Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Supporting information Appendix S1 | Estimating crown volume of individual trees ................................................ 2 Appendix S2 | Predicted canopy packing .............................................................................. 5 Appendix S3 | Canopy packing models ................................................................................. 6 Accounting for non-target species ..................................................................................... 6 Model selection .................................................................................................................. 6 Appendix S4 | Crown expansion and light interception ........................................................ 9 Crown vertical and lateral expansion ................................................................................. 9 Vertical stratification in species tree heights ................................................................... 10 Relating crown volume and light interception ................................................................. 10 Appendix S5 | The influence of shade tolerance on canopy packing .................................. 12 [1] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Appendix S1 | Estimating crown volume of individual trees The volume of a tree’s crown can be estimated in the same way that the volume of a solid of revolution is calculated. For a given curve determined by an equation of the form y = αxβ, rotating the curve around the x axis will outline a solid whose shape will depend on β (Fig. S1). If β = 1, the solid will take the shape of a cone, β < 1 results in convex shape (i.e., paraboloid) which becomes increasingly cylindrical as β approaches 0, while β > 1 outlines a concave object (i.e., neiloid). Fig. S1 | Schematic representation of the crown profile of a tree, where H corresponds to the height at the top of the tree, CD is the depth of the crown, CRmax is the crown’s maximum radius (assumed to be at the crown’s base) and CRh is the crown radius at a given height h from the base of the crown. The volume of the crown is estimates in the same way as that of a solid of revolution, where the crown profile is rotated around the height axis to give a 3D object. The shape of the crown is determined by a function of the form y = αxβ which describes how crown radius changes along the height of the crown (red section of the crown profile). The volume of the solid between points a and b along the x axis is given by: π ππππ’ππ = ∫π π π¦ 2 ππ₯ (1) In the same way, a tree’s crown volume (CV) can be approximated by: 2 π» πΆπ = ∫π»−πΆπ· π (π(β)) πβ (2) where H is the height of the tree, CD the depth of the crown, and f(h) describes how the tree’s crown radius (CRmax) changes along the height of the crown. The function which predicts the crown radius at a given height h along the crown (CRh) was obtained from Caspersen et al. (2011) and is: πΆπ β = πΆπ πππ₯ ( πΆπ·−β π½ πΆπ· ) (3) [2] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes where CRmax is the crowns maximum radius, which for the purposes of this study is assumed to be at the base of the crown, and β is the shape parameter which determines the curvature of the crown. At the base of the crown h = 0, and therefore CRh = CRmax, while at the top of the crown h = CD and CRh = 0. Substituting f(h) with equation (3) gives: 2 πΆπ·−β π½ π» πΆπ = π ∫π»−πΆπ· (πΆπ πππ₯ ( πΆπ· ) ) πβ (4) Deriving equation (4) gives: π» πΆπ·−β 2π½ 2 πΆπ = π πΆπ πππ₯ ∫π»−πΆπ· ( πΆπ· ) πβ β πΆπ = π πΆπ = − πΆπ = − πΆπ = − πΆπ = πΆπ·(1− ) π» πΆπ· 2 πΆπ πππ₯ ∫π»−πΆπ· (− 2π½+1 2 π πΆπ πππ₯ πΆπ· 2π½+1 2 π πΆπ πππ₯ πΆπ· 2π½+1 2 π πΆπ πππ₯ πΆπ· 2π½+1 β ) πβ (6) ] (7) π»−πΆπ· 2π½+1 ((1 − πΆπ·) πΆπ· 2π½+1 (− (πΆπ·) 2π½+1 2π½+1 π» [(1 − πΆπ·) π» (5) (1 − ) ≡ − 2 π πΆπ πππ₯ πΆπ· π»−πΆπ· 2π½+1 πΆπ· ) 2 π πΆπ πππ₯ πΆπ· 2π½+1 ) (8) (−1) (9) (10) 2π½+1 Estimating the shape parameter β requires measuring a tree’s crown radius at successive height intervals in order to reconstruct its crown profile. As