Silicon Dale

Core Loss

This is a problem which plagues the resource modeller. What should you do with core data where there are significant core recovery problems ? It is actually an acute practical example of the 'partially missing data' problem.

It is common practice for regular-length samples to be obtained from drill core, for assaying. Where the core is of hard, competant rock there is not usually a problem, and a metre length of core contains a metre length of rock. Where the rock itself includes cavities, or if it is soft or friable, however, the metre length of core will often contain less than a metre length of rock. There may be gaps representing cavities, or the core may be fragmented with a proportion of fragments lost.

There are a number of ways in which core assay data is conventionally processed when there are core loss problems.

  1. A conservative approach is first to eliminate from consideration any samples with less than a specified percentage recovery (commonly between 50% and 80%). For this to be truly a conservative approach, the intervals which these samples represent must, of course, also be excluded from the ore body envelope.
  2. Alternatively, if the samples actually do contain mineralised material, it might be considered desirable to include the samples in the resource modelling exercise. The assays will then need to be recalculated to take into account the lost material. Conservatively, (a) it can be assumed that all material lost is at zero grade, and therefore the recalculation will simply reduce the reported grade by the recovery percentage. Less conservatively, (b) it might be accepted that the lost material is of the same grade as the material assayed. Implicitly this is what is usually done when core loss is small or is not even recorded. Where there is a large amount of material lost, however, such an assumption requires strong justification.

Both of the above procedures are mechanical, and take no account of the actual geology. Possibly a more realistic approach is to examine each case on its merits.

  1. (a) For example, core samples from an ore zone within limestone host rock could well include genuine cavities in the rock which lead to apparent core loss even if all rock drilled was actually recovered. In such a case, the reported grades may be correct, but the average density will be reduced by the proportion of cavities encountered - i.e. the apparent core loss.
  2. (b) Alternatively, it might be found that mineralisation is associated with alteration such as kaolinisation, producing friable ore within a harder host rock. In such a case, it is quite possible that high grade material is lost preferentially, and the reported grades underestimate the 'true' grade in such zones. The effect on reported mineral grades would, however, tend to be highly erratic, and it will be impossible to know sample by sample what allowance should be made.

Indeed, the problem in any such situation is that it is difficult to quantify the effects in order to yield an objective and unbiased data set for resource modelling. So are there any general rules that can be applied ? Unfortunately not. However, in the context of resource modelling, it is important to err on the conservative side (or, more correctly, to avoid erring in the direction of over-estimation). Hence method 1 cannot be faulted - provide that the ore zone envelopes are properly re-drawn. Method 2(a) is perhaps better, in that there is no need to change the geological interpretation and ore zone envelopes. The mineralised zones remain as defined on geological criteria, but the grades are still defined conservatively. Method 2(b) should be avoided unless there is real objective and independent corroboration - for example by comparison of core material with bulk samples. It may be that regression curves of reported grade against core loss percentage could give some information - but this needs to be treated with extreme caution as there is still strictly no information about the nature of the material which has been lost. If one is dealing with carbonate or evaporite sequences in which there are known to be cavities, then approach 3(a) could be used if it can be established through examination of the core (or by geophysical methods) that cavities are indeed intersected.

Beyond this, little can be done. Certainly detailed geological study of the core can shed light on the reasons for core loss, but what it cannot do is provide any real data on the properties of the material that was lost. That is, in the truest sense of the expression, 'missing data'.

Stephen Henley
Matlock, England

Copyright © 2003 Stephen Henley
Mumbo Jumbo (of the Mumbo Jumbo kind): Earth Science Computer Applications, v.18,no.9