking the difference general conclusions that can be readily interpreted by forest- between the daily precipitation and potential evapotranspo- ers and modelers who use functional groups to implement ration, then averaging across the same time period. Potential models or management strategies. evapotranspiration was calculated following Turc (1961) as In its most basic form, the model quantifies the growth implemented in Zimmermann and Kienast (1999). rate of a species cohort as a negative exponential function of the plot basal area (BA): Data processing GB= AG e−uBA  (3) For each species, we calculated the basal area (BAs) and fm ax  growth in basal area (Gs) at the plot level:

ns where BAf is the basal area of the focal species, Gmax is its 10000 maximum relative growth rate, and u quantifies the rate its BAs = ∑BAi (1) i=1 Ai growth decreases as the plot basal area increases (including the basal area of the focal species itself). As explained further ns 10000 below, the magnitude of growth suppression can also vary GBs = AIi (2) with the species composition of the plot (Eq. 6). Note that ∑ A i=1 i for the sake of simplicity, we did not analyze growth at the

1789 level of individual trees, as Kunstler et al. (2016) did. Rather, area of the focal species (BAf) is also included in BA (see we calculated the growth in basal area of all the trees in a plot Supplementary material Appendix 2 for further explanation). that belong to a given species. The last term in Eq. 6 includes a set of ‘pairwise dissimilarity To facilitate linear regression analysis, we first log trans- effects’ that quantify the positive or negative effect of neigh- formed Eq. 3 to linearize the relationship between growth bors possessing traits that the focal species does not possess and competition: (t ≠ f). These pairwise dissimilarity effects (ρ) fall into one of three categories: logG =+logBAlfmogGuax − BA (4) Exacerbation: the trait exacerbates the suppression of the focal species, compared to the suppression exerted by a Then, we included various fixed and random effects to heterospecific competitor with the same traits as the focal allow the maximum growth rate to vary with environmental species. Note that the term ‘dissimilarity effect’ specifically conditions: refers to interactions between functional groups, because this extra competitive effect is not exerted on species that =+αβ++γδ++εε logGmaxpTSA s (5) share the same traits (belong to the same functional group). Also note that the term ‘pairwise effect’ refers to interactions where logGmax is the maximum relative growth rate of the between specific pairs of functional groups, so a functional species cohort s in plot p, and the Greek letters are estimated group may exert several such dissimilarity effects, one for regression coefficients. each trait defining the group. The first four terms in Eq. 5 specify how maximum growth Facilitation: the trait offsets the effect of competition, varies around the intercept (α) as a function of mean annual thereby facilitating the growth of the focal species. For exam- temperature (T), slope (S), and aspect (A). The remaining ple, N-fixing neighbors actually increase the growth of non- terms, εS and εp, are normally-distributed random effects fixing species, indicating that the supply of extra nitrogen (with a mean of zero) that allow growth to vary among spe- more than offsets any competitive effects, such as shading. cies and plots. Note that we included species as random Again, note that this facilitative effect does not benefit species effects because our goal was to quantify differences in growth that share the same trait as the neighbor. Also note that facili- among functional groups, rather than differences in growth tation is not necessarily reciprocal, since two separate dissimi- among species per se. larity effects were estimated for each interaction (e.g. ρ We did not include trait-specific terms in Eq. 5 because pre- N-fixing and ρnon-fixing), one of which can be negligible (e.g. ρnon-fixing). liminary analyses indicated that the functional groups do not Amelioration: the trait ameliorates the suppression of the differ in their maximum growth rate (an assumption that we focal species, compared to the suppression exerted by a het- tested along with several others, as explained further below). erospecific competitor with the same traits as the focal spe- However, given the difference in wood density between cies. For example, deciduous species may suppress evergreen gymnosperm and angiosperms, one can assume that broadleaf species less than evergreen species suppress one another (given species accrue more mass per unit growth in basal area. that they cast shade throughout the year). This will result in Differences in relative growth do, however, emerge under overyielding (complementarity) provided that evergreen spe- competition. Thus, we added two terms to Eq. 4 to allow cies do not suppress deciduous species more than deciduous the suppression of the focal species to vary not only with the species suppress one another (exacerbation), in which case relative abundance of conspecific and heterospecific neighbors, the two asymmetric effects may cancel one another. It is but also the traits of both the focal species and its neighbors: important to note that Kunstler et al. (2016) did not allow 3 for asymmetry because they only estimated one parameter logG =+logBAlogGu−−BA π BA fmax ∑ t=1 tf that applies to both of the interacting species, affecting the 3 + ρ BA (6) suppression of both in proportion to the absolute value of the ∑ t=1 tf≠≠tf difference of the continuous trait values. where BAf is the basal area of a focal species possessing three Accounting for the effect of temperature particular traits (t), and BAt≠f is the basal area of neighbors possessing one of the three traits that the focal species does According to the stress gradient hypothesis, species are more not possess (t≠f). Note that BAt≠f takes three different values, likely to facilitate one another’s growth at low temperatures, one for each binary trait. For example, if the focal species is or at least ameliorate the effects of competition observed at larch, then the three values are BAbroadleaf, BAevergreen, BAN-fixer. higher temperatures (Jucker et al. 2016). Thus, we included With the addition of these new terms, µ now quantifies an additional term that allows dissimilarity effects to vary the intensity of interspecific competition, and πt quantifies with temperature (T): the incremental intensity of intraspecific competition. For 3 logG =+logBAlogGB−−µπABA example, if πbroadleaf = µ, then broadleaf species are twice as fmax ∑ t=1 tf sensitive to intraspecific competition than they are to inter- 3 ++((ρλBA BA × T))) (7) specific competition (all else being equal), since the basal ∑ t=1 tf≠≠tf tf≠≠tf

