Moderate wildfire is associated with depth-dependent reorganization of soil microbial communities in a subtropical forest

Wildfires are becoming more frequent and intense under climate change, with important consequences for soil microbial communities. However, it remains unclear how wildfire severity and soil depth jointly shape multiple dimensions of post-fire microbial reorganization, including microbial diversity, community composition, association networks, assembly patterns, and functional potential, in subtropical montane forests. We investigated an observed wildfire-severity gradient in a subtropical coniferous forest in Huaxi District, Guizhou Province, China, including unburned, light, moderate, and severe wildfire plots. Soil samples were collected at two depths (0–20 and 20–40 cm) four months after a single wildfire event, yielding 32 samples. Shotgun metagenomic sequencing was used to characterize taxonomic diversity and composition, association networks, neutral-model fits, and functional profiles. Microbial responses were severity- and depth-dependent, and the direction of change differed among indicators. Bacterial observed richness was higher in burned topsoil than in the unburned control and reached the highest values under severe wildfire; in the subsoil, richness decreased slightly under light wildfire, increased under moderate wildfire, and was highest under severe wildfire. Bacterial diversity and evenness peaked under light wildfire in the topsoil but under moderate wildfire in the subsoil. Fungal richness increased under light and moderate wildfire in the topsoil and remained relatively high under severe wildfire, whereas fungal diversity and evenness decreased under light and moderate wildfire but increased markedly under severe wildfire. In the subsoil, fungal diversity and evenness were lowest under light wildfire and were higher under moderate and severe wildfire. Bray–Curtis ordination and PERMANOVA showed clear shifts in bacterial and fungal community composition across wildfire-severity classes and soil depths. Network analyses indicated that bacterial networks had the highest edge number and average degree under moderate wildfire, whereas fungal networks had the largest node and edge numbers under severe wildfire. Neutral-model fits were higher for bacterial than fungal communities. At the functional level, KO Shannon diversity remained relatively stable, but KEGG and COG functional composition shifted significantly across wildfire-severity and soil-depth combinations. Wildfire severity and soil depth jointly shaped post-fire microbial reorganization in subtropical forest soils, with moderate wildfire associated with distinct community and association-network patterns, together with broader wildfire- and depth-related functional compositional shifts along the observed severity gradient. These findings indicate that post-fire soil assessment should not rely only on surface-soil responses or burn-severity classification alone. Instead, depth-specific microbial indicators, especially community composition, association-network structure, and functional profiles, may provide useful evidence for evaluating early post-fire soil recovery and guiding restoration priorities in subtropical plantation forests.