We compiled observational data from ecosystems characterized by herbaceous or low-statured vegetation which we refer to collectively here as grasslands. To maximize spatial coverage, we synthesized plot data from observational networks and individuals. Observational networks contributed most of the plot data, and included the Diversity NP Network (Scheifes et al. 2024, Wassen et al. 2021) (63.5% of overall database), the Nutrient Network (Borer et al. 2014) (NutNet; 22.5% of overall database; saved as separate subset, see above), the National Ecological Observatory Network (NEON - Herbaceous clip harvest, NEON - Plant presence and percent cover) (6.9% of overall database), the Biodiversity Exploratories (Fischer et al. 2010) (3.0% of overall database), the South African Environmental Observational Network (Burkepile et al. 2017, Koerner et al. 2014, Wilcox et al. 2020) (SAEON; 0.4% of overall database), and the Polar Data Catalogue (Elmendorf et al. 2012) (0.5% of overall database). Individual contributors provided data from grasslands in China (Jing et al. 2015, Peng et al. 2021) (3.2% of overall database). To ensure consistency and comparability, we applied the following plot selection criteria: i) observations were made without manipulations (i.e., natural ecosystems or control plots in experimental manipulations), and ii) aboveground peak biomass and plant species composition was assessed. This yielded a database of 4,961 plots across 3,414 sites. Biomass values were calculated per m² based on the reported sampling area. If available, we also included data measured in multiple years from the same site (376 plots; 7.6% of overall data), but as our main goal was to maximize spatial coverage, we also included sites for which only one point in time was sampled (4,585 plots; 92.4% of overall data). Total aboveground peak biomass (dry weight in g m⁻²), including live and dead biomass, is a good estimate of aboveground productivity in grasslands and was quantified via destructive harvest or reliable indirect estimates. An exception was the South African savanna dataset, where aboveground biomass was measured non-destructively using a rising plate meter. Biomass was estimated using established site-specific calibrations based on destructive harvests from the same system, a method shown to provide reliable aboveground biomass estimates. Plant species composition was quantified as species-specific cover or biomass. Percent cover was measured to the nearest 1% of each species rooted in the plot. These values provide the relative abundance of each species in a plot.

This dataset contains plot-level observations of aboveground biomass, community-weighted mean (CWM) plant traits, trait coverage metrics, and environmental variables across grassland sites compiled from multiple ecological databases. All community-level belowground trait data was calculated based on community-level vegetation composition (quantified in % species coverage per plot, or in the case of "PolarData (Polar)" as species-specific biomass per plot) derived from the respective databases. All trait data derived from the UNDERPLOT database. Each row represents one plot in one year. The subset containing data from the Nutrient Network (NutNet) used in this study is published separately due to network data sharing policies (see related works below). In addition, this dataset is accompanied with the RCode used in Andraczek et al. "Climate influences how belowground traits regulate grassland biomass". See the readme for a description of the workflow and how the script is supposed to be used. Note that this script works with both the main dataset, and in addition, also the subset containing data from the Nutrient Network which is published separately in Zenodo (see related works below).

This dataset contains plot-level observations of aboveground biomass, community-weighted mean (CWM) plant traits, trait coverage metrics, and environmental variables across grassland sites compiled from multiple ecological databases. All community-level belowground trait data was calculated based on community-level Vegetation composition (quantified in % species Coverage per plot, or in the case of "PolarData (Polar)" as species-specific biomass per plot) derived from the respective databases. All trait data derived from the UNDERPLOT database. Each row represents one plot in one year. The subset containing data from the Nutrient Network (NutNet) used in this study is published separately due to network data sharing policies (see related works below). This data is used to explore the relationships between plant functional traits and aboveground biomass and how this is modified by climate in grassland ecosystems. 

For detailed meta data on all variables included in this subset, see ReadMe attached to this repository: Andraczek_et_al_Belowground_traits_and_aboveground_biomass_main_dataset_metadata.txt