1790 where the sign of λ denotes whether the dissimilarity effect functional group is zero, because the mean of the random increases or decreases with temperature, and the sign of ρ species effects is zero. Then, we calculated the percentage determines whether the dissimilarity effect is positive or neg- increase in growth of mixed stands relative to a monocul- ative at low temperatures. Note that if λ and ρ differ in sign, ture, assuming a standardized total basal area of 30 m2 ha–1 then the dissimilarity effect may switch from positive (ame- for both mixtures and monocultures as basis for comparison. lioration, facilitation) to negative (exacerbation) as tempera- Finally, we plotted the increased growth of mixtures (y-axis) ture increases (or vice versa). as a function of stand composition, as quantified by the proportion of basal area comprised by one of the two species Alternative models and comparison of AIC values (x-axis).

We fitted our model in R (< www.r-project.org >) using the Data deposition lme4 package for mixed effects models (Bates et al. 2015), and calculated Aikaike’s information criterion (AIC) to com- Long-term storage and availability of the data (< https://doi. pare it to various alternative models (Supplementary material org/10.21258/1000001 >) is guaranteed by Swiss law (SR Appendix 3). We found that the model described above (Eq. 510.620 and in the ‘Bundesgesetz über die Archivierung’ 5, 7) had the lowest AIC, indicating that it is the most par- SR 152.1). Tables with the results of the four Swiss NFIs simonious model, and that additional terms did not signifi- are freely available for download from the website < www.lfi. cantly improve model fit (Supplementary material Appendix ch/resultate/ >, while the complete list of available tree- and 4). Finally, we also used AIC comparisons to assess terms for plot-level attributes can be found at < www.lfi.ch/dienstleist/ individual traits (e.g. ρnon-fixing) and dropped them from the katalog.php?lang=de >. Plot and tree data from the Swiss NFI model if including them did not significantly improve the fit can be provided free of charge within the scope of a contrac- of our model based on AIC comparisons. tual agreement (< www.lfi.ch/dienstleist/daten-en.php >).

Calculating the growth rate of mixtures relative to monocultures Results

We used the trait-based model (Eq. 5, 7) and the estimated For all species, growth declined at the same rate with parameters to quantify the magnitude of overyielding in increasing abundance of heterospecific neighbors, provided two-species mixtures. We first calculated the growth rate of they were in the same functional group (Table 1), reach- mixtures and monocultures of the two species, assuming that ing 49% of the maximum rate at a basal area of 40 m2 ha–1 species from the same functional group have the same growth (Fig. 1). The additional effect of intraspecific competition rate: while any given species pair may differ in growth, the varied among functional groups, since broadleaf species were average difference between species belonging to the same more sensitive to intraspecific competition than needleleaf

Table 1. Estimated parameters of a trait-based model of competition in mixed gymnosperm–angiosperm forests (Eq. 5, 7). Parameters α, β, γ, and δ quantify the effect of environmental conditions on the maximum growth rate (Eq. 5). Parameter µ (Eq. 7) quantifies the competitive effect of heterospecific neighbors with the same traits as the focal species (black line in Fig. 1). Parameterπ (Eq. 7) quantifies the incremental competitive effect of conspecific neighbors (lighter lines in Fig. 1). Parametersρ and λ (Eq. 7) determine whether neighbors with different traits decrease (or increase) suppression (Fig. 3, 5), relative to that exerted by heterospecific neighbors with the same traits (black line in Fig. 3, 5). Note that the lower and upper bounds are 95% confidence limits, and that terms for individual traits were dropped from the model

(e.g. ρnon-fixing) if including them did not significantly improve model fit.

Equation Symbol Trait Estimate Lower bound Upper bound Maximum growth rate 5 Intercept α – 0.030 0.025 0.035 Temperature β – 0.045 0.030 0.059 Slope γ – –0.00340 –0.00430 –0.00250 Aspect δ – –0.00020 –0.00041 –0.00011 Competition (Fig. 1) 7 Interspecific (black line) µ – 0.017 0.015 0.019 Intraspecific π (dark gray line) needleleaf 0.0072 0.0048 0.0095 (light gray line) broadleaf 0.0160 0.0120 0.0190 Dissimilarity effects (Fig. 3, 5) 7 ρ broadleaf 0.020 0.005 0.032 N-fixer 0.018 0.005 0.032 λ needleleaf 0.00087 0.00048 0.00012 broadleaf –0.00370 –0.00560 –0.00170

1791 100

s 75

50 (% of maximum)

Growth of focal specie 25 All species, when heterospecific neighbors belong to the same functional group Needleleaf species, when neighbors are conspecific Broadleaf species, when neighbors are conspecific 0 0 10 20 30 40 Basal area (m2 ha–1)

