Data extraction
Data, such as the control and treatment means, variances, and sampling effort were extracted either from the text, tables, or figures using Web Plot Digitizer (https://apps.automeris.io/wpd/). When data were presented as a time-series, the last time point in the series was used for the control and treatment means to maximise independence . When a gradient of impacts was presented, e.g., different fertilizer levels, the most extreme level was used as the treatment. Data were extracted at the highest level of taxonomic resolution possible but above family level (except enchytraeids, which are an important group commonly reported only at family level). When presented as functional groups (as is often seen in nematode and earthworm studies), the data were pooled together into the taxonomic group.
Each study could provide more than one comparison between a control and treatment, henceforth referred to as a ‘case’. Most commonly, primary papers presented data for multiple biodiversity metrics (for example, abundance, biomass and species richness), different taxonomic groups, different GCs or environmental stressors, or from different habitat types. Each was extracted as an individual case. Where a paper presented multiple GCs/environmental stressors, each was extracted independently from any other.