these data were not available for our study species, we instead relied on published estimates of β available for 250 North American tree species (Purves et al. 2007). We first filtered the species list by excluding taxa that belong to a different genus to those found in our study. We then grouped the remaining species into conifers and broadleaves, and averaged their β estimates to obtain mean values for each of the two functional groups. Details on how to calculate a species’ β value can be found in the supporting information of Purves et al. (2007). Ideally we would have preferred to directly measure both the inter- and intra-specific variation in β (as opposed to using published estimates). However, this would have required taking measurements of crown radius at multiple heights along the crown for each tree, something which we were unable to do given the large sample sizes of the present study (≈ 13000 trees). Nonetheless, we expect equation (10) to provide an adequate approximation of each trees’ CV, as most of the inter-tree variation in CV is driven by the numerator of the equation (i.e., CRmax and CD, both of which were measured for all trees). Evidence of this can be seen when looking at the distribution of β values reported in Purves et al. (2007), where [3] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes 95% of species have β values which range between 0.26 to 0.44 (calculations based on the subset of species used to estimate β for conifers and broadleaves in this study). Using equation (10) we can compare the effect of the variability in β on CV estimates versus the effect of inter- and intraspecific variation in CRmax and CD (e.g., values reported in Table S4). Even when assuming the extremes of the distribution in β values (i.e, differences in CV when β = 0.26 vs β = 0.44), the effect of β on CV is much smaller than that of inter- and intraspecific variation in either CRmax or CD (Fig. S2). Fig. S2 | Comparison of the effects of variation in β on crown volume versus that of variation in (a) crown radius and (b) crown depth. Equation (10) was used to generate crown volume estimates in which (a) crown radius was allowed to vary within the 95% range of observed values in the data while crown depth was kept at the mean (9.5 m) and (b) in a scenario in which crown depth varied along the 95% range of the data and crown radius was kept at the mean (2.4 m). This was repeated for β values at both extremes of the spectrum (β = 0.26 in red and β = 0.44 in black). In both cases, variation in crown dimensions had a much larger impact on crown volume than variation in β. Caspersen, J. P., Vanderwel, M. C., Cole, W. G. & Purves, D. W. (2011) How stand productivity results from size- and competition-dependent growth and mortality. PLoS ONE, 6, e28660. Purves, D. W., Lichstein, J. W. & Pacala, S. W. 2007 Crown plasticity and competition for canopy space: a new spatially implicit model parameterized for 250 North American tree species. PLoS ONE, 2, e870. [4] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Appendix S2 | Predicted canopy packing To test whether canopy packing increases with diversity as a consequence of vertical stratification or as a result of changes in crown allometries at the species level, we used field data to develop regression models aiming to explain the primary drivers of changes in CV among trees (Table S1). Table S1 | Comparison of regression models aiming to explain variation in CV among trees. The AIC of the best fitting model (M4) is shown in bold. Species richness and plot basal area are abbreviated to SR and BA, respectively. An interaction term was fitted in order to allow CV~DBH relationships to vary among species. Model Structure M0 M1 M2 M3 M4 log (CV) ~ log (DBH) AIC log (CV) ~ log (DBH) x Species identity log (CV) ~ log (DBH) x Species identity + BA log (CV) ~ log (DBH) x Species identity + SR log (CV) ~ log (DBH) x Species identity + BA + SR 30400 22152 22015 22045 21900 We found that tree size explains much of the variation in CV among trees (R2=0.81 for M1 in Table S1). Based on the