Testing the Exactitude of Estimation Methods in the Presence of Outliers: An accounting for Robust Kriging

Subject Areas : Mineralogy

Shahrokh  Paravarzar ¹ (Department of Mining and Metallurgical Engineering, Amirkabir University of Technology, 1359767191 Tehran, Iran)
Nasser  Madani ² (Advance Mining Technology Center (AMTC), Santiago, Chile)
Abbas  Maghsoudi ³ (Department of Mining and Metallurgical Engineering, Amirkabir University of Technology, Tehran, Iran)
Peyman  Afzal ⁴ (Department of Mining Engineering, South Tehran Branch, Islamic Azad University, Tehran, Iran)

Received: 2016-06-05 Accepted : 2016-06-05 Published : 2014-03-21

Keywords: gold, Robust Kriging, Lognormal Kriging, Zarshuran,

Abstract :

Estimation of gold reserves and resources has been of interest to mining engineers and geologists for ages. The existence of outlier values shows the economic part of the deposits subject to the fact that don’t depend on the human or technical errors. The presence of these high values causes a pseudo dramatically increment in variance estimation of economical blocks when applying conventional methods as such as kriging. Conventional approach such as replacing the upper limit (capping), log-normal and Median-indicator kriging have been somehow resolved the problem of estimations but they still have numerous drawbacks. The Zarshuran gold deposit is one of the nominated gold deposits in Iran that choosing the appropriate method is a great chance in the estimation procedure. Robust kriging is a method that is able to reduce the allocated pseudo kriging weights to the blocks surrounding with high value data, and correct them in a manner that not only enhance the accuracy of the estimation but also reduce the estimation variance to the reasonable level. In this method, the estimation can be implemented in the presence of outlier and do not need to eliminate, reduce or transfer the data to another space (normalization). In this paper an attempt has been made for estimating the gold in the ore used to be compared with capping high value method. Results showed that this method has a high efficiency and is able to run estimation properly and produces high precision due to the high value reduction.