Figure 1. Growth suppression due to inter- and intraspecific competition among species that share the same traits (belong to the same functional group). The distance between the black line (interspecific competition) and the other two lines reflects the incremental strength of intraspecific competition: broadleaf species are more sensitive to intraspecific competition than needleleaf species. The vertical lines denote 95% confidence intervals (in subsequent figures as well). species (Table 1). For example, in monocultures with a basal the maximum rate (Fig. 1). Thus, 50/50 mixtures of two area of 40 m2 ha–1, the growth rate of broadleaf species was different broadleaf species were 27% more productive than reduced to 25% of the maximum, whereas the growth of broadleaf monocultures (Fig. 2). In contrast, needleleaf needleleaf species was only reduced to 37% of the maximum species grew only 32% faster when surrounded by other growth rate (Fig. 1). evergreen species, increasing from 37% to 49% of the max- Since broadleaf species are more sensitive to intraspe- imum rate (Fig. 1). Thus, 50/50 mixtures of two needleleaf cific competition, they respond more to the alleviation species were only 11% more productive than needleleaf of intraspecific competition. At 40 2m ha–1, for example, monocultures (Fig. 2). broadleaf species grew twice as fast when surrounded by These yield curves are symmetrical along the x-axis other broadleaf species, increasing from 25% to 49% of because there are no dissimilarity effects, and because both

35 Broadleaf species 30 Needleleaf species )

10 Increased growth of mixtures (% 5

0 00.2 0.40.6 0.81

Proportion of basal area comprised of 2nd species

Figure 2. Overyielding in two-species mixtures of broadleaf species and two-species mixtures of needleleaf species, calculated using the estimated parameters (Table 1) of the trait-based model of competition (Eq. 5, 7). Increased growth is the percentage increase in growth of broadleaf and needleleaf mixtures compared to their respective monocultures (assuming basal areas of 30 m2 ha–1). The increased growth of broadleaf mixtures is greater because broadleaf species are more sensitive to intraspecific competition (Fig. 1). Since two species from the same functional group do not exert dissimilarity effects on one another, overyielding is the result of alleviating intraspecific competition.

1792 monocultures of the same functional group are assumed Dissimilarity effects of evergreen and deciduous species to have the same growth rate. We made this assumption because the mean of the random species effects is zero, There was no evidence of complementarity between evergreen and because the goal was simply to demonstrate that and deciduous species. Deciduous species exerted no signifi- overyielding is caused by the alleviation of intraspecific cant dissimilarity effects on evergreen species (Table 1), indi- competition. However, adding random species differences cating that deciduous species do not reduce the suppression of would reduce the magnitude of overyielding, and this evergreen species, compared to that exerted by other evergreen effect declines as the difference in growth between mono- species. For example, mixing larch with other gymnosperms cultures increases (all else being equal). Large differences does not reduce the effect of heterospecific competition would render mixtures less productive than the most pro- among evergreen species, even though evergreen species cast ductive monoculture. It is also important to note that the shade on one another throughout the year (further discussion magnitude of overyielding varies with the metric used to in Supplementary material Appendix 5). Thus, mixing larch calculate the mixing proportions (Pretzsch 2005, Forrester with other gymnosperms only results in overyielding due to and Pretzsch 2015). the alleviation of intraspecific competition (Fig. 2). Dissimilarity effects of needleleaf and broadleaf species Dissimilarity effect of N-fixing species The interactions between needleleaf and broadleaf species N-fixing species facilitated the growth of non-fixing spe- depended strongly on temperature (Table 1). At cold tem- cies (Fig. 3), and this effect was independent of temperature peratures (3°C annual mean), they both exerted a positive (Table 1), indicating that the supply of extra nitrogen more dissimilarity effect on one another, though it did not amount than offsets any negative effects of heterospecific competition, to facilitation. Rather, they ameliorated the effect of het- regardless of temperature. However, the dissimilarity effects erospecific competition, indicating that needleleaf species were non-reciprocal, because non-fixing species exerted no suppress broadleaf species less than broadleaf species suppress significant dissimilarity effects on N-fixing species, though one another (Fig. 3), and vice versa (Fig. 5). Thus, in mon- they did suppress their growth as heterospecific competitors tane forests, mixtures were up to 15% more productive than (i.e. as different deciduous species). needleleaf monocultures (Fig. 6), which are more productive Mixing N-fixing angiosperms with non-fixing angio- than broadleaf monocultures, because needleleaf species are sperms results in substantial overyielding (Fig. 4). 50/50 less sensitive to intraspecific competition (Fig. 1). mixtures of N-fixing and non-fixing species are 50% more In lowland forests, needleleaf species also ameliorated the productive than monocultures of either functional group, effect of competition on broadleaf species (Fig. 3), because because N-fixing species facilitate the growth of non-fixing this positive effect increased with temperature (Table 1). In species (Fig. 3). However, about half of the overyielding is contrast, the positive effect of broadleaf species decreased due to the alleviation of intraspecific competition, as shown with temperature (Table 1), such that the sign of the interac- by the relative height of the lower curve, which excludes the tion effect was negative in lowland forests. Thus, broadleaf dissimilarity effect. species exacerbate the effect of competition on needleleaf

125

100

(% of maximum) N-fixing neighbors Needleleaf neighbors at 9 °C 25

Growth of braodleaf species Needleleaf neighbors at 3 °C Heterospecific neighbors 0 010203040

Basal area (m2 ha–1)

Figure 3. Dissimilarity effects on the growth of broadleaf species. The black line is a reference case (also shown in Fig. 1) quantifying the effect of heterospecific neighbors with the same traits. The other lines indicate whether neighbors with different traits decrease (or increase) the suppression of the broadleaf species, relative to that exerted by heterospecific neighbors with the same traits (black line).

