Evolution of species complementarity in response to drought in a grassland biodiversity experiment


 Growing threats from extreme climatic events and biodiversity loss have raised concerns about their interactive consequences for ecosystem functioning. Evidence suggests that biodiversity is crucial to buffer ecosystem functioning facing climatic extremes. However, whether evolutionary processes in species mixtures underpin such biodiversity-dependent stabilizing effects remains elusive. We tested this hypothesis by exposing experimental mixtures of grassland species to eight recurrent summer droughts vs. control in the field. Seed offspring of 12 species were subsequently grown individually, in monocultures or in 2-species mixtures and subjected to a novel drought event in the glasshouse. Comparing mixtures with monocultures, drought-selected plants showed greater between-species complementarity than ambient-selected plants when recovering from the drought event, which led to greater biodiversity effects on community productivity and better recovery of drought-selected mixtures after the drought. These findings suggest biodiversity can buffer the impacts of extreme climatic events through evolution of species complementarity.

between intolerant and tolerant species 38 (Supplementary Fig. 1a). Alternatively, 85 drought may exclude intolerant genotypes from both intolerant and tolerant species, 86 which may reduce the niche overlap and competition intensity between the two 87 species (Supplementary Fig. 1b). Second, trait expression in one species may 88 influence trait expression in another interacting species, leading to heritable changes 89 in species interactions 37 . That is, selection may occur at the community level and not  In the present study, we assessed how an 8-year selection treatment of summer 99 droughts affected plant diversity effects on ecosystem functioning and species 100 interactions before, during and after a novel drought event. We exposed experimental 101 mixtures of grassland species to eight recurrent summer droughts vs. control in a 102 long-term biodiversity experiment in the field (the Jena Experiment 23,46,47 ). Seed 103 offspring of 12 species were subsequently grown individually, in monocultures or in 104 2-species mixtures and subjected to a novel drought event in the glasshouse ( Fig. 1;  6 of two weeks without watering (drought conditions, "during drought") and (3) after a 110 third phase of seven weeks with regular watering (ambient conditions for recovery, 111 "after drought"). 112 We used the harvested aboveground biomass in four sets of analyses. First, we 113 assessed how drought selection affected biodiversity effects on productivity before, 114 during and after the drought event in the glasshouse. We used the biomass difference 115 between 2-species mixtures vs. monocultures to calculate the net biodiversity effects, 116 which we further partitioned into complementarity effects (indicating niche 117 differentiation or facilitation) and sampling effects (indicating a disproportionate 118 contribution to community productivity of one or the two species; which is also 119 referred as "selection effect" elsewhere, a term we did not use in this study to avoid 120 confusion) 48 . Second, we assessed the effects of drought selection on resistance 121 (biomass ratio during vs. before the drought), recovery (biomass ratio after vs. during 122 the drought) and resilience (biomass ratio after vs. before the drought) of productivity 123 19,20 . Third, we tested whether altered plant interactions drove the above differences 124 between the two selection treatments. We calculated the intensities of neighbor

136
Effects of drought selection on biodiversity effects on productivity 137 We found that the net biodiversity effects for different species pairs were higher when whereas ten had negative sampling effects, and six of these were statistically 146 significant (P < 0.05; Fig. 2r). In contrast, of eleven ambient-selected species pairs 147 with sufficient replication for separate statistical analyses, six had negative 148 complementarity effects and one of these was significant (P < 0.05; Fig. 2o); whereas 149 only two pairs had significantly negative sampling effects (P < 0.05; Fig. 2q).