relationship between CV and DBH, we predicted the CV of each tree in the dataset, using these values to obtain a measure of predicted canopy packing at the plot level. Predicted canopy packing values were strongly related to those obtained from field measurements (Fig. S3a). However, we found that the influence of species richness of canopy packing weakened when considering predicted as opposed to observed values (see main text and Fig. S3b). The reason for this is that trees exhibit crown plasticity, tending to have larger crown in mixed stands and decreased CV in plots with high basal area. The strong influence of both species richness and basal area on CV is confirmed by the fact that models which account for these two factors significantly outperform ones that do not (Table S1). Fig. S3 | Predicted vs observed canopy packing values. In panel (a) the relationship is shown for each country, with the dashed line corresponding to a 1:1 relationship. In panel (b) separate lines are fitted to plots in each species richness level, highlighting a progressively larger difference between observed and predicted canopy packing (i.e., observed > predicted) when going from monocultures to mixtures with 4 or more species. [5] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Appendix S3 | Canopy packing models Accounting for non-target species Canopy packing models were fitted to a subset of the total FunDivEUROPE plot network (199 of the 209 plots). Ten plots were excluded from the analyses as they contained too high a proportion of non-target trees within the stand (Fig. S4): 6 in Poland, 3 in Germany and 1 in Romania. The criteria used to exclude plots was based on proportion of total canopy volume belonging to non-target trees (>20% in excluded plots). Plots with a high proportion of nontarget trees were excluded as we did not have sufficient allometric data to model the crown volume of non-target species. This precluded us from robustly teasing apart vertical stratification effects from those of crown plasticity for these plots, as this requires substituting measured CV with predicted values from models to tease out the effects of crown plasticity. Nonetheless it should be noted that analyses performed on the entire dataset led to almost identical results to those performed on the subset of the data which we present in the main document. Fig. S4 | Scatter plot of canopy volume estimates for each plot (in 1000 m3) based on all trees within the stand and only accounting for target species. Dashed line corresponds to a 1:1 relationship. Points falling to the right of red line (n = 10) correspond to plots in which non-target species make up > 20% of the total canopy volume. These were excluded from the analyses presented in the main document. Model selection To determine the importance of species richness as a predictor of canopy packing and to gauge the appropriate level of model complexity, a number of alternative mixed effects models of canopy packing were fitted and compared using AIC (Table S2). Models included different combinations of three predictors: species richness, plot basal area (to account for the fact that plots with a greater density of trees are expected to have greater total canopy volume) and country (to account for the fact that baseline canopy packing values vary among forest types; treated a random effect in the model). [6] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Table S2 | Comparison of mixed-effects models of canopy packing. The AIC of the best fitting model (M7) is shown in bold. Species richness and plot basal area are abbreviated as SR and BA, respectively. Model structure follows lme4 syntax in R, where (1 | Country) indicates that country is included as a random effect, which is normally distributed and affects the intercept, while (SR | Country) corresponds to a varying slope and intercept model (i.e., testing whether the effects of species richness on canopy packing varies among countries). Model M0 M1 M2 M3 M4 M5 M6 M7 M8 Structure logit (Canopy packing) ~ (1 | Country) logit (Canopy packing) ~ SR + (1 | Country) logit (Canopy packing) ~ BA + (1 | Country) logit (Canopy packing) ~ SR + BA + (1 | Country) logit (Canopy packing) ~ SR + (SR | Country) logit (Canopy packing) ~ SR + BA + (SR | Country) logit (Canopy packing) ~ BA + (BA | Country) logit (Canopy packing) ~ SR + BA + (BA | Country) logit (Canopy packing) ~ SR + BA + (BA | Country) + (SR | Country) AIC 369.8 365.0 335.7 331.5 368.9 335.5 318.7 309.4 315.4 We chose basal area as a descriptor of stand structure in the models as it correlates strongly with both the mean size of trees within plots and the total number of stems within a stand (Fig. S5). Fig. S5 | Scatter plots of (a) stand basal area and quadratic mean stem diameter (ρ = 0.74; P < 0.0001) and (b) quadratic mean stem diameter and total number of stems within a plot (ρ = –0.77; P < 0.0001). Relationships are plotted on a logarithmic scale. For a given level of complexity (e.g., M1 vs M0), models which included a species richness term always outperformed ones that did not. However, models which included a random slope term for species richness were not supported (i.e., M4 vs M1 and M8 vs M7). This suggests that although canopy packing values vary among forest types, the effect of species richness on canopy packing does not (Fig. S6). [7] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Fig. S6 | Relationship between canopy packing (logit transformed) and species richness for each site. Fitted relationships were obtained from a mixed-effects model in which the effect of species richness on canopy packing was allowed to vary across sites (i.e., random slope effect). Variation in slope among sites was minimal in the model (± 0.02 from the overall slope of 0.12). Overall, the best performing model (the results of which are presented in the main text) included both the effects of species richness and plot basal area on canopy packing, with the effect of plot basal area on canopy packing varying among countries (M7). Species richness and plot basal area were only weakly correlated (Fig. S7), suggesting that collinearity should not affect parameter estimates in the models. Fig. S7 | Relationship between plot basal area and species richness across the entire plot network (black fitted line), and for each site separately (see figure legend for colour coding). Pearson’s correlation coefficients (ρ) between plot basal area and species richness, both within and among countries, were weak: ρ = 0.13 across all plots, with a minimum in Spain (-0.16), and a maximum in Poland (0.21) . [8] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Appendix S4 | Crown expansion and light interception Crown vertical and lateral expansion Increased crown volume (CV) can be the result of one of two processes: trees can either expand their crown laterally by increasing their crown radius (CRmax) and/or vertically by increasing their crown depth (CD). To determine whether changes in CV where primarily driven by changes in CRmax or CD, we modelled the two crown architectural components separately. In addition to this, we also tested whether tree height (H) – diameter (DBH) scaling relationships varied in response to species richness: log (CRmax) ~ log (DBH) + SR + BA + (log (DBH) | Species identity) (11) log (CD) ~ log (H) + SR + BA + (log (H) | Species identity) (12) log (H) ~ log (DBH) + SR + BA + (log (DBH) | Species identity) (13) Across species, both CRmax and CD increased significantly in response to species mixing, suggesting that increased crown volume of trees in mixture is the result of both a lateral and vertical expansion of the crown (Table S3). Interestingly, CD increased even though trees became significantly shorter (for a given diameter) when growing in mixture (Table S3). The fact that tree in mixture allocate less carbon to height growth vs diameter increment suggests that species mixing is alleviating competition for light among neighbouring trees (Jucker et al. 2014). In addition to species richness, plot basal area also had a strong influence on crown morphology (Table S3). In general, trees growing in densely