A three-dimensional numerical model of sediment transport, erosion and deposition within a network of channel belts and associated floodplain is described. Sediment and water supply are defined at the upstream entry point, and base level is defined at the downstream edge of the model. Sediment and water are transported through a network of channels according to the diffusion equation, and each channel has a channel belt with a width that increases in time. The network of channels evolves as a result of channel bifurcation and abandonment (avulsion). The timing and location of channel bifurcation is controlled stochastically as a function of the cross-valley slope of the floodplain adjacent to the channel belt relative to the down-valley slope, and of annual flood discharge. A bifurcation develops into an avulsion when the discharge of one of the distributaries falls below a threshold value. The floodplain aggradation rate decreases with distance from the nearest active channel belt. Channel-belt degradation results in floodplain incision. Extrinsic (extrabasinal, allogenic) and intrinsic (intrabasinal, autogenic) controls on floodplain dynamics and alluvial architecture were modelled, and sequence stratigraphy models were assessed. Input parameters were chosen based on data from the Rhine–Meuse delta. To examine how the model responds to extrinsic controls, the model was run under conditions of changing base level and increasing sediment supply. Rises and falls in base level and increases in sediment supply occurred over 10 000 years. Rising base level caused a wave of aggradation to move up-valley, until aggradation occurred over the entire valley. Frequency of bifurcations and avulsions increased with rate of base-level rise and aggradation rate. Channel-belt width varied with water discharge and the lifespan of the channel belt. Wide, connected channel belts (and high channel-deposit proportion) occurred around the upstream inflow point because of their high discharge and longevity. Less connected, smaller channel belts occurred further down-valley. Such alluvial behaviour and architecture is also found in the Rhine–Meuse delta. During base-level fall, valley erosion occurred, and the incised valley contained a single wide channel belt. During subsequent base-level rise, a wave of aggradation moved up-valley, filling the incised valley. Bifurcation and avulsion sites progressively moved upstream. Relatively thin, narrow channel belts bordered and cut into the valley fill. These results differ substantially from existing sequence stratigraphy models. The increase in sediment supply from upstream resulted in an alluvial fan. Most bifurcations and avulsions occurred at the fan apex (nodal avulsion), and channel belts were the widest and the thickest here (giving high channel-deposit proportion) due to their high discharge and longevity. The width and thickness of channel belts decreased down-valley due to decreased discharge, longevity and aggradation rate. This behaviour occurs in modern alluvial fans. Intrinsic controls also affect floodplain dynamics and alluvial architecture. Variation of aggradation rate, bifurcation frequency and number of coexisting channel belts occurred over periods of 500 to 2000 years, compared with 10 000 years for extrinsic controls. This variation is partly related to local aggradation and degradation of channel belts around bifurcation points. Channel belts were preferentially clustered near floodplain margins, because of low floodplain aggradation rate and topography there.