150
When we used all species pairs in a combined statistical analysis, testing the 151 overall drought selection effect against the variation among species pairs as error 152 term, we found that drought selection had increased post-drought net biodiversity 153 effects in mixtures independent of the particular species pair considered and could 154 thus be generalized across the entire experiment ( Fig. 3; Supplementary Table 3). 155 Furthermore, drought-selected species pairs consistently had more positive 156 complementarity and more negative sampling effects than ambient-selected species selected plants in mixtures. However, before and during the drought, the effects of 160 drought selection were not statistically significant (Fig. 3; Supplementary Table 3). 161 Initially, the positively biodiversity effect was mainly due to positive sampling effect 162 (Fig. 3a,   The difference in recovery rates between mixtures and monocultures (i.e.,  Table 5). This is consistent 176 with the more positive biodiversity effects on productivity for drought-selected than 177 ambient-selected plants after the drought event described in the previous section.

178
However, biodiversity effects on resistance were more negative for drought-selected 179 than for ambient-selected plants, thus leading to similar biodiversity effects on 180 resilience between the two selection treatments (Fig. 4a, c).  Effects of drought selection on plant traits 208 We measured six traits that were closely related to plant water or carbon use on plants  Table 9). The drought event in the glasshouse reduced leaf 214 stomatal conductance ( Supplementary Fig. 5). However, leaf stomatal conductance 215 did not vary significantly between the two selection treatments, neither before nor 216 during the drought event ( Supplementary Fig. 5). These results suggest that the    We found that drought selection caused a significant difference (relative to  The selection for niche differentiation after the drought event could be related to 267 effects of drought on soil nutrients. For example, droughts can increase leaf litter 52 268 and reduce the mobility of soil nutrients, the activity of soil microbes and the rate of 269 litter decomposition 53-55 . These constrained resources can be released after droughts 270 33,34,54 , which increases the potential for niche partitioning among species. 271 We did not test the relative importance of genetic vs. epigenetic variation in 272 explaining our results, but know from a previous study that adaptive changes that had study, showed that more diverse communities were better able to compensate for 298 drought-driven productivity loss, which led to stabilizing effects of biodiversity. The 299 results from this study suggest that biodiversity-dependent evolutionary adaptation 300 may be an important mechanism driving the above compensatory recovery observed 301 in the field. Biodiversity effects on resistance were more negative for drought-selected 302 than for ambient-selected plants, thus leading to similar biodiversity effects on 303 resilience between the two selection treatments. These findings suggest that drought-304 selected plants competed more strongly between species than did ambient-selected    Table 1). However, including or excluding these communities 398 produced qualitatively similar results. Thus, we present the results including these two 399 species in this paper (Figs. 3, 4 and 6).

400
During a first phase of three month in the glasshouse (Fig. 1), pots were watered 401 regularly ("before drought"). After 14-18 weeks, when most of the species had 402 reached peak aboveground biomass, we harvested them (first harvest). This was 403 followed by a second phase of two weeks without watering ("during drought"). Soil 404 moisture decreased from more than 40% to less than 10% after 10 days since the 405 drought initiation. During a third phase of seven weeks, pots were watered regularly 406 again for recovery until most plants reached a new aboveground biomass peak again 407 ("after drought"). We conducted three harvests: 14-18 weeks (before the drought 408 event), 20 weeks (at the end of the drought event) and 27 weeks (after complete recovery from the drought event) after establishment of the experiment in the 410 glasshouse. At each harvest, we cut the plants at about 3 cm above soil and dried the 411 harvested plant material at 70°C for 48 hours to obtain the dry biomass. We refer to 412 the three harvests as before, during and after drought, and use the aboveground 413 biomass as a proxy for productivity.