packed stands are taller and have slimmer crowns that are less deep. Table S3 | Regression models testing the effects of species richness and plot basal area on crown radius, crown depth, tree height and crown light interception (equations 11-13 and 15). Equation Response variable 11 Crown radius 12 Crown depth 13 Tree height 15 Light interception Predictor Slope (CI95) Species richness Plot basal area Species richness Plot basal area Species richness Plot basal area Species richness 0.0351 (0.0057) -0.0035 (0.0007) 0.0282 (0.0058) -0.0092 (0.0007) -0.0252 (0.0040) 0.0068 (0.0005) 0.0664 (0.0175) Plot basal area -0.0166 (0.0022) [9] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Vertical stratification in species tree heights Table S4 | Mean and maximum height (H), crown depth (CD) and crown radius (CR) of each study species. Maximum values correspond to the 99th percentile of the data for each allometric measurement. Study site Finland Germany Italy Poland Romania Spain Species name Mean H Max H Mean CD Max CD Mean CR Max CR Betula pendula 18.1 Picea abies 16.3 24.8 9.7 16.7 2.3 4.0 24.3 11.5 19.5 2.2 3.5 Pinus sylvestris 17.8 24.5 8.0 12.9 2.2 3.5 Acer pseudoplatanus 20.8 33.9 9.6 23.0 2.6 7.2 Fagus sylvatica 21.1 37.0 12.8 27.1 3.2 7.0 Fraxinus excelsior 25.8 38.7 9.3 23.5 2.6 7.0 Picea abies 21.4 35.5 10.9 23.7 2.2 5.0 Quercus petraea 30.8 37.6 16.5 26.4 4.5 7.0 Castanea sativa 14.9 22.1 8.2 16.9 2.1 4.5 Ostrya carpinifolia 14.8 24.2 6.2 15.4 1.8 4.3 Quercus cerris 19.1 29.3 9.2 20.0 2.8 7.0 Quercus ilex 13.3 22.9 8.0 18.4 2.3 7.0 Quercus petraea 18.0 27.9 9.6 19.5 2.9 8.4 Betula pendula 31.8 38.3 14.1 21.6 3.8 7.1 Carpinus betulus 20.8 30.5 13.7 23.3 3.5 8.0 Picea abies 25.6 38.5 15.9 28.3 2.7 5.0 Pinus sylvestris 32.7 39.6 11.1 19.0 3.5 6.0 Quercus robur 29.6 40.3 15.4 24.5 4.0 8.1 Abies alba 26.7 40.1 11.5 26.4 2.1 4.2 Acer pseudoplatanus 26.3 35.5 11.2 20.3 2.5 5.0 Fagus sylvatica 25.4 39.9 12.9 24.5 2.7 6.0 Picea abies 28.3 43.5 14.6 30.5 2.1 4.0 Pinus nigra 11.3 19.5 5.9 11.7 2.4 5.0 Pinus sylvestris 12.8 21.2 5.6 11.4 2.2 6.6 Quercus faginea 7.5 14.0 4.5 9.8 1.4 4.0 Quercus ilex 4.8 10.4 3.1 6.5 1.3 3.3 Relating crown volume and light interception Crown volume is considered a good proxy of a tree’s ability to intercept light (Binkley et al. 2013). Nevertheless, even a tree with a large crown may exhibit slow growth if neighbouring trees cast it in shade. To determine whether crown volume estimates effectively capture the extent to which a tree is able to intercept light, we used field data to calculate a light interception (LI) index for each tree using the approach developed by King et al. (2005): πΏπΌ = πΆππ΄ × πΆπΌ 2 (14) where CPA is the crown projected area of each tree (in m2; calculated using the crown radius measurements taken in the field) and CI is the crown illumination index, which scores each [10] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes tree on a scale of 1 to 5 based on exposure to direct sunlight (Clark & Clark 1992). Suppressed crowns with no direct access to light receive a CI score of 1, trees exposed only to lateral light are assigned to class 2, trees experiencing overhead light on a portion of their crown are scored as 3, crowns with complete access to overhead light belong to class 4, while fully exposed dominant crowns are assigned to class 5. Crown volume was found to be a strong predictor of light interception (Fig. S8). Fig. S8 | Relationship between light interception and crown volume for all tree in the plot network (axes are on a log-log scale). Tree species are grouped by country. In addition to this, we also tested whether light interception varied in relation to species richness by fitting a linear mixed-effects model of: log (LI) ~ log (DBH) + SR + BA + (log (DBH) | Species identity) (15) As was found for crown volume, this revealed that light interception increased significantly in response to species richness (Table S3). Binkley, D., Campoe, O. C., Gspaltl, M. & Forrester. D. I. 2013 Light absorption and use efficiency in forests: Why patterns differ for trees and