1793 60 with dissimilarity effect without dissimilarity effect 50 )

20 Increased growth of mixtures (% 10

0 00.2 0.40.6 0.81 Proportion of N-fixing species

Figure 4. Overyielding in two-species mixtures of N-fixing and non-fixing angiosperms, both with (ρ > 0) and without the dissimilarity effect (ρ set to 0). Increased growth is the percentage increase in growth of mixtures relative to monocultures (assuming a basal area of 30 m2 ha–1), and was calculated using the estimated parameters (Table 1) of the trait-based model of competition (Eq. 5, 7). Note that about half of the increased growth of mixtures is due to the alleviation of intraspecific competition, as shown by the relative height of the lower curve, which excludes the dissimilarity effect. species (Fig. 5), completely offsetting the positive effect that groups that possess markedly different traits. Yet, many pair- needleaf species have on broadleaf species, such that needle- wise replacement studies do not explicitly quantify the inten- leaf-broadleaf mixtures were not more productive in lowland sity of competition and how it varies with trait differences forests (Fig. 6). (Uriarte 2010, Kunstler et al. 2012, Kunstler et al. 2016, Levine 2016). As a result, it often remains uncertain why, and in what combinations, mixtures are more productive than Discussion monocultures. In particular, it remains uncertain whether mixing gymnosperms and angiosperms increases productiv- It is generally recognized that complementarity arises when ity due to the differences in phenology and leaf morphology, one species reduces the suppression of another species (and or unknown niche differences between species that happen to vice versa), particularly when the species belong to functional belong to different functional groups.

100

50 (% of maximum) 25 Broadleaf neighbors at 3 °C Growth of needleleaf species Heterospecific neighbors Broadleaf neighbors at 9 °C 0 010203040

Basal area (m2 ha–1)

Figure 5. Dissimilarity effects on the growth of needleleaf species. The black line represents the reference case (also shown in Fig. 1) quantifying the effect of heterospecific neighbors with the same traits. The dashed lines indicate whether neighbors with different traits decrease (or increase) the suppression of needleleaf species, relative to that exerted by heterospecific neighbors with the same traits (black line).

1794 20 3 °C 9 °C

) 10

–10

–20

–30

Increased growth of mixtures (% –40

–50 00.2 0.40.6 0.81 Proportion of the needleleaf species

Figure 6. Overyielding in needleleaf–broadleaf mixtures, calculated using the estimated parameters (Table 1) of the trait-based model of competition (Eq. 5, 7). The increased growth of mixtures is calculated relative to the respective needleleaf monocultures (assuming a basal areas of 30 m2 ha–1), which grow faster than broadleaf monocultures because needleleaf species are less sensitive to intraspecific competition (Fig. 1).

In order to dispel this uncertainty, we estimated a set of 100% of their nitrogen intake from the atmosphere, which ‘functional dissimilarity effects’ that quantify how trait differ- is made available to neighboring plants through decompo- ences reduce the intensity of competition between functional sition, root exudation, and mycorrhizal transfer (Rothe and groups. Our method allows for interactions that are both Binkley 2001, Richards et al. 2010). Indeed, we found that asymmetric and non-reciprocal, and it allows us to determine N-fixing species facilitate the growth of other species (but not the extent to which overyielding is caused by the alleviation vice versa) such that mixtures are 50% more productive than of intraspecific competition. Our results show that both monocultures. This effect was independent of temperature asymmetric and non-reciprocal interactions occur between (Table 1), but it remains uncertain whether overyielding is the main functional groups of Swiss tree species, including so large at the coldest end of the temperature gradient, since needleleaf and broadleaf species (Table 1, Fig. 1, 3). However, there were only three plots with nitrogen fixers at tempera- we found no evidence of complementarity between evergreen tures below 4.5°C (Supplementary material Appendix 1). and deciduous species. These results are consistent with other studies that Dissimilarity effects have long been considered a key mech- show mixtures with alder species are more productive than anism by which traits might affect competition (MacArthur monocultures without alder, particularly on poor soils and Levins 1967), but evidence for them is limited (Mayfield (Binkley et al. 1992, Mason and Connolly 2013). It is also and Levine 2010). While the recent global analysis by consistent with a recent meta-analysis of 18 studies which Kunstler et al. (2016) found that competition within species found that N-fixing mixtures were significantly more produc- was stronger than between species, they also found that the tive than monocultures in most studies (Forrester et al. 2006). degree of trait dissimilarity between species had little influ- However, we found that half of the overyielding in N-fixing ence in reducing competition. In particular, they found that mixtures is due to the alleviation of intraspecific competition, the intensity of competition does not vary with continuous and therefore attributable to unknown differences between differences in specific leaf area, contrary to the dissimilar- species that happen to belong to different functional groups. ity effects between needleleaf and broadleaf species reported Thus, caution is warranted in interpreting pairwise studies here. This discrepancy likely reflects the fact that their model that fail to disentangle the effects of intraspecific competition did not allow for asymmetry, because they only estimated and nitrogen fixation, lest overyielding be falsely attributed one parameter that applied to both of the interacting species to nitrogen fixation. jointly, ameliorating the suppression of both in proportion to the absolute value of the difference in continuous trait values. Mixing evergreen and deciduous gymnosperms

