Complex geology estimation using the iterative adaptive Gaussian mixture filter

Stochastic Parameterizations for Improving the Estimation of Channelized Reservoirs in an Assisted History Matching Context

“…In this study, we focus on the second category using the MPS as a geological simulation model in combination with a channelized training image. In [24][25]28], we introduced a parameterization bridging the MPS with the truncated Gaussian and pluri-Gaussian simulation models. This parameterization was successfully applied for estimation and uncertainty quantification of the complex 3D-channelized reservoir of the Brugge field.…”

“…Here we propose two parameterizations based on the same idea, but the random fields are different. In [24][25], the marginally Gaussian random fields were defined by the conditional mean of a standard normal variable. In this study, we explore two directions: the first idea is to randomly sample from the conditional distribution using different random seeds for each grid-cell within each ensemble member, and the second idea is to randomly sample from the conditional distribution using a unique random seed for all the grid cells of an ensemble member, but different seeds across ensemble members.…”

“…In this study, we explore two directions: the first idea is to randomly sample from the conditional distribution using different random seeds for each grid-cell within each ensemble member, and the second idea is to randomly sample from the conditional distribution using a unique random seed for all the grid cells of an ensemble member, but different seeds across ensemble members. In the original parameterization ( [24][25]), the choice of the conditional mean is deterministic and, this could cause a deficiency in randomness and provide a smaller uncertainty. The characterization of the prior is crucial in any assisted history matching (AHM) procedure and, consequently, we wanted to explore if, by increasing the prior uncertainty of the parameter field, we obtain better results in terms of estimation and/or data match and prediction.…”

Delineating Facies Spatial Distribution by Integrating Ensemble Data Assimilationand Indicator Geostatistics With Level‐Set Transformation

“…The intrinsic scale of the facies is dependent on the properties of interest, which could vary from centimeters in laboratory to meters in field (Sassen et al, 2012). Developing mathematically general methods for delineating the spatial distribution of facies based on existing data has been an important and active research area (Lu & Zhang, 2015;Nejadi et al, 2015;Sebacher et al, 2016).…”