Spatial distribution of soil invertebrate effects on forest leaf litter decomposition
Forest litter decomposition by invertebrates was the highest in the tropics and decreased with increasing absolute latitude (Fig. 2a). Although overall microorganism effect (69% on average) on forest litter decomposition twice that of invertebrate effect (31% on average), the contribution of invertebrates to litter decomposition in the tropics approaches that of microorganisms. In boreal forests, the invertebrate contribution is low, with microorganisms acting as the major decomposers (Fig. 2a). Overall, invertebrates had significantly positive effects on forest litter decomposition, leading to a 31% increase in decomposition where there was invertebrate access (LRR++ = 0.368, Qt = 37687.144, df = 475, P < 0.001) (Fig. 2b, Table S1). Soil invertebrates increased litter decomposition significantly by an average of 40% in tropical forests (LRR++ = 0.513, df = 161, P < 0.001) compared with 26% in non-tropical forests (LRR++ = 0.303, df = 311, P < 0.001), respectively (Fig. 2b, Tables S1, S2). Tropical invertebrates contributed 1.5 times more to forest litter decomposition in situthan did non-tropical invertebrates (Fig. 2b, Table S2).
Across different forest biomes, soil invertebrates contributed significantly to litter decomposition in all forest biomes, increasing decomposition by 43% in tropical wet forests (LRR++ = 0.554, P < 0.001), 28% in evergreen broadleaf forests (LRR++ = 0.323, P = 0.001), 26% in deciduous forests (LRR++= 0.303, P < 0.001), 25% in coniferous forests (LRR++ = 0.293, P < 0.001), and 24% in cold or dry woodlands (LRR++ = 0.274, P < 0.001) (Fig. 2b). Compared with non-tropical forests, tropical wet forests held the highest invertebrate-mediated litter decomposition (Table S2). Among forests in different zoogeographic realms measured, we found the higher soil invertebrate effects on decomposition in Neotropical forests (48%, LRR++ = 0.656, P < 0.001) compared with those in the Nearctic (29%, LRR++ = 0.338, P = 0.001), the Palearctic (25%, LRR++ = 0.292, P < 0.001), and the Sino-Japanese forests (22%, LRR++ = 0.247, P < 0.001), respectively (Fig. 3, Table S2). There were no significant differences in invertebrate effects in the Neotropical forests compared with forests in the Oriental (29%, LRR++ = 0.344, P < 0.001), Panamanian (46%, LRR++ = 0.625, P < 0.001), and Australian (35%, LRR++ = 0.426, P = 0.004) data, respectively (Table S2).
Drivers of soil invertebrate effects on forest leaf litter decomposition
Mixed-effect meta regression showed a significantly positive relationship between termite diversity and invertebrate effect sizes (Qm = 17.842, P < 0.001, df = 446, Fig. 4a). For litter traits, our meta regressions showed significantly negative relationships between invertebrate effect sizes and C: N ratios in litter (Qm = 5.475, P = 0.019, df = 266, Fig. 4b) and lignin: N ratios (Qm = 5.634, P = 0.018, df = 153, Fig. 4c), whereas no significant relationships were observed between effect sizes and litter C, N, P contents (as percentage of dry litter weight) (Table S1). In terms of climatic variables, there were significant positive relationships between invertebrate effect sizes and mean annual temperature (Qm = 8.133, P = 0.004, df = 474, Fig. 4d) and mean annual precipitation (Qm = 25.672, P < 0.001, df = 475, Fig. 4e). We also found that soil pH negatively correlated with invertebrate effects on forest litter decomposition (Qm = 11.226, P = 0.001, df = 450, Fig. 4f).
In our dataset, the decomposition durations of 73% of observations are ≤ 1 year, and the remaining 23% are from 1 to 2 years. Mixed-effect meta regression showed that soil invertebrate effects weakened with the increasing of decomposition duration (Qm = 5.942, P = 0.015, Table S1), indicating soil invertebrates contribute differentially to the early stages of leaf litter decomposition (Fig. S1a). We did not find a significant impact of faunal exclusion protocol (chemical vs. physical) on invertebrate effects (Table S1). The reliability of our results was supported by Egger’s test for funnel plot asymmetry (z = 0.8708, P = 0.3838), showing that invertebrate effect sizes were not affected by publication bias (Fig. S2).