415
Additive partitioning 416 We used the additive partitioning approach 48 to decompose the net biodiversity effect 417 (NE) on aboveground biomass into complementarity effect (CE) and sampling effect 418 (SE), separately for each harvest, selection treatment and block. We did not perform 419 the partitioning for mixtures with zero biomass 63 . For monocultures with zero 420 biomass, we kept the ones which had alive plants in the previous harvest but excluded 421 the ones which had already lost all of them. For example, when we did the 422 partitioning for the second harvest, we kept the monocultures that had zero biomass in 423 the second harvest but non-zero biomass in the first harvest; we excluded the 424 monocultures that had zero biomass already in the first harvest. This was to assure 425 that communities that died before the drought could not reappear during or after the 426 drought, and communities that had died during the drought could not reappear after 427 the drought. 428 We used mixed-effects models to assess the influences of drought vs. ambient 429 selection treatments on biodiversity effects separately for each harvest (Fig. 3,   430 Supplementary Table 3). Block and selection treatment were set as fixed-effects terms, to mixed-effects models that estimate variance components for random-effects terms 445 with maximum likelihood 64 . 446 We additionally tested the significance of biodiversity effects on productivity 447 separately for each selection treatment and harvest (Supplementary Table 2). We set 448 block and species composition as fixed-and random-effects terms, respectively. The  Table 11). Therefore, we excluded the history of 458 functional group richness from further analyses and presentation in this paper.  In the same way as net biodiversity effects on productivity were calculated for 478 additive partitioning, we calculated biodiversity effects on biomass stability as the 479 difference between each mixture and its corresponding monocultures. Then, we tested 480 the influence of selection treatment on the biodiversity effects on biomass stability.

481
Block and selection treatment were set as fixed-effects terms; species composition and 482 its interaction with selection treatment were set as random-effects terms (Fig. 4,   483 Supplementary Table 5). The log-transformed biomass at the first harvest was also included as a covariate 19 . To assess the significance of biodiversity effects on biomass 485 stability for each selection treatment, we fitted another set of simplified models, with 486 block and log-transformed biomass as fixed-effects terms, and species composition as 487 random-effects term (Fig. 4).

489
Neighbor interactions 490 We assessed interactions between neighboring plants within pots using the metrics of expected. Values greater than -0.75 indicate some sort of overyielding due to higher 508 density or higher density and higher diversity.

509
To assess how selection treatment modified interactions between plants, we 510 tested the effects of selection treatments on neighbor interaction intensity separately 511 for monocultures and mixtures. We included block and selection treatment as fixed-512 effects terms, species composition and its interaction with selection treatment as 513 random-effects terms ( Supplementary Fig. 3, Supplementary Table 6). 514 We calculated the difference between heterospecific interaction in a mixture and 515 conspecific interactions in its two corresponding monocultures. A positive value of 516 this difference indicates a weaker heterospecific than conspecific competition (i.e., 517 niche differentiation) or stronger heterospecific than conspecific facilitation, which 518 may lead to a positive complementarity effect. We tested the effects of selection 519 treatments on interaction difference for each harvest by fitting block and selection 520 treatment as fixed-effects terms, and species composition and its interaction with 521 selection treatment as random-effects terms (Fig. 6, Supplementary Table 8). We also 522 tested the significance of the interaction difference for each selection treatment by 523 fitting block and species composition as fixed-and random-effects term, respectively 524 ( Fig. 6, Supplementary Table 7).  We used linear mixed-effects models to assess the influence (generalized across 558 species) of selection treatments on trait values (red lines in Supplementary Figs. 4-6).

559
Block and selection treatment were set as fixed-effects terms; species and its 560 interaction with selection treatment were set as random-effects terms. Alternatively, 561 we set species, selection treatment and their interaction as fixed-effects terms to assess 562 whether species responded differently to the selection treatments (Supplementary 563   Table 9). LMA, LA, leaf osmotic potential, leaf stomatal conductance and root-shoot 564 biomass ratio were log-transformed to improve normality of residuals. 565 We also measured leaf relative chlorophyll content, LA and LMA in mixtures 566 before the drought event (Supplementary     Table   846 2), i.e., positive or negative average biodiversity effects. Grey points represent means 847 for species pairs (standard errors for species pairs were not shown). Grey lines 848 connect the same species pair between the two selection treatments.  Table 8). Red points and error bars show means ± standard error.

880
The filled red point shows a significant difference from zero (horizontal dashed lines; 881 see Supplementary Table 7), i.e., weaker heterospecific than conspecific competition.

882
Grey points represent means for species pairs (standard errors for species pairs were 883 not shown). Grey lines connect the same species pair between the two selection 884 treatments.