Large rivers have been previously shown to be vertically heterogeneous in terms of suspended particulate matter (SPM) concentration, as a result of sorting of suspended solids. Therefore, the spatial distribution of suspended sediments within the river section has to be known to assess the riverine sedimentary flux. Numerous studies have focused on the vertical distribution of SPM in a river channel from a theoretical or experimental perspective, but only a few were conducted so far on very large rivers. Moreover, a technique for the prediction of depth-integrated suspended sediment fluxes in very large rivers based on sediment transport dynamics has not yet been proposed. We sampled river water along depth following several vertical profiles, at four locations on the Amazon River and its main tributaries and at two distinct water stages. Depending on the vertical profile, a one- to fivefold increase in SPM concentration is observed from river channel surface to bottom, which has a significant impact on the ‘depth-averaged’ SPM concentration. For each cross section, a so-called Rouse profile quantitatively accounts for the trend of SPM concentration increase with depth, and a representative Rouse number can be measured for each cross section. However, the prediction of this Rouse number would require the knowledge of the settling velocity of particles, which is dependent on the state of aggregation affecting particles within the river. We demonstrate that in the Amazon River, particle aggregation significantly influences the Rouse number and renders its determination impossible from grain-size distribution data obtained in the lab. However, in each cross section, the Rouse profile obtained from the fit of the data can serve as a basis to model, at first order, the SPM concentration at any position in the river cross section. This approach, combined with acoustic Doppler current profiler (ADCP) water velocity transects, allows us to accurately estimate the depth-integrated instantaneous sediment flux. Copyright © 2010 John Wiley & Sons, Ltd.