Tree diversity effects on forest productivity increase through time because of spatial partitioning

Background Experimental manipulations of tree diversity have often found overyielding in mixed-species plantations. While most experiments are still in the early stages of stand development, the impacts of tree diversity are expected to accumulate over time. Here, I present findings from a 31-year-old tree diversity experiment (as of 2018) in Japan. Results I find that the net diversity effect on stand biomass increased linearly through time. The species mixture achieved 64% greater biomass than the average monoculture biomass 31 years after planting. The complementarity effect was positive and increased exponentially with time. The selection effect was negative and decreased exponentially with time. In the early stages (≤3 years), the positive complementarity effect was explained by enhanced growths of early- and mid-successional species in the mixture. Later on (≥15 years), it was explained by their increased survival rates owing to vertical spatial partitioning — i.e., alleviation of self-thinning via canopy stratification. The negative selection effect resulted from suppressed growths of late-successional species in the bottom layer. Conclusions The experiment provides pioneering evidence that the positive impacts of diversity-driven spatial partitioning on forest biomass can accumulate over multiple decades. The results indicate that forest biomass production and carbon sequestration can be enhanced by multispecies afforestation strategies.

Within each plot, trees were laid out hexagonally with 50-cm spacing, resulting in a density of 91 48,000 trees·ha −1 . In the mixture plot, trees were planted such that their closest neighbors were 92 always heterospecific (Fig. 1). The initial heights of all trees were set equal to ca. 35 cm by using    , 1988, 1989, 1990, 2002, 2003, 2009,   values averaged across all the species. The first and second components of Eq. 1 represent the 133 complementarity and selection effects, respectively.

134
The temporal changes in net diversity, complementarity, and selection effects (Y) across the 135 experiment year (X) were analyzed by fitting linear models = α + β · and exponential 136 models = α · β to the data. I used exponential models because Y can accumulate over time in which 21 trees were not measured (thus n=307) (see Table S1). Tree heights among treatments 153 and species in each year were analyzed by means of two-way ANOVA followed by Tukey's 154 multiple comparison tests. The treatment, species identity, and their interaction were included as 155 explanatory variables. See Table S1 for sample sizes. All statistical analyses were conducted using 156 R 3.5.2 (R Core Team 2018).

159
The net diversity effect and complementarity effect showed positive values and tended to 160 increase across the experiment year (Fig. 2a). The selection effect showed negative values after 161 the first year and decreased thereafter (Fig. 2a). The linear model had a lower AIC than the 162 exponential model for the net diversity effect (R 2 = 0.91), while the exponential models were  (Table S2). State space models accounting for temporal autocorrelations also showed that 188 the net diversity and complementarity effects increased over time while the selection effect 189 decreased in negative direction (Table S3). The survival rates of individual trees did not differ 190 significantly between the monoculture and mixture plots during the first three years (Fig. 2b, Table   191 S4). From the 15th year onwards, however, trees in the mixture showed higher survival rates than 192 those of the same species in the monocultures (Fig. 2b, Table S4). Interspecific differences in 193 survival rates became significant from the second year onwards (Table S4). The survival rates of  Table   196 S1). Trees that were shorter in height tended to show lower survival rates than taller trees of the 197 same species (Fig. S1).

198
Three years after the onset of the experiment, B. maximowicziana and Q. crispula 199 monocultures showed larger AGB than the average monoculture AGB (Fig. 3a). While B.
200 maximowicziana showed the largest AGB among the monocultures, it accounted for a relatively 201 small proportion of the total AGB in the mixture (Fig. 3a). This explains the negative selection 202 effect in the third year (Fig. 2a). Fifteen years after planting, A. sachalinensis showed larger AGB 203 than the average monocultue AGB, but constituted for the smallest proportion in the mixture (Fig.   204 3a). This discrepancy became even more pronounced 31 years after planting, explaining the 205 escalating, negative selection effect aross time (Fig. 3a). On the other hand, B. maximowicziana 206 and Q. crispula monocultures showed relatively low AGB after 15 years, and this trend became 207 even clearer after 31 years (Fig. 3a). Their relative AGB were, however, comparatively higher in 208 the mixture than in the monoculture (Fig. 3a). This explains the increasing complementarity effect 209 through time (Fig. 2a). studies also showed that the tree diversity effect increased over the first eight years of experiment 236 (Huang et al. 2018). Adding on to these earlier findings, the Furano experiment showed that the 237 species mixture achieved 64% greater biomass than the monocultures 31 years after planting (Fig.   238 3a). It also revealed that the complementarity effect increased more largely than has the selection 239 effect decreased in negative direction with time (Fig. 2a). These results indicate that the impacts 240 of species mixing can accumulate over time at least in some forest ecosystems, encouraging 241 multispecies afforestation strategies to enhance long-term biomass production and carbon 242 sequestration.

243
During the first three years of the experiment, early-and mid-successional species (B. 244 maximowicziana and Q. crispula) grew significantly larger in the mixture than in the monocultures 245 (Fig. 3b), a result which was attributable to increased availability of canopy space (Williams et al. 246 2017). The survival rates, on the other hand, did not differ significantly between the mixture and 247 monocultures (Fig. 2b). These results coincide with previous findings in young tree stands that shifted from tree growth to survival. Trees in the mixture showed higher survival rates than those 251 of the same species in the monocultures (Fig. 2b). This can be potentially explained by the 252 reduction in competition-induced mortality owing to vertical partitioning -i.e., alleviation of 253 self-thinning via canopy stratification (Fig. 3b). Note here that the vertical partitioning per se 254 unlikely resulted from overyielding, but rather from the inherent differences in species' 255 successional niches, namely maximum growth rates under sufficient resources. I also found that 256 the complementarity effect became increasingly large with time (Fig. 2a). This was likely because 257 the intraspecific competition in each canopy layer intensified as crowns enlarged, and this impact 258 was especially strong in the monocultures that had a single layered canopy (Fig. 3b).

259
The late-successional species (A. sachalinensis) achieved the highest monoculture biomass 260 after ≥15 years (Fig. 3a). By contrast, in the mixture, its total biomass remained significantly small 261 (Fig. 3a), which led the selection effect to take negative values (Fig. 2a). The low total biomass of 262 A. sachalinensis was attributable to their suppressed growths (Fig. 3b), and was not explained by 263 their survival rate, which was higher in the mixture than in the monoculture (Fig. 2b). In a and Q. crispula (Fig. 2b) to some extent. It is also possible that the positive impacts of species 281 mixing on forest biomass growth (Fig. 1a)  tended to show lower survival rates than taller trees of the same species (Fig. S1). Taken together, 287 it is likely that disturbance played minor roles compared to light competition in the Furano 288 experiment. I should, however, stress that this interpretation requires future verifications, given the 289 fact that the experiment lacks plot replication and disturbance-related data. that the experiment will enter the next stage of stand development in the coming decades, where 304 the early-successional species would die out and the remaining species take over the canopy space.

305
The experiment would continue to provide, together with other tree diversity experiments, 306 irreplaceable opportunities for long-term BEF research in forests.