Extended Data Fig. 2: Effects of treatments on wood decomposition and insect colonization.

a, b, Coefficients and confidence intervals from post hoc tests assessing all three pairwise comparisons between the uncaged, closed-cage and open-cage treatments for annual mass loss (a; same structure as the model shown in Table 1 based on 3,578 logs) and insect colonization (b; binomial model for insect presence and absence based on 3,430 logs) of wood of native tree species. The 95% confidence intervals that do not intersect the zero line (dashed) indicate significant differences. c, Pairwise comparison of fitted annual mass loss (%) between each of the three treatments in the global deadwood decomposition experiment. Points represent the predicted values for angiosperm species at 55 sites and gymnosperm species at 21 sites based on three Gaussian generalized linear mixed log-link models for 3,758 logs with site-specific random effects and temperature, precipitation, treatment (closed cage versus uncaged, open cage versus uncaged and closed cage versus open cage), host division, as well as their interactions, as fixed effects. In a and b, the largest differences in both response variables were observed between uncaged and closed-cage treatments. Annual mass loss was higher in the uncaged than open-cage treatment and higher in the open-cage than in closed-cage treatment, although the latter was not significant. This indicates that the open cage, despite its openings for insects, has a clearly reduced decomposition rate compared with the uncaged treatment. Insect colonization for the open cage differed significantly from both uncaged and closed-cage treatment, but was more similar to the uncaged than closed-cage treatment. This indicates that open cages were colonized by insects, but not as frequently as the uncaged treatment. Open cages thus excluded parts of the wood-decomposing insect community, which may explain the rather small difference in annual mass loss between closed cages and open cages. These results suggest that the comparison of uncaged wood versus closed cages provides a more reliable estimate of the net effect of insects on wood decomposition than the comparison of closed-cage versus open-cage treatments, which is likely to underestimate the net effect of insects. In c, the difference between annual mass loss in closed-cage and both treatments with insect access (uncaged and open cage) increased from boreal to tropical biomes, whereas the difference between uncaged wood and open cages hardly deviated from the 1:1 line. This indicates that the reported mass loss differences between closed-cage and uncaged treatments, as well as the accelerating effect of temperature and precipitation (Table 1), can be attributed to insects and are not an artefact of potential microclimatic effects of the cages (Supplementary Information section 1).