From outcrop to process-based reservoir modelling of fluvial meandering systems. The key issue of parameter choice

International audienceThe need of models for heterogeneous reservoirs has stimulated the development of new models both process-based and stochastics. Such a model of fluvial meandering systems, developed at the school of Mines, is a comprehensive one. It comprises a realistic migration pattern of the channel centerline, (based on hydraulic studies) and reproduces event processes such as overbank floods, levee breaches, crevasse splays and avulsions. Various architectures are to be produced by changing frequency and intensity of overbank floods or avulsions, or by constraining the aggradation rate by an equilibrium profile. The number of parameters which rule the model has been kept limited: channel parameters, floodplain slope, frequency and intensity of other elements such as overbank floods and avulsions. In order to address the potential of such a modelling to field data, a study has been conducted on the Miocene Loranca formation (central Spain). This meandering succession is around 100m thick. Channel depth is based on measurement of point-bar and channel-fill thicknesses, giving a mean value of 3.5 m. The complex pattern of a meandering system makes difficult the estimation of the channel width from the outcrop. The channel width is thus estimated by using a classic hydraulic formula, (d=24.6h1.45). The aggradation rate is controlled by overbank floods and avulsions. Evaluation of overbank intensity is based on the levee thicknesses (0.43 m), characteristics of sheet flood deposits that show little decrease in thickness and grain size away from the channel. Levee breaches are attested by regular crevasse-splays. The rate of successful avulsions is taken at 2000 years, based on paleosol maturity stage and lateral extent of the point bars. Amongst the needed parameters, erodibility coefficient, floodplain slope and overbank flood frequency could not be directly inferred from the field. Erodibility coefficient and slope value are taken by default at 2 10-8 and 0.001, respectively. Overbank frequency is adjusted to reproduce the N to G that varies between 16 and 37% along the studied interval. These ratios could easily be obtained by changing the frequency of the overbank floods (from 190 to 600 iterations), while keeping the other parameters constant. Thus, the simulated 3D block and the outcrop display similar sand bodies and facies distribution. A further step in the potential of the model is demonstrated through a conditional simulation, run with the on conditional parameters and using measured sections as wells. The number of iterations is very similar to that of the non conditional simulations, indicating that processes were respected during the conditional simulation, despite a 10% relative increase in the N to G. Matching of simulated sandbodies with data is close to 90%