Turbidite reservoirs frequently consist of massive sandstones with excellent reservoir properties, but showing a heterogeneity which is difficult to characterize with only well and seismic data. The physical and numerical modeling of the depositional processes is then a way to predict the heterogeneity pattern and to assist the geological interpretation. Recently, Cellular Automata (CA) modeling was adapted to simulate turbidite flow deposits. In this study, CA modeling incorporates the main submarine physical processes involved in turbulent flows, such as water entrainment, erosion, deposition and particle fall-out. These processes are developed through CA simulations, in which the cells interact by exchanging energy and flow properties. In this work, the CA modeling was applied in real oilfields of the Campos Basin (offshore Brazil), in a case where the palaeotopography played a major role on trapping of turbidite sand deposits. The sensitivity tests performed on this case study highlighted that the parameters of flow concentration, composition of the substratum and the deposition model greatly impact the simulation results. The simulation results also realistically reproduced sedimentation patterns, such as successive filling of contiguous sub-basins, increasing flow velocities in confined settings, run-up effects with lateral deposition of fines and concentration of coarser sediments in topographic lows. An important characteristic of the studied turbidite reservoirs is the presence of multiple-stacked depositional cycles. For this reason, CA code was adapted to allow the simulation of multiple flow events and, by this way, to reproduce stacked turbidite cycles. The reservoir distribution and the thicknesses of the geological model fit very well with the results of CA simulations for multiple flow events. A blind test performed thanks to the drilling of a new well in the study area also confirmed the forecast capacity of the CA modeling for both sediment distribution and thickness. The results of the simulations are consistent with the geological model of the study area, and predict reservoir distribution in locations away from the wells. These results also point to the potential of such numerical techniques in improving the prediction of the turbidite reservoir extension, especially in the case where palaeotopography controlled the turbidite sedimentation.Read less <