Belowground biomass plays a central role in carbon storage within mangrove ecosystems, yet remains one of the least quantified components due to methodological challenges. In scrub forests of Avicennia germinans on the Ajuruteua Peninsula, northeastern Amazon coast, allometric models were developed to estimate root biomass across distinct height classes. Using destructive sampling and structural predictors, including stem dimensions, crown attributes, and wood density, robust equations were generated to characterize biomass allocation among root compartments. The models demonstrated high predictive accuracy and revealed marked differences in belowground biomass among vegetation classes, reflecting strong environmental filtering driven by salinity and water limitation. Patterns of carbon allocation emphasized the dominance of belowground compartments in stressed scrub formations. These findings refine estimates of root carbon stocks, improve understanding of carbon partitioning in Amazonian mangroves, and offer methodological advances relevant for climate mitigation assessments and emerging blue carbon initiatives.