Facilitation by nitrogen fixers We found no evidence for temporal niche complementarity among gymnosperms. In particular, we found that mixing It is well known that certain functional groups exert a the deciduous larch with other gymnosperms species only profound and asymmetric influence on competition and pro- increases productivity due to the alleviation of intraspecific ductivity. N-fixers can acquire anywhere from 10% to nearly competition. Other studies have also reported overyielding

1795 when larch is mixed with other gymnosperms (Morgan et al. Stress gradient hypothesis 1992, Cameron and Watson 2000, Pretzsch 2005), but these studies did not attempt to disentangle the effects of intraspe- We found that dissimilarity effects only increase the pro- cific competition and seasonal light partitioning. Thus, the ductivity of needleleaf–broadleaf mixtures at one end of the evidence for temporal niche complementarity between gym- altitudinal temperature gradient. This result is consistent nosperms remains inconclusive. with the stress gradient hypothesis (Maestre et al. 2009, However, recent studies have shown that complementarity Jucker et al. 2016), as well as with studies showing that the is more likely to occur in vertically stratified stands in which strength of the correlation between diversity and productiv- understorey trees belong to a different species or functional ity increases as temperature decreases with latitude (Paquette group than overstory trees (Dănescu et al. 2016, Madrigal- and Messier 2011, Potter and Woodall 2014, Wu et al. 2015, González et al. 2016). This suggests that vertical stratification Jucker et al. 2016). This suggests that altitudinal and latitudi- of evergreen–deciduous stands (i.e. an evergreen understory nal trends may reflect the same underlying mechanisms, and below a deciduous canopy) may allow them to intercept more perhaps temporal light partitioning is not one of them. Yet, light throughout the year than stands comprised of only one the mechanisms underlying the complementarity of angio- of the functional groups. Yet, our method does not account sperms and gymnosperms remain uncertain, and none of the for tree size or stand structure, so it must be modified to assess other hypotheses invoke productivity or temperature as an whether vertical stratification alleviates the size asymmetry of important factor. competition for light. Other than temporal light partitioning (Sapijanskas et al. 2014), the most common hypothesis is that overyielding is A long-lasting debate about mixing gymnosperms and the result of vertical stratification of complementary crown angiosperms shapes, which allows mixed stands to fill canopy space more effectively than stands comprised of a single crown shape, The complementarity of gymnosperms and angiosperms particularly if crowns exhibit greater plasticity in mixed has been actively debated since the dawn of silviculture in stands (Pretzsch 2014, Jucker et al. 2015, Pretzsch and Europe, with early practitioners arguing both for and against Schütze 2016). This hypothesis generally assumes that both mixed stands (Pretzsch 2005). The subsequent establish- gymnosperms and angiosperms benefit from stratification ment of conifer stands throughout Europe attests to the and crown plasticity, regardless of the productivity of the fact that they provide higher volume yields in monoculture environment. Indeed, the positive effect of mixing on canopy (Pretzsch et al. 2015), which is consistent with our results packing is strongly conserved across the full latitudinal gra- and other studies suggesting that conifer stands are more pro- dient of Europe (Jucker et al. 2015). Other light partition- ductive (Augusto et al. 2015). However, the relative produc- ing hypotheses emphasize differences in physiology and the tivity of mixed stands has remained a topic of debate until display of foliage (Ishii and Asano 2010), while water and the present. soil nutrients have also been hypothesized to allow for spatial Interest in this topic has surged in recent years because resource partitioning (Pretzsch et al. 2015). However, most conifer stands are being converted back to mixed stands at hypotheses are invoked to explain overyielding and therefore an ever-increasing rate (Knoke et al. 2008). Thus, numer- assume the interactions to either be symmetric and positive ous studies have examined the productivity of mixed stands or non-reciprocal, rather than asymmetric under certain comprised of common temperate species such as Norway conditions, as we found in lowland temperate forests. spruce and European beech (Jones et al. 2005, Pretzsch 2005). These studies have found that some pairs exhibit Asymmetry and the role of litter quality overyielding across a wide range of conditions, such as Scots pine and European beech (Pretzsch et al. 2015), while others Pairwise studies have also reported negative interactions, only under certain conditions, such as Norway spruce and including asymmetric interactions, in which one species sup- European beech (Pretzsch and Schütze 2009, Toïgo et al. presses another, but is itself unaffected, or even benefits from 2015). Again, these pairwise studies generally do no attempt the interaction (Jones et al. 2005, Pretzsch 2005). While some to disentangle the effects of phenology and leaf morphology, studies report that broadleaf species benefit at the expense of and many do not rule out the possibility that overyielding needleleaf species (Mård 1996, Valkonen and Valsta 2001, simply reflects unknown niche differences between species Pretzsch 2005), as we found in lowland forests, the evidence that happen to belong to different functional groups. Yet, is by no means consistent across studies. However, the lack some studies do identify a mechanism that causes positive of consistent evidence for asymmetric interactions is not interactions between the two species, but often it is not nec- surprising, given that pairwise interactions are generally essarily related to the fact that one is a gymnosperm and the expected to alleviate intraspecific competition, potentially other is an angiosperm, such as differences in shade toler- offsetting any negative effects (Fig. 5b), which would not be ance (Pretzsch 2005). Thus, it remains uncertain why mixing evident without using methods such as those presented here. gymnosperms and angiosperms increases productivity under Thus, it also remains uncertain why angiosperms benefit at certain circumstances. the expense of gymnosperms under certain conditions.

