Joint stochastic simulation of petrophysical properties with elastic attributes based on parametric copula models

• Published in: Geofísica internacional Vol. 62; no. 2; pp. 487 - 506
• Main Authors: Vázquez-Ramírez, Daniel, Le, Van Huong, Díaz-Viera, Martín A., del Valle-García, Raúl, Erdely, Arturo
• Format: Journal Article
• Language: English
• Published: Universidad Nacional Autónoma de México, Instituto de Geofísica 01-04-2023
• Subjects: archimedean; bernstein; co-simulation; copulas; elastic seismic attributes; Geochemistry & Geophysics; petrophysical properties; Arquimedeanas; atributos sísmicos elásticos; propiedades petrofísicas; Archimedean; co-simulación; cópulas; Bernstein; elastic seismic attributes; co-simulation; copulas; petrophysical properties
• ISSN: 2954-436X; 2954-436X
• DOI: 10.22201/igeof.2954436xe.2023.62.2.1593

Abstract The spatial stochastic co-simulation method based on copulas is a general method that allows simulating variables with any type of dependency and probability distribution functions. This flexibility comes from the use of a copula model for the representation of the joint probability distribution function. The method has been mainly implemented through a non-parametric approach using Bernstein copulas and has been successfully applied for the simulation of petrophysical properties using elastic seismic attributes as secondary variables. In the present work this method is implemented through two other approaches: parametric and semi-parametric. Specifically, for the parametric approach the family of Archimedean copulas is used. First, the parametric approach is validated against a published case, and then a comparison of the three approaches in terms of accuracy and performance is made. The results showed that the parametric approach is the one that reproduces the data statistics worse and presents greater uncertainty with a lower computational cost, while the non-parametric approach was the one that best reproduces the dependence of the data at a high computational cost. The semi-parametric approach reduces the computational cost by 10% compared to the non-parametric approach, but its accuracy is significantly degraded.