Restoration programs that promote the functioning of restored ecosystems are in urgent demand. Although several biodiversity and ecosystem functioning (BEF) experiments have demonstrated the importance of functional complementarity enhancing plant community performance, no BEF study has yet experimentally manipulated facilitation testing its contribution to how the complementarity effect modulates community performance.

We built a restoration experiment manipulating diversity and facilitation in a tropical semiarid forest. We planted 4704 seedlings of 16 native tree species to assemble 147 experimental communities with 45 different compositions comprising 1, 2, 4, 8 or 16 species. Facilitation was included in the experimental design by creating a gradient of communities from low to high facilitation potential (based on prior research). We measured functional diversity and functional identity using species above and below-ground traits to investigate how they modulate the effects of species diversity and facilitation on leaf biomass production, and its additive partition biodiversity effects (NE, CE & SE).

The joint influence of diversity and facilitation was tested separately for leaf biomass production and Net Biodiversity Effect using Linear Mixed Models (LMMs). We subsequently ran LMMs including functional diversity and functional identity. We hypothesised that facilitation would increase community productivity and functioning and that functional dispersion and functional identity related to above and below-ground traits would explain facilitation performance.

Facilitation positively influenced leaf biomass production as predicted, but unexpectedly, neither of the functional traits were important for modulating the facilitation process. Positive values for Complementarity Effect (CE) showed that plants performed better in mixtures in comparison to monocultures. Selection Effect (SE) negative values, showed that species with below-average performance in monocultures, performed better in mixtures. Unexpectedly, CE did not increase as species diversity or facilitation increased. SE was influenced negatively by facilitation leading to a more equal distribution of biomass production between species in mixtures.

Synthesis: Facilitation improves biomass production in restored communities and increases biomass equitability among plant species and thus ecosystem reliability. To improve restoration success, plant communities should be built using facilitating plants.

Methods

The work was developed at the Brazilian semiarid Caatinga, the largest and most diverse Seasonally Dry Tropical Forest of the Americas (da Silva et al., 2017). Its total area encompasses 826,411 km², and occupies 11% of the Brazilian territory, although half of its original area has been degraded, the other half is represented by 47100 forest fragments that suffer constant anthropogenic disturbance (Antongiovanni et al., 2018; Antongiovanni et al., 2020). This study was conducted at the Açu National Forest, in the northeast of Brazil (05º34'20"S, 36º54'33"W), in which there is an agreement between this university, the protection area and the environmental council of Brazil. The National Forest encompasses 528 ha, where vegetation is characterised by trees varying from 2 to 15 m in height, with mean annual temperature of 24ºC and mean annual rainfall of 607mm. Plants lose their leaves during the dry season and produce new leaves at the beginning of the rainy season that extends from January to July (Silva & Souza, 2018).

Experimental design

The study was performed at the BrazilDry experiment which is part of a worldwide network of BEF experiments that have built experimental communities using only tree species (Paquette et al., 2018). The BrazilDry experiment was implemented during the 2016 rainy season, in a 3.3 ha degraded area. The experiment consists of 147 experimentally built tree communities of 13 m x 18 m plots, each with 32 individual trees planted approximately 2 m apart. We built communities with different levels of diversity: monocultures, two, four, eight and 16 species and with different random compositions assigned within each level of diversity. The species randomisation to build experimental tree communities followed the random partition design (Bell et al., 2009), where all 16 species are present at all levels of diversity and have the same number of individuals in the experiment. We built 16 monoculture compositions, 16 compositions with two species, eight compositions with four species, four compositions with eight species, each replicated three times. Communities with the total amount of species studied (16 species) were replicated 15 times (see detailed specifications of the communities’ composition in Supporting Information Table S1). To include facilitation in our experimental design, we limited part of the randomisation using groups of species with net facilitative, neutral and competitive potential based on existing empirical data. At each diversity level, we formed two groups. The first one was completely at random and the second one limited the choice of species in a way that half of the communities had facilitative and neutral species and half had competitive and neutral species. In this way, facilitation potential became a factor included in our design that was reasonably independent of our diversity levels (see Supporting Information Fig. S1).

