Exploitation of unconventional resources could prolong the gas production in the North Sea. Low-net-to-gross fluvial intervals have tough-gas reservoir potential in thin-bedded crevasse splays. To assess economic risks associated to the development of these reservoirs, a reliable depositional model is required. Sparse areal data availability for reservoir models commonly results in the use of stochastic interpolation. Numerical models offer the possibility to support these methods with proven physical concepts. To this end, simulations were conducted with Delft3D process-based modelling software. Input parameters and the validation data sets for these models are derived from outcrop studies in the present-day R\'{i}o Colorado fluvial system in the Altiplano Basin, Bolivia. In total 174 simulations were run, of which five models were analysed in detail. The model output consists of discrete morhpological, hydrodynamical, and grain-size data. Petrophysical parameters where derived from continuous grain size distributions parameters. This study demonstrates a statistical approach to extract these from discrete grain-size data. A sensitivity study shows that using the right size and amount of grain-size classes is crucial to minimize errors. Simulations demonstrate the effect of different conditions through their direct influence on physical processes. Crevasse-splay formation only occurs for water levels that are higher then the base of the levee breach and lower then the top of the levee. The resulting morphology is dependent on the relation between the driving hydrodynamic gradient and the counteracting frictional forces. Both outflow from the river onto the floodplain and reflux of water back into the river leads to an erosional channel network. Crevasse channels, together with their related overbank and mouth-bar deposits, form the architectural elements that make up crevasse splays. A quantitative comparison was made between field and simulated data. It shows that the model overestimates the grain size as sediment is transported away from a channel. The conclusion is that this model is not valid for quantitative purposes. The process-based model with its current abilities and limitations can be used as a tool for heuristic research. It allows the study of the different governing processes. These ideas should be taken back into the field to test, validate, and further study the exact morphodynamics of the system. Resulting concepts can be used to accurately predict the crevasse-splay architecture and corresponding sedimentary trends. Subsequently these can be used as an input in reservoir models to improve their reliability.