Pruning management drives early accumulation of deep soil carbon following agroforestry expansion

Despite the political momentum and financial support for tropical agroforestry as a nature-based climate solution, the true carbon capture potential is highly uncertain because evidence from large-scale field experiments remains rare. We implemented a large-scale, multi-site agroforestry experiment focusing on management (i.e., tree pruning intensity) and historical land use contexts in southern Bahia, Brazil. Across three sites, cocoa ( Theobroma cacao ) based agroforestry plots (14.5 x 13.4 m) were established in a randomized block design of replicated moderate and intense pruning treatments established over (1) treeless pastures and (2) integrated pasture-banana ( Musa spp.) system. After two growing years, results indicated that soil organic carbon (SOC) responses to agroforestry expansion (0–90 cm soil layer) varied by pruning intensity and depended on prior land use. In areas previously under a pasture-banana consortium, SOC stocks remained unchanged under moderate pruning (P = 0.645) but decreased by 47.7 ± 15.7 Mg C ha −1 in plots under intense pruning (P = 0.085). Conversely, in areas previously under treeless pastures, SOC stocks increased by 40.7 ± 11.1 Mg C ha −1 (P = 0.030) under moderate pruning but remained stable under intense pruning. A mechanistic assessment through carbon isotopes showed that moderate pruning boosted SOC stocks through new agroforestry-derived SOC accumulation at deeper soil layers (>30 cm), which offset surface losses of pasture-derived SOC. These findings suggest that land use history influences SOC outcomes in response to agroforestry management. This study supports moderately pruned systems as an effective nature-based climate solution and prioritizing non-integrated pastures as optimal areas for tropical agroforestry expansion.