The deposits of meandering fluvial successions constitute important hosts for natural resources including hydrocarbons, water and ore deposits. Their successful exploitation requires knowledge of both the architecture of the preserved meander-belt depositional elements and their internal facies composition. However, these features are complex as meandering fluvial systems undertake a variety of types of channel migration including expansion, translation, rotation and combinations thereof. Such behaviour results in the generation and accumulation of bar deposits that are characterised by a range of external geometries and internal lithofacies compositions, giving rise to considerable stratigraphic complexity in the depositional record. Typically, point-bar and channel-fill deposits are preserved as fragments of abandoned meanders. Attempts to reconstruct the evolution and migratory behaviour of formative river courses therefore requires careful analysis and reconstructive modelling before predicting the arrangement of preserved facies. This study uses inverse and forward numerical stratigraphic modelling approaches (ChaRMigS and PB-SAND), statistically informed by a database (FAKTS) that records facies architectures from 290 analogue studies of fluvial systems, to predict three-dimensional internal facies distributions within preserved fluvial meander belts and associated point-bar elements. ChaRMigS is an inverse modelling method that uses the plan-view shape of remnant abandoned meander loops and the last active channel path as inputs, to propose multiple possible geometries of the formative river courses back through a series of reverse time steps. Stochastic modelling outputs consist of different scenarios of temporal steps that record channel and point-bar evolution, including episodes of loop abandonment due to neck or chute cutoff. In contrast, PB-SAND is a forward model that reconstructs point-bar facies architecture based on channel trajectories in plan-view at key time steps. It then uses this information to predict the internal anatomy of resultant channel and point-bar deposits, and the associated lithofacies distributions. This model is constrained by reference to examples documented from outcrop data, seismic images, or relationships derived from a quantitative sedimentological database (FAKTS). PB-SAND returns stochastic realizations of 3D facies distributions in meander-belt deposits.