Erratum: Global distribution of earthworm diversity (Science DOI: 10.1126/science.aax4851)

After publication of the Report “Global distribution of earthworm diversity,” the authors discovered a bug in their data-preparation code. This bug affected only some datasets, but it removed from certain datasets all sites where no earthworms were recorded (“zero-sites”). Unaffected datasets still retained zeros. This bug affected all analyses in the report, generally causing the models to overestimate earthworm richness, abundance, and biomass, which was reflected in the predicted overall means decreasing in the revised analyses. After reanalysis, there were also changes in the inferred geographic patterns, but only minor effects on significant environmental drivers. In total, 40% of studies (51 datasets, 64 studies) in the richness model, 33% of studies (56 datasets, 69 studies) in the abundance model, and 28% of studies (30 datasets, 36 studies) in the biomass model were affected (see the figure). Of those affected, most datasets had only a relatively small number of zero-sites. However, 13 affected studies used in the richness model, 14 affected studies used in the abundance model, and 7 affected studies in the biomass model more than doubled in size when the zero-sites were re-added to the underlying datasets. Four datasets were also removed during the reanalysis, because upon detailed inspection they did not meet the criteria for inclusion in relation to number of sites. deviate considerably from the previous patterns of species richness and abundance. Because a large number of zero-sites were previously omitted from North American studies, the model now fits better in the northern United States/Canada region. The predictions in certain parts of Europe and Asia are likely affected by the North American data, and thus, the predictions of both regions are now less extrapolated across environmental space. The patterns in the global map of earthworm biomass are consistent in both versions. Incorporation of the zero-sites has reduced the Report’s global estimates considerably. The mean global estimates are now, for richness, 0.95 species, SD = 0.96 (previously 2.42 species, SD = 2.19); for abundance, 20.59 individuals/m², SD = 24.84 (previously 77.89 individuals/m², SD = 98.94); and for biomass, 51.18 g/m², SD = 4881.2, median = 6.16 (previously 2772.8 g/m², SD = 1,312,782, median = 6.69). Climate and habitat cover are still more important drivers of earthworm communities than soil properties, although there are some changes to the importance of the different themes. Most noticeably, habitat cover has increased in importance in both the abundance model and the biomass model. The importance of temperature has also changed in the abundance and biomass models (decreasing and increasing, respectively). The text and figures now reflect these changes. The error did not quantitively affect any of the overall conclusions drawn in the original paper. A full description of the reanalysis and all changes to the text are on GitHub (DOI:10.5281/zenodo.3901273). The updated data are now available on the iDiv Data portal (DOI:10.25829/idiv.1818-13-3001). This erratum also adds an additional author to the author list.