1796 Leaf litter quality has a profound influence on decom- Darwin’s abominable mystery (Bond 1989, Berendse and position and nutrient cycling (Hättenschwiler 2005, Scheffer 2009). Cornwell et al. 2008), and the quality of needle and broadleaf litter could not be more different. Needleleaf litter The need for experiments generally has higher lignin concentration, and lower concen- trations of N and P, which are released more slowly because We sought to generalize the results of pairwise replacement the rate of decomposition is roughly half that of broadleaf studies by examining the competitive interactions observed litter (Cornwell et al. 2008, Augusto et al. 2015). Thus, at in an inventory dataset that encompasses a broad variety of first glance it seems reasonable that litter quality could medi- species, functional groups, and climatic conditions. While ate the interactions between gymnosperms and angiosperms our regression-based method revealed that complementar- through its effect on soil resources. For example, if litter mix- ity arises from dissimilarity effects that are both asymmet- ing effects were strictly additive (Ball et al. 2008), then the ric and non-reciprocal, it does not allow us to elucidate the addition of broadleaf litter would increase the rate of nutri- underlying mechanisms. Furthermore, it would be difficult ent cycling in needleleaf stands, which is consistent with our to separate the effect of variables that co-vary along the observation that broadleaf species ameliorate the suppres- altitudinal gradient, such as temperature and soil resource sion of needleleaf species in montane forests. However, the availability. Thus, there is a need to conduct controlled addition of needleleaf litter would slow the rate of nutrient biodiversity–productivity experiments to study overyield- cycling in broadleaf stands, which is inconsistent with our ing in gymnosperm–angiosperm mixtures, similar to those observation that needleleaf species ameliorate the suppression conducted in seasonal tropical forests that exhibit temporal of broadleaf species. niche complementarity between evergreen and deciduous Yet, it is possible that non-additive litter mixing effects angiosperms (Sapijanskas et al. 2014). While these experi- could promote complementary interactions between gym- ments suffer their own limitations, due to the long life span nosperms and angiosperms (Gessner et al 2010). For of trees and slow dynamics of forests, using high-density example, microbes and detritivores can make complemen- mixtures holds promise for elucidating the mechanisms tary use of resources by obtaining different resources from underlying the dissimilarity effects identified in this study low- and high-quality litter, particularly when nutrients are (Tobner et al. 2014). transported between the two substrates via fungal hyphae (Schimel and Hättenschwiler 2007). This complementary Acknowledgements – JC would like to thank Hongcheng Zeng for use of resources can accelerate decomposition and nutrient assistance with processing data. turnover in mixed litter (Hättenschwiler 2005), which may Funding – JC was funded by a Sabbatical Fellowship from the Swiss in turn promote complementarity between gymnosperms Fed. Res. Inst. WSL, as well as a Discovery grant from NSERC and angiosperms, provided they both benefit from the non- (RGPIN-2015-04452). additive litter mixing effect. Our results suggest that they Author contributions – JC conceived the ideas and designed both benefit from the non-additive litter mixing effect in methodology, in collaboration with all the other authors. ET montane, boreal-type forests, whereas angiosperms benefit at compiled the data. JC analyzed the data, with critical input from all the expense of gymnosperms in lowland temperate forests. the other authors. JC led the writing of the manuscript, with critical While positive non-additive effects on decomposition are input from all the other authors. widely observed in litter-mixing experiments, negative and neutral interactions have been reported as well, and there is no consensus regarding mixtures of needle and broadleaf References litters (Hättenschwiler 2005, Gessner et al 2010). Thus, it Augusto, L. et al. 2015. Influences of evergreen gymnosperm and remains uncertain whether litter quality mediates the inter- deciduous angiosperm tree species on the functioning of actions between gymnosperms and angiosperms through its temperate and boreal forests. – Biol. Rev. 90: 444–466. effect on soil resources. Furthermore, it remains unclear why Ball, B. A. et al. 2008. Consequences of non-random species loss mixing litters and accelerating nutrient turnover would only for decomposition dynamics: experimental evidence for additive favor angiosperms in lowland forests. and non-additive effects. – J. Ecol. 96: 303–313. However, it is widely recognized that broadleaf angio- Bates, D. et al. 2015. Fitting linear mixed-effects models using sperms are able to acquire and recycle resources faster than lme4. – J. Stat. Software 67: 1–48. needleleaf gymnosperms, whose conservative use of resources Berendse, F. and Scheffer, M. 2009. The angiosperm radiation allows them to dominate in less productive environments revisited, an ecological explanation for Darwin’s ‘abominable mystery’. – Ecol. Lett. 12: 865–872. (Grime 2006, Coomes et al. 2005). Thus, it is possible that Bertness, M. D. and Callaway, R. 1994. Positive interactions in accelerating nutrient turnover only favors angiosperms in communities. – Trends Ecol. Evol. 9: 191–193. lowland forests because this advantage is only realized in more Binkley, D. et al. 1992. Biogeochemistry of adjacent conifer and productive environments. Indeed, the fact that angiosperms alder–conifer stands. – Ecology 73: 2022–2033. profit from increased nutrient turnover has been recently Brassel, P. and Lischke, H. (eds) 2001. Swiss National Forest advanced to explain the historical displacement of gymno- Inventory: methods and models of the 2nd assessment. sperms from productive environments, otherwise known as Birmensdorf, Swiss Federal Res. Inst. WSL.

