Although several studies have identified the effects of functional trait diversity (FTD) and/ or identity, i.e., the community-weighted mean (CWM) of a trait, on aboveground biomass (AGB) along abiotic conditions, these effects on AGB in global forest metacommunities are still largely unexplored. Here, we modelled the effects of abiotic (i.e., climate, soil and plot physical conditions) and biotic [i.e., FTD, CWM of conservative traits (CWMCT), CWM of acquisitive traits (CWMAT), and functional dominance (FunDom; based on CWM of plant maximum height or diameter)] factors on AGB in 76 forest metacommunities (from 24 studies). Using multiple linear regression models and piecewise structural equation modeling (pSEM), we tested the hypothesis that both abiotic and biotic factors regulate AGB, but that the mass ratio mechanism underpins AGB of metacommunities in global forests better than the niche complementarity mechanism. We found that abiotic and biotic factors contributed 45.39% and 54.07%, respectively, to the explained variance in AGB (R² = 0.59), and as such, abiotic factors shaped FTD (R² = 0.42 to 0.48), CWMCT (R² = 0.33 to 0.36), CWMAT (R² = 0.27 to 0.33), and FunDom (R² = 0.59 to 0.61) through divergent effect sizes and directions. The final best-fitted pSEM showed that FunDom increased (β = 0.49) but CWMCT (β = -0.35) and CWMAT (β = -0.11) decreased AGB (R² = 0.52) as compared to the negligible effect of FTD (β = 0.04). This study supports the mass ratio effect, specifically the overruling role of tall-stature or dominant trees on AGB, at a macroecological scale, and hence, suggests that a suitable species' functional strategy is important to promote carbon sequestration in forest metacommunities that underpins human well-being. We expect that our study will advance the field of biodiversity – ecosystem functioning at a macroecological scale by using the metacommunity concept and approach.

Data assembly: FTD, CWMAT, CWMCT, FunDom, and AGB

We searched and identified the literature using the ISI Web of Science (WoS) database (https://clarivate.com/products/web-of-science/; up to and including November 2020). We used a combination of “functional trait” AND “aboveground biomass” or “functional trait” AND “biomass” or “functional trait” AND “aboveground carbon stock or storage” or “functional trait” AND “productivity”. We then reduced this list of the general literature (because it was possible to get the studies from grasslands) by reading the abstracts of most relevant papers and by using an additional keyword “forest” to above keywords. By reading the references and citations to the selected papers, we found extra relevant papers. We followed a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) model of the literature collection (605 papers), identification (251 papers), screening (109 papers), eligibility (69 papers) and inclusion (24 papers) (Fig. S1 in Appendix A). The final included 24 studies are published during 2013 – 2020. We used Webplot-digitizer software to extract the data from figures or graphs (https://automeris.io/WebPlotDigitizer/).

Extraction of abiotic factors

For each metacommunity, we extracted the mean Lat (in decimal degrees) and longitude (Lon, in decimal degrees) data as well as Nplots and PS (ha) from the eligible published studies, and we then extracted elevation (Elev, m) data. After that, using the geographic information of each metacommunity, we extracted 10 climate and 8 soil variables from open-access global databases, using rgdal, ncdf4, raster, and sp packages of R. The temperature-related climate factors were mean annual temperature (MAT; °C), diurnal temperature range (DTR; °C) and solar radiation (SRAD; kJ m^-2 day^-1), and the water-availability-related climatic factors were cloud cover (CLD; %), mean annual precipitation (MAP; mm year⁻¹), vapor pressure (VAP; Kilopascals), wet day frequency (WET; days), potential evapotranspiration (PET; mm year⁻¹), aridity index (AI) and frost day frequency (FRS; days).

For each metacommunity, soil properties, at an average of 0-100 cm depth, were collected from the Regridded Harmonized World Soil Database v1.2. These soil properties were soil cation exchange capacity (CEC; cmol kg^-1), pH (PH), bulk density (BD; kg dm^-3), available water storage capacity (ordinal variable ranging from 1 to 7) and textural properties including gravel (GRAVEL; %), clay (CLAY; %), silt (SILT; %), and sand (SAND; %).

Sheet 1 includes basic information of the meta-data. Sheet 2 includes the final dataset used in the statistical analysis, and so, R codes can be applied to reproduce the results. Sheets 3 and 4 also provide some data for reproducing results.