Carbon recovery overestimates soil ecological recovery during forest regeneration after pasture abandonment in the Amazon

Land-use changes in the Amazon, particularly forest-to-pasture conversion, alter soil properties, microbial communities, and soil carbon dynamics. Here, we assessed how soil and litter carbon stocks, soil organic matter (SOC), and microbial community composition respond to forest conversion and secondary forest restoration in degraded pasture soils of the Eastern Amazon. Using 16S rRNA and ITS amplicon sequencing, we evaluated microbial taxonomic and functional composition across land uses and applied random forest models to predict SOC and identify key predictors. In the 0–30 cm layer, pastures stored less carbon and nitrogen than forests, with carbon stocks reduced by 13.4 and 11.2 Mg C ha⁻¹ relative to primary and secondary forests and nitrogen stocks lower by 0.88 and 0.55 Mg N ha⁻¹ and showed more humified organic matter at depth. Soil δ¹³C distinguished land uses, indicating C4-derived inputs in pastures and C3 dominance in forests, whereas δ¹⁵N patterns suggested altered nitrogen cycling along the soil profile. Secondary forests showed partial recovery of carbon stocks and litter δ¹³C during vegetation regeneration, but nitrogen content and δ¹⁵N remained similar to pastures. Microbial communities differed among land uses: archaeal and bacterial communities dominated pastures and showed the highest predictive power for SOC (~62% explained variation), whereas fungi were more associated with forested systems. Despite partial recovery, secondary-forest communities remained distinct from primary forests, indicating persistent pasture legacy effects on carbon and nitrogen cycling and microbial structure. These results highlight the importance of integrating vegetation-litter-microbe interactions in regulating carbon resilience under land-use change.