1797 Bond, W. J. 1989. The tortoise and the hare: ecology of angiosperm Kunstler, G. et al. 2012. Competitive interactions between forest dominance and gymnosperm persistence. – Biol. J. Linn. Soc. trees are driven by species’ trait hierarchy, not phylogenetic or 36: 227–249. functional similarity: implications for forest community Brändli, U.-B. (ed) 2010. Schweizerisches Landesforstinventar. assembly. – Ecol. Lett. 15: 831–840. Ergebnisse der dritten Erhebung 2004–2006. Birmensdorf Kunstler, G. et al. 2016. Plant functional traits have globally Eidgenössische Forschungsanstalt für Wald, Schnee und consistent effects on competition. – Nature 529: 204–207. Landschaft WSL. Bern, Bundesamt für Umwelt, BAFU. Levine, J. M. 2016. Ecology: a trail map for trait-based studies. Cameron, A. D. and Watson, B. A. 2000. Growth and wood – Nature 529: 163–164. properties of Sitka spruce (Picea sitchensis (Bong.) Carr.) in Lu, H. et al. 2016. Overyielding of temperate mixed forests occurs nursing mixtures established on nitrogen-deficient mineral in evergreen–deciduous but not in deciduous–deciduous species soils. – Scand. J. For. Res. 15: 237–246. mixtures over time in the Netherlands. – For. Ecol. Manage. Cannell, M. G. et al. 1992. The ecology of mixed-species stands of 376: 321–332. trees. – Blackwell. MacArthur, R. and Levins, R. 1967. The limiting similarity, Chabot, B. F. and Hicks, D. J. 1982. The ecology of leaf life spans. convergence, and divergence of coexisting species. – Am. Nat. – Annu. Rev. Ecol. Syst. 13: 229–259. 101: 377–385. Coomes, D. A. et al. 2005. The hare, the tortoise and the crocodile: Madrigal-González, J. et al. 2016. Complementarity effects on tree the ecology of angiosperm dominance, conifer persistence and growth are contingent on tree size and climatic conditions fern filtering. – J. Ecol. 93: 918–935. across Europe. – Sci. Rep. 6: 32233. Cornwell, W. K. et al. 2008. Plant species traits are the predominant Mård, H. 1996. The influence of a birch shelter (Betula spp.) on control on litter decomposition rates within biomes worldwide. the growth of young stands of Picea abies. – Scand. J. For. Res. – Ecol. Lett. 11: 1065–1071. 11: 343–350. Dănescu, A. et al. 2016. Structural diversity promotes productivity Mayfield, M. M. and Levine, J. M. 2010. Opposing effects of of mixed, uneven-aged forests in southwestern Germany. competitive exclusion on the phylogenetic structure of – Oecologia 182: 319–333. communities. – Ecol. Lett. 13: 1085–1093. Forrester, D. I. and Pretzsch, H. 2015. Tamm review: on the Mason, W. L. and Connolly, T. 2013. Mixtures with spruce species strength of evidence when comparing ecosystem functions of can be more productive than monocultures: evidence from the mixtures with monocultures. – For. Ecol. Manage. 356: 41–53. Gisburn experiment in Britain. – Forestry 87: 209–217. Forrester, D. I. and Bauhus, J. 2016. A review of processes behind Maestre, F. T. et al. 2009. Refining the stress gradient hypothesis diversity–productivity relationships in forests. – Curr. For. Rep. for competition and facilitation in plant communities. – J. Ecol. 2: 45–61. 97: 199–205. Forrester, D. I. et al. 2006. Mixed-species plantations of Eucalyptus Mina, M. et al. 2017. Multiple factors modulate tree growth with nitrogen-fixing trees: a review. – For. Ecol. Manage. 233: complementarity in central European mixed forests. – J. Ecol. 211–230. 106: 1106–1119. Gessner, M. O. et al. 2010. Diversity meets decomposition. Morgan, J. L. et al. 1992. Nitrogen relations of mixed-species – Trends Ecol. Evol. 25: 372–380. stands on oligotrophic soils. – In: Morgan, J. L. et al. (eds), The Givnish, T. J. 2002. Adaptive significance of evergreen vs deciduous ecology of mixed-species stands of trees. Blackwell, pp. 65–86. leaves: solving the triple paradox. – Silva Fenn. 36: 703–743. Mouillot, D. et al. 2011. Functional structure of biological Grime, J. P. 2006. Plant strategies, vegetation processes, and communities predicts ecosystem multifunctionality. – PloS One ecosystem properties. – Wiley. 6: e17476. Hättenschwiler, S. 2005. Effects of tree species diversity on litter Paquette, A. and Messier, C. 2011. The effect of biodiversity on quality and decomposition. –In: Scherer-Lorenzen, M. et al. tree productivity: from temperate to boreal forests. – Global (eds), Forest diversity and function. Springer, pp. 149–164. Ecol. Biogeogr. 20: 170–180. Ishii, H. and Asano, S. 2010. The role of crown architecture, leaf Potter, K. M. and Woodall, C. W. 2014. Does biodiversity make a phenology and photosynthetic activity in promoting comple- difference? Relationships between species richness, evolutionary mentary use of light among coexisting species in temperate diversity, and aboveground live tree biomass across US forests. forests. – Ecol. Res. 25: 715–722. – For. Ecol. Manage. 321: 117–129. Jones, H. E. et al. 2005. Functioning of mixed-species stands: Pretzsch, H. 2005. Diversity and productivity in forests: evidence evidence from a long-term forest experiment. – In: Scherer- from long-term experimental plots. – In: Scherer-Lorenzen, Lorenzen, M. et al. (eds), Forest diversity and function. Springer, M. et al. (eds), Forest diversity and function. Springer, pp. 111–130. pp. 41–64. Jucker, T. et al. 2015. Crown plasticity enables trees to optimize Pretzsch, H. 2014. Canopy space filling and tree crown morphology canopy packing in mixed-species forests. – Funct. Ecol. 29: in mixed-species stands compared with monocultures. – For. 1078–1086. Ecol. Manage. 327: 251–264. Jucker, T. et al. 2016. Climate modulates the effects of tree diversity Pretzsch, H. and Schütze, G. 2009. Transgressive overyielding on forest productivity. – J. Ecol. 104: 388–398. in mixed compared with pure stands of Norway spruce and Kikuzawa, K. 1991. A cost-benefit analysis of leaf habit and leaf European beech in central Europe: evidence on stand level and longevity of trees and their geographical pattern. – Am. Nat. explanation on individual tree level. – Eur. J. For. Res. 128: 138: 1250–1263. 183–204. Knoke, T. et al. 2008. Admixing broadleaved to coniferous tree Pretzsch, H. and Schütze, G. 2016. Effect of tree species mixing on species: a review on yield, ecological stability and economics. the size structure, density and yield of forest stands. – Eur. J. – Eur. J. For. Res. 127: 89–101. For. Res. 135: 1–22.

