The development of the Leeville semi-dynamic and largely implicit geology model
CIM Montreal 2015
Richard Inglis (Newmont USA Limited), Nathan Earle (Newmont USA Limited)
In most Mineral Reserve and Mineral Resource estimates the geology model is a fundamental component of the estimate. The geology model is usually the foundation for the domaining of assay, mineralogical, geotechnical, geometallurgical and density data for the subsequent estimates. This paper outlines the benefits delivered from the implementation of a semi-dynamic, largely implicit geology model as well as some of the challenges that were overcome.
Newmont's Leeville underground operation is 43 kilometres north of Carlin in Nevada, North America and has been producing gold since 2006. The previous explicit geology model consists of formations, faults, a three dimensional grade domain for flagging the gold estimation volume and a polygonal grade domain that was used for flagging sample data. For short term production estimates the explicit geology model is updated every six weeks, however due to time constraints the formations and faults are only updated annually. The method of updating the explicit geology model is to adjust each cross section by digitising, adjusting the tag strings that control triangulation between sections and re-linking each section. Boolean operations are used to ensure final volumes are complete and not overlapping. Typically the explicit process takes 12 person weeks to complete for an entire geology model update.
With the advent of implicit geology modeling techniques it is now possible to create a semi-dynamic, largely implicit model that removes much of the digitising and manual generation of volumes. The new geology model was developed by firstly increasing the focus on data validation; collar and downhole survey errors were identified and rectified and spatially inconsistent geological logging was reviewed and modified. The new model is semi-dynamic as the following steps need to be followed when new data is added to the model:
1. Check consistency of new data,
2. Add new fault points to existing fault points, create new faults if necessary and
3. Identify formation contact points that are greater than 9 metres from all faults.
The new model is largely implicit as the formations are modeled directly from the formation contact points that are greater than 9 metres from all faults. However to make local improvements to the model 168 digitised points are added making the model explicit in places. This compares with 226,463 points digitised in the explicit model. The grade shells are completely dynamic and implicit as they are modeled using indicator interpolants and no manual is intervention required.
The semi-dynamic, largely implicit model is in the process of being reconciled with previous production before being implemented at Newmont's Leeville operation. The testing and validating phase included the production of eight alternative methods for building formations; the final model was selected based on a visual assessment of geological continuity and the numerical match between modeled volumes and the logged formation. Determining a process to handle faults in the model with imperfect logging and fault surfaces was a key to producing a visually acceptable result for the formations; at Leeville this is achieved by ignoring almost all formation contacts within 9 metres of any fault. The grade shells were produced by combining a formation control and indicator interpolants to produce a "balanced" shell for flagging the estimation volume and a "Russian doll" shell for flagging sample data in an attempt to minimise the conditional bias introduced with the use of a grade shell as described in Inglis (2013). Updating the model now takes 1 person approximately a day to complete.
The development of a semi-dynamic, largely implicit geology model at Leeville has reduced the updating time from 12 person weeks to 1 person day without sacrificing model quality. This enables personnel to focus their time on quality data collection and improving the geological interpretation rather than digitising and resolving errors in the generation of triangulations.
INGLIS, R. J., 2013. How to select a grade domain – a gold mine case study. Exploration, resource and mining geology conference / Cardiff, Wales, UK, 21-22 October 2013, p. 21.
implicit, model, geology, dynamic