T187 Mathematics, Statistics, and Machine Learning in the Geosciences. How Can We Solve Today’s Challenges through Data Mining and Artificial Intelligence?

Tuesday, 11 October 2022
1:30 PM - 5:30 PM
502

IMPROVING RESOURCE MODELS WITH RECURSIVE PARTITIONING FOR DOMAIN DEFINITION: CASE STUDY OF MNCO3 AT NSUTA, SOUTH WESTERN GHANA

Paper number: T187-2
Program: Technical Sessions
Time: 1:35 PM – 1:50 PM

Authors

  • Francis Fosu Presenting Author
    University of Mines and Technology, Tarkwa , Ghana, Nsuta, Tarkwa

Abstract

This presentation describes the use of Recursive Partitioning (RP) to support the creation of domains for resource estimation at the Nsuta manganese carbonate deposit in southwestern Ghana. The geological database for logging was used in conjunction with the assay database to create a recursive partitioning tree. The terminal nodes in this decision tree were then coded as indicators, 0s for weakly mineralized and 1s for strongly mineralized. Following analysis of the spatial continuity of the indicators, an inverse distance interpolation was used to estimate the probability of strong mineralization, guided by the indicator variogram analysis. A threshold level for the probabilities was then chosen to partition the block model into two domains for resource estimation.

The advantage of this approach over more conventional methods is twofold. First, it avoids the direct use of grade data as the basis for creating estimation domains. Many authors have previously recognized the shortcomings of this “grade zone” approach, which risks over-estimation.

The second advantage of this approach is that it can construct domains that do not depend on a single geological criterion. The use of Recursive Partitioning allows one to create indicators that are based on several geological factors, more complex than what human interpreters are easily able to achieve.

The recursive partition analysis confirmed that the barren lithologies are predominantly the foot-wall or hanging-wall materials and also identified the geological characteristics of the carbonate ore materials that are the main host of mineralization.

The presentation discusses the data analysis and cleaning (QA/QC) necessary as a first step before Recursive Partitioning. It discusses the calculation of probabilities, the checking of the appropriate threshold for partitioning the deposit into domains, and the use of customized software to convert the block model partition into DXF wireframes that are needed by the software that performs the final resource estimation. Resource estimation was done by ordinary kriging and checked by inverse distance interpolation, with the two methods coming within 5% of each other on tonnage and grade.

Keywords: Recursive partitioning, resource estimation, domaining

Geological Society of America Abstracts with Programs. Vol. 54, No. 6

doi: 10.1130/abs/2022AM-382068

© Copyright 2022 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.