Data extraction
For invertebrate exclusion and/or inclusion treatments of each article, we recorded sample sizes (n), means of mass loss or decomposition rates, and standard deviations (SD) from tables directly or extracted data from figures by performing Web-PlotDigitizer (Burda et al. 2017). Standard errors (SE) reported in the original articles were converted into SD using the formula \(\mathbf{S}\mathbf{D=SE\times\sqrt{}n}\). Means of mass loss were converted into annual decomposition rates using the negative exponential decomposition equation described by Olson (1963). Other information we recorded from the original articles include latitude, longitude, biome, mean annual temperature (MAT, °C), mean annual precipitation (MAP, mm yr-1), soil pH, litter traits (carbon (C), nitrogen (N), C:N ratio, lignin:N ratio), duration of decomposition, and the method to exclude invertebrates (physical vs. chemical).
All sites were classified into geographic groups for testing regional variations. First, we grouped sites into ‘the tropics’ and ‘the non-tropics’ based on biomes as stated in the original articles which we further checked by spatial coordinates. Specifically, tropical wet and dry forests were grouped into ‘the tropics’ (96% belonged to tropical wet forests); other forest biomes were grouped into ‘the non-tropics’. Biomes are powerful biogeographic units for studying large-scale patterns of carbon and energy fluxes (Yi et al. 2010; Mucina 2019). Our classification of biomes followed Dinerstein et al.(2017). Fig. 1 was plotted using ArcGIS (version 10.2, ESRI, 2020). We also assigned sites into zoogeographic realms to explore potential biogeographic effects (e.g. dispersal and evolutionary histories). Zoogeographic information of each observation followed Holt et al. (2013) which is based on vertebrates but is generally pertinent to the assessment of invertebrate distributions (Liria et al. 2021).
To explore potential moderators of regional variation of the effects of invertebrate son decomposition, we tested several potential explanatory factors: termite diversity (a decomposer group the diversity of which is different in the tropics and non-tropics), litter traits (C, N, C:N and lignin:N ratios), climate and soil pH. Termite diversity values were extracted from a corresponding prediction model. The diversity predictions are estimated from a model which was ′trained′ using alpha-diversity values from 700 sites (Woon et al., in preparation). We acknowledge that species diversity and richness do not always confer higher contribution to ecosystem services compared with functional diversity, but, currently, this is the best proxy we have to identify global patterns of species distribution of the group. Where data were absent from focal studies we obtained missing litter quality data from the TRY plant trait database (Kattge et al. 2020), missing soil pH data from the Harmonized World Soil Database (https://www.fao.org/soils-portal/en/, resolution = 5′), and missing climate data (mean annual temperature, MAT and mean annual precipitation, MAP) from the Worldclim database (http://www.worldclim.org/, resolution = 5′).