Assessing Local Uncertainties with Geostatistical History Matching with Local Probability Distributions

Iterative geostatistical history matching techniques based on stochastic sequential simulation, update the model parameters at each iteration globally or using a regional perturbation criterion (e.g., Voronoi polygons centred ad the well locations). While these techniques ensure the convergence of the iterative procedure towards the observed production data, the exploration of the model parameters, and the uncertainty space, is limited by each region considered. To avoid this effect, we propose the use of stochastic sequential simulation conditioned to local probability distributions at each grid cell to account for local uncertainties. The local probability distribution functions are built at the end of each iteration from the petrophysical realisations that generated simulated production curves with a misfit score below a given threshold. The proposed methodology was applied in a challenging non-stationary synthetic reservoir and the results show the advantages of the proposed technique when reproducing the geological features of interest when compared against conventional iterative geostatistical history matching.