stands. Forest Ecol. Manag., 288, 5–13. Clark, D.A. & Clark, D.B. 1992 Life history diversity of canopy and emergent trees in a neotropical rain forest. Ecol. Monogr., 62, 315–344. Jucker, T., Bouriaud, O., Avacaritei, D., Danila, I., Duduman, G., Valladares, F. & Coomes, D. A. 2014 Competition for light and water play contrasting roles in driving diversity– productivity relationships in Iberian forests. J. Ecol., 102, 1202–1213. King, D. A., Davies, S. J., Nur Supardi, M. N. & Tan, S. 2005 Tree growth is related to light interception and wood density in two mixed dipterocarp forests of Malaysia. Func. Ecol., 19, 445–453. [11] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes Appendix S5 | The influence of shade tolerance on canopy packing To test whether mixtures which include species with a greater range of shade tolerance (i.e., mixtures of shade tolerant and light demanding species) are able to partition aboveground space most efficiently we calculated a metric of functional diversity in shade tolerance (FDshade) for each plot. To do so we made use of a continuous measure of shade tolerance developed by Niinemets & Valladares (2006) which has since been widely used in forest ecology (e.g., Kunstler et al. 2011). Estimates of shade tolerance were obtained for each species from Appendix A in Niinemets & Valladares (2006) and are reported in Table S5. Table S5 | Shade tolerance index for each species in the FunDivEUROPE plots (± standard errors) as reported in Appendix A of Niinemets & Valladares (2006). Low values identify light demanding species, while high values indicate species with a greater degree of shade tolerance. Estimates were available for all species with the exception of Quercus faginea (†), for which a genus-level mean was used. Species name Shade tolerance Abies alba Acer pseudoplatanus Betula pendula Carpinus betulus Castanea sativa Fagus sylvatica Fraxinus excelsior Ostrya carpinifolia Picea abies Pinus nigra Pinus sylvestris Quercus cerris Quercus faginea† Quercus ilex Quercus petraea Quercus robur 4.60 ± 0.06 3.73 ± 0.21 2.03 ± 0.09 3.97 ± 0.12 3.15 ± 0.23 4.56 ± 0.11 2.66 ± 0.13 3.94 ± 0.18 4.45 ± 0.50 2.10 ± 0.43 1.67 ± 0.33 2.55 ± 0.11 2.51 ± 0.07 3.02 ± 0.19 2.73 ± 0.27 2.45 ± 0.28 To calculate the diversity in shade tolerance of each plot (FDshade) we used Rao’s quadratic entropy index (de Bello et al. 2010). For a given plot, we calculated FDshade as: πΉπ·π βπππ = ∑πππ πππ ππ ππ (15) where S is the number of species in the plot (species richness), dij is the distance between the species pair i and j, weighted by the relative abundance pi and pj of the two species (estimated on the basis of relative basal area within a plot). The dissimilarity between species (dij) is quantified using a multivariate functional distance matrix based on Euclidean distance. FDshade was calculated using the dbFD function in the R package FD. [12] Crown plasticity enables trees to optimize canopy packing in mixed-species forests Tommaso Jucker, Olivier Bouriaud and David A. Coomes FDshade emerged a strong predictor of canopy packing (Fig. S9), with plots that included species with a greater range in shade tolerance exhibiting more efficient use of aboveground space (P = 0.0004; Slope and 95% confidence intervals = 0.50 ± 0.27). Fig. S9 | Relationship between canopy packing (logit transformed) and FDshade (a measure of the diversity in shade tolerance of species within a plot) across the FunDivEUROPE network. The red line corresponds to the fit of a linear model, with the shaded region marking the 95% confidence intervals of the regression line. Niinemets & Valladares. 2006 Tolerance to shade, drought, and waterlogging of temperate northern hemisphere trees and shrubs. Ecol. Monogr., 76, 521–547. de Bello, F., Lavergne1, S., Meynard, C.N., Lepš, J. & Thuiller, W. 2010 The partitioning of diversity: showing Theseus a way out of the labyrinth. J. Veg. Sci., 21, 992-1000. Kunstler et al. 2011. Effects of competition on tree radial-growth vary in importance but not in intensity along climatic gradients. J. Ecol., 99, 300–312. [13]