Facilitation potential applied in the experiment design

The potential facilitative effect of the 16 species was estimated in a previous field experiment in the Açu National Forest. Within a degraded area of 5 ha, we mapped isolated adults of the 16 tree species. Then, we randomly select five individuals of each species and planted in a block design, below the adult canopies and in an adjacent open area without the influence of any vegetation, one sapling of each of the following target tree species: Anadenanthera colubrina (Fabaceae), Myracrodruon urundeuva (Anacardiaceae) and Cenostigma pyramidale (Fabaceae). Thus, we had five replicates for each 16 tree species, with a total of 80 adult trees and 480 individuals of the target species. During 265 days, we monitored three variables of plant performance: height, number of leaves and the number of days in which saplings survived. We then calculated the relative interaction intensity index - RII (Armas et al., 2004) for each adult tree species. The RII was calculated for the three performance variables separately:

Where, P_a is the performance of the plant under the adult plant and P_o is the performance of the sapling in open spaces. The RII index varies from -1 to 1, where negative values indicate competition, while positive values indicate facilitation.

To rank species based on their facilitation potential, we calculated an average and standard deviation of RII values considering the three performance variables (RII number of leaves, RII height and RII days of survival). We then calculated a t-value for each species dividing the average RII value by its standard deviation, which we defined as the facilitative strength considering the variability of responses between the three performance variables. The species were then ranked based on their t values and divided into four groups for randomization: High facilitative; Neutral-facilitative; Neutral-competitive and High competitive (Please, see species facilitation values in Supporting Information Table S2).

Greenhouse Seedling Production

In a greenhouse located at Açu National Forest, we produced 4704 seedlings of the 16 Caatinga tree species. All seeds were sampled from at least five distinct adult trees present in the Caatinga inside the reserve. After germination, seedlings were planted in 1m long PVC tubes to allow deep root growth before planting, which reduce mortality after planting (unpublished data, G. Ganade). The seedlings received water twice a day during the first 2 months of development, after which the plants were watered once a day until they reached six months. In the fifth month, the shade cloth of the greenhouse (50%) was removed to allow plants to acclimatise in full sunlight for a one-month period until they were transplanted into the field. At the time of planting, the plants had six months of development in the greenhouse, roots 1m deep and heights varying between 0.5 and 2 meters, depending on the species. The robust size of the transplants allowed rapid development of the plant community from the beginning of the experiment.

Experimental monitoring

One year after planting in 2017, plant mortality was recorded, and the 1245 dead plants were replaced. After the replacement of the dead plants, all 4704 individual plants had their number of leaves counted and recorded as their initial size. The leaves of all plants were counted in the following year, 2018, to calculate community leaf biomass productivity.

Community Productivity

Leaf biomass was estimated in 2017 and 2018 for each plant by counting the number of leaves produced and multiplying it by an average leaf dry weight (measured previously for each plant species). The average leaf dry weight was calculated as the average of three sun-exposed leaves from five individuals of each tree species. To calculate the leaf biomass production, we used the leaf biomass difference between 2017 and 2018. Plants that died between 2017 and 2018 had negative leaf biomass production. The community leaf biomass productivity was calculated by summing leaf biomass production of all individuals within plots. In Caatinga forests, plant species lose their leaves during the dry season and during the rainy season, tree species flush new leaves and keep them for about six months. Thus, the number of leaves measured is a good indicator of annual leaf biomass production.

Functional traits

The functional characteristics of the species were sampled from individual plants produced in greenhouses, under the same conditions and for the same period as the individual plants used to set up the experimental communities. For each species, five individual plants were randomly sampled and the following traits above and below ground were measured: Leaf area (mm²), Specific leaf area - SLA (fresh area divided by dry weight), Shoot height (cm), Root depth (cm), Trunk wood density (g/cm³) and Root wood density (g/cm³) and specific root area (fresh area divided by dry weight). See Supporting Information Table S3 for details. All traits were measured following the Pérez-Harguindeguy et al. (2013) protocol.

Functional dispersion

Functional dispersion was calculated for each experimental community, as a proxy for functional complementarity. We considered all the functional traits measured, using the functional dispersion index of Laliberté & Legendre (2010). The method uses a distance matrix in pairs with trait values and an abundance matrix to calculate the mean distance from the community centroid. This method does not present correlation with species diversity, allowing us to estimate the role of functional complementarity between communities with distinct diversity levels.

Community functional identity

The community functional identity was calculated for each sampled trait: LA, SLA, Height, Wood density, Root density, and SRA. We obtained the Community Weighted Mean (CWM) of each trait by calculating the mean value of the trait of all species, weighted by their abundance per experimental community.