Aquatic islands in the sky: 100 years of research on water-filled tree holes

For the meta-analyses, we compiled raw data from unpublished or published community studies where possible. If data was available as a figure only we used the web-based tool WebPlotDigitizer (https://automeris.io/WebPlotDigitizer/) to extract the data points. We also used Pearson’s correlation coefficients (r) from published studies reporting the effects of tree-hole size or detritus amount on the abundance and richness of organisms. Furthermore, t, F and χ2 values were converted to correlation coefficients in some instances. Correlation coefficients between tree-hole size or detritus amount, respectively, and organism abundance or richness, respectively, were then used as effect sizes in four meta-analyses after transformation to Fisher’s z. We ran additional analyses using partial correlation coefficients for those studies for which we had raw data on both detritus amount and size to control for the effects of size on detritus amount. Data on separate groups of organisms or response variables from one study were used as separate data points. Annual mean temperature and annual precipitation were used as reported by the studies. If this information was not available from the publication, temperature and precipitation data were extracted from the WorldClim database (version 1.4, www.worldclim.org). For those studies that used several sites, averages of the climate data between the sites were used. If several sites per study were used, these were never more than 300km apart and climates did not differ much.