Accurate estimates of current forest carbon stocks are required for efforts to reduce emissions from tropical deforestation and forest degradation. The relative contributions of different vegetation types to carbon stocks and potential emissions are poorly understood in highly heterogeneous forest mosaics, and further field-based measurements are necessary from severely undersampled regions and forest types to improve regional scale extrapolations based on remote sensing. We assessed the aboveground biomass (AGB) of two contiguous western Brazilian Amazonian protected areas totalling 886,176ha, which contain vast expanses of seasonally flooded várzea (VZ) forest along the floodplain of the Juruá river and adjacent terra firme (TF) forest farther inland. Estimates were based on equations incorporating wood specific gravity (WSG) and tree height in addition to DBH, and derived from a network of 200 forest plots of 0.1ha (=20ha) sampled across adjacent areas of flooded and unflooded forest. A large number of small plots stratified by forest type allowed a more representative sample, encompassing the considerable variation in forest structure and composition both within and between forest types. Mean basal area per plot was higher in várzea forest plots than in terra firme plots (VZ: 37.6±1.2m ha ; TF: 32.4±0.9m ha ) but AGB was lower in várzea (VZ: 281.9±12.0Mgha ; TF: 358.4±14.4Mgha ) due to lower WSG and tree height. Linear mixed effects models showed the overriding effect of forest type on AGB, and the roles of water stress and a historical signature of selective logging pressure, particularly within várzea forests. ALOS ScanSAR generated categories of flood duration provided a more relevant description of water stress than SRTM elevation data; AGB within várzea forest was higher in plots subjected to longer flood duration. Várzea forests store significant levels of forest carbon despite their lighter-wooded trees and lower canopy stature, and yet are heavily settled by rural Amazonians, and are increasingly vulnerable to deforestation and logging. This study helps understand how baseline environmental gradients and human disturbances in these unique forests affect their carbon storage value, and highlights their importance both within and outside existing protected areas.