A v-transformed copula-based simulation model for lithological classification in an Indian copper deposit

Area of Science:

• Geostatistics
• Computational Geology
• Data Science

Background:

• Multivariate data often exhibit non-linear and non-Gaussian dependencies, necessitating advanced modeling techniques.
• Lithological classification in spatial domains involves categorical variables, posing challenges for traditional statistical methods.
• Copula functions offer a flexible framework for modeling complex multivariate relationships.

Purpose of the Study:

• To apply a non-Gaussian copula, specifically the v-transformed copula, for lithotype classification in a spatial context.
• To compare the performance of the v-transformed copula discriminant function with other simulation methods for geological classification.
• To evaluate the classification accuracy using ground-truth data from blast hole information.

Main Methods:

• Utilized a v-transformed copula model coupled with Markov chain Monte Carlo (MCMC) simulation and a copula discriminant function.
• Simulated four lithotypes: granite, quartz, basic, and aplite within an Indian copper deposit.
• Compared the v-transformed copula simulation results against Gaussian copula, t copula, and sequential indicator simulations.

Main Results:

• The v-transformed copula simulation achieved a classification accuracy of 76%.
• This accuracy is higher than that obtained by Gaussian copula (70%), t copula (69%), and sequential indicator simulations (70%).
• The study demonstrates the superior performance of the v-transformed copula for lithotype classification.

Conclusions:

• The v-transformed copula provides a more accurate method for lithotype classification compared to traditional approaches.
• This advanced modeling technique enhances the understanding and characterization of geological deposits.
• The findings support the use of non-Gaussian copulas in spatial geological classification tasks.