1798 Pretzsch, H. et al. 2015. Growth and yield of mixed versus pure Tobner, C. M. et al. 2014. Advancing biodiversity–ecosystem func- stands of Scots pine (Pinus sylvestris L.) and European beech tioning science using high-density tree-based experiments over (Fagus sylvatica L.) analysed along a productivity gradient functional diversity gradients. – Oecologia 174: 609–621. through Europe. – Eur. J. For. Res. 134: 927–947. Toïgo, M. et al. 2015. Overyielding in mixed forests decreases with Pretzsch, H. et al. 2017. Mixed-species forests: ecology and site productivity. – J. Ecol. 103: 502–512. management. – Springer. Turc, L. 1961. Evaluation des besoins en eau d’irrigation, Reich, P. B. et al. 2004. Species and functional group diversity évapotranspiration potentielle, formule simplifié et mise à jour. independently influence biomass accumulation and its – Ann. Agron. 12: 13–49.

response to CO2 and N. – Proc. Natl Acad. Sci. USA 101: Uriarte, M et al. 2010. Trait similarity, shared ancestry and the 10101–10106. structure of neighbourhood interactions in a subtropical wet Richards, A. E. et al. 2010. The influence of mixed tree plantations forest: implications for community assembly. – Ecol. Lett. 13: on the nutrition of individual species: a review. – Tree Physiol. 1503–1514. 30: 1192–1208. Valkonen, S. and Valsta, L. 2001. Productivity and economics of Rothe, A. and Binkley, D. 2001. Nutritional interactions in mixed mixed two-storied spruce and birch stands in southern Finland species forests: a synthesis. – Can. J. For. Res. 31: 1855–1870. simulated with empirical models. – For. Ecol. Manage. 140: Sapijanskas, J. et al. 2014. Tropical tree diversity enhances light 133–149. capture through crown plasticity and spatial and temporal niche van Ommen Kloeke, A. E. E. et al. 2012. Global quantification of differences. – Ecology 95: 2479–2492. contrasting leaf life span strategies for deciduous and evergreen Scherer-Lorenzen, M. et al. 2005. Forest diversity and function: species in response to environmental conditions. – Global Ecol. temperate and boreal systems, vol. 176. – Springer. Biogeogr. 21: 224–235. Schimel, J. P. and Hättenschwiler, S. 2007. Nitrogen transfer Wu, X. et al. 2015. The relationship between species richness and between decomposing leaves of different N status. – Soil Biol. biomass changes from boreal to subtropical forests in China. Biochem. 39: 1428–1436. – Ecography 38: 602–613. Thornton, P. E. et al. 1997. Generating surfaces of daily meteoro- Zimmermann N. E. and Kienast F. 1999. Predictive mapping of logical variables over large regions of complex terrain. – J. alpine grasslands in Switzerland: species versus community Hydrol. 190: 214–251. approach. – J. Veg. Sci. 10: 469–482.

Supplementary material (available online as Appendix oik- 05360 at < www.oikosjournal.org/appendix/oik-05360 >). Appendix 1–5.

1799