Predictive Geological Mapping Using Closed-Form Non-stationary Covariance Functions with Locally Varying Anisotropy: Case Study at El Teniente Mine (Chile)
Résumé
This paper is concerned with the problem of predicting the surface elevation of the Braden
breccia pipe at the El Teniente mine in Chile. This mine is one of the worlds largest and
most complex porphyry-copper ore systems. As the pipe surface constitutes the limit of the
deposit and the mining operation, predicting it accurately is important. The problem is
tackled by applying a geostatistical approach based on closed-form non-stationary covariance
functions with locally varying anisotropy. This approach relies on the mild assumption
of local stationarity and involves a kernel-based experimental local variogram a weighted
local least-squares method for the inference of local covariance parameters and a kernel
smoothing technique for knitting the local covariance parameters together for kriging purpose.
According to the results, this non-stationary geostatistical method outperforms the
traditional stationary geostatistical method in terms of prediction and prediction uncertainty
accuracies.