One major and under-appreciated aspect of stratigraphic modeling by diffusion process is the range of diffusion coefficients used to simulate natural examples without considering their meaning in term of transport and deposition processes. Most of the time, stratigraphic simulation tools are indeed used as a semi-inversion tool based on a "best-fit" approach to reproduce well-constrained sedimentary architectures. The aim of this work is to consolidate inputs of stratigraphic numerical modeling such as stratigraphic architectures, slopes of sedimentary systems, grain-size distributions and diffusion co- efficients calibrated on natural examples of passive margin deltas: the Plio-Pleistocene Ogooué Delta in Gabon and the Mio- to Pleistocene Zambezi Delta in Mozambique. We calculate diffusion coefficients from high resolution seismic stratigraphy and well analyses in three-steps: (1) calibration of sand/shale ratio variations (from wells) along the depositional profile; (2) restoration of the slope of sedimentary systems at time of deposition (including differential compaction corrections); (3) calculation of accumulation rates using a high resolution age model and quantification of uncompacted volumes for each time step in various stratigraphic context. Most of calculated diffusion coefficients range over two orders of magnitude (x0,01 to 10 km2/ka), in agreement with most of the published diffusion process-based stratigraphic modeling (x0,0001 to x10 km2/ka). Our results suggest that: (1) 621neither the stratigraphic context nor the sand/clay ratio impact the diffusion coefficients but, (2) they strongly depend on the slopes along the de- positional profiles. We also observe (3) a high variability of coefficients on the shelf and the basin floor that could reflect the occurrence of sedimentary processes that cannot be simplified to a simple diffusion (e.g. waves or flood on the shelf; oceanic currents, turbiditic channels or hemipelagites in the most distal domain). We use the diffusion coefficients values to calibrate the stratigraphic model developed by Yuan et al. (COLORS project, funded by Total) based on Bayesian inversions and optimization scheme of geometrical parameters of observed stratigraphic architectures (such as the angle of migration of the offlap-break or slope variations along depositional profiles).