B0476
Title: Fast mixture spatial regression: A mixture in the geographical and feature space applied to predict oil in the post-salt
Authors: Marcos Prates - Universidade Federal de Minas Gerais (Brazil) [presenting]
Lucas Michelin - Universidade Federal de Minas Gerais (Brazil)
Lucas Godoy - University of Connecticut (United States)
Abstract: The extraction of geological resources, such as hydrocarbon fluids, requires significant investments and precise decision-making processes. Porosity, a key attribute of reservoir rocks, plays a crucial role in determining fluid storage capacity. Geostatistical techniques, such as kriging, have been widely used for estimating porosity by capturing spatial dependence in sampled point-referenced data. However, the reliance on geographical coordinates for determining spatial distances may present challenges in scenarios with widely separated points. A mixture model is developed that combines the covariance generated by geographical space and available covariates to enhance estimation accuracy. Developed within the Bayesian framework, the approach utilizes flexible Markov Chain Monte Carlo methods and leverages the nearest-neighbour Gaussian process strategy for scalability. A controlled empirical comparison is presented, considering various data generation configurations, to assess the performance of the mixture model in comparison to the marginal models. Applying the models to a three-dimensional reservoir simulation demonstrates its practical applicability and scalability. A novel approach is presented for improved porosity estimation by integrating spatial and covariate information, offering the potential for optimizing reservoir exploration and extraction activities.
