From time to time, over the years, Micon has been asked by clients whether they can make use of old or historic assay data in a mineral resource estimate.
This situation typically involves a project which has been around for quite a while and has been drilled multiple times by different operators. It may even have a prior production history. Some of the phases of drilling (or underground sampling) often go back decades. The client usually has the drill logs and assay results. They may even have some or all of the old core and will know the core diameter (and therefore the sample size). Rarely will they have assay certificates which will describe details of the sample preparation and assaying methods employed (although there may be historic records and reports describing some of this). Typically, if the data predate National Instrument (NI) 43-101 (February 1, 2001), information about any quality assurance/quality control (QA/QC) results (certified reference standards, assay duplicates and blanks) may be absent and, if they are more than a couple of decades old, there may not have been any independent data verification performed by the operator at all.
Figure 1 shows a theoretical longitudinal section of a deposit, outlined in red, with black circular pierce points of the historic drill holes used to explore it. These may have been drilled from surface and/or underground.
Theoretical Longitudinal Section Showing Historical Drill Hole Pierce Points
As can be seen in Figure 1, the amount of data involved may be substantial and duplicating it could cost hundreds of thousands or even millions of dollars. The client usually wants to use as much of it as possible. What can be done to complete best practices data verification and justify its use?
Twinning of drill holes is often suggested as a verification technique. This involves locating old drill hole collar locations and setting up a drill a metre or two away to complete a hole of the same azimuth, dip and length of the selected hole.
Such a hole will likely confirm the intersected width of the mineralization and its location in three dimensional space. However, especially in gold deposits, it may not confirm the grade, given the often nuggety nature and short range grade variability of the deposit type. Confirming the whole database may also involve the drilling of a large number of twinned holes at significant expense. Sometimes finding the collars is problematic, and, if located underground, access may be impossible.
Some years ago Micon encountered a resource estimator from Vancouver, Gary Giroux, using another validation method which is sometimes able to confirm much, or all, of the database without drilling a single hole. Micon has used it several times since.
It is now 2017 and NI 43-101 has been in force for over 16 years. Therefore it is quite possible that the most recent round of drilling completed at a project will have been performed with industry best practice QA/QC and data confirmation/validation. The results will likely be documented, at least in summary fashion, in an NI 43-101 Technical Report filed on SEDAR (www.sedar.com). In addition, given modern electronic data storage, the original assay certificates and QA/QC results may be available from the assay laboratory. These data sources may allow the new Qualified Person (QP) to verify and validate the assay data results from the most recent drill program.
Part of the data verification procedures used by a QP responsible for resource estimation will often be the calculation of univariate statistics for the data set as well as the plotting of such graphs as box and whisker plots, log histograms and log probability plots. This will be in addition to the control charts for the reference standards, assay duplicates and blanks, which are typically absent in the older data sets.
Figure 2 shows the same theoretical longitudinal section as in Figure 1, but with some more recent drill hole pierce points than in Figure 1, which are shown as red stars.
Theoretical Longitudinal Section Showing Historical Drill Hole Pierce Points and Additional Recent Drilling
Verification of the older assay data may be performed if, QA/QC data are available for the most recent vintage of drilling and the data verification steps of univariate statistics calculation and graphs are prepared separately for each vintage of drilling. Comparison of the statistics and graphs from the recent drill data can be made against the results from the older vintages of drilling. If similar population distributions are seen between the old data set(s), and the newer, validated data set, then the old data can be considered as suitable for use in a mineral resource estimate.
Several caveats apply when using this technique.
- The assay data must be properly domained prior to analysis, with data-by-vintage comparisons being made only within a mineralized domain of common geological and mineralogical characteristics (likely the same domains being used for mineral resource estimation). Assays outside the domain must be excluded. In the case seen in Figure 2 only domain assays from those black circle intercepts within the red lines would be selected. If multiple zones are present, they too may need to be separated unless determined to be one domain. Modern resource estimation software makes this selection and compilation relatively easy.
- When selecting an area for extraction of data it is important that each vintage of data be evenly distributed therein. Figure 2 shows evenly distributed black circles and red stars throughout the zone, although the density of the red stars is much lower. This density difference is not critical, but the even distribution of intercepts within the domain is, so that the zone is evenly sampled.
- If one of the vintages of drilling is absent in a portion of the domain then a sub area, where each of the distributions is even, must be selected.
- The comparison requires that several hundred assays be available in each data vintage set for valid analysis.
- If there is a significant variation in sample length within the domains then common-length assay composites should be used for comparison. Short samples are sometimes taken around visible gold in core. This can result in assay results which are skewed high relative to those from standard sample lengths.
Figure 3 shows a comparison of log probability plots of real gold project drill hole assay data used by Micon to verify 30-year-old historic drill data from a project. Similar population distributions can be seen between the new, validated client data and the 30-year-old data, which was then considered validated for use in a mineral resource estimate.
Comparison of Real Gold Assay Data by Vintage
When comparing probability plots from very old data, particularly data generated at a mine assay laboratory, one often finds a different lower limit of detection which may be much higher than that received from a modern commercial laboratory. This lower limit difference may vary from 5 ppb to 500 ppb for gold assays.
Figure 4 shows more real gold assay data. In this case comparing old mine assay data generated by the mine’s assay laboratory, to newer, validated, yet historical exploration data. The older mine data comprise a very large data set of drill holes and channel samples in stopes and development heading, some too low grade to mine, and were collected over many years.
Comparison of Real Gold Assay Data: 60-Year-Old Mine Data vs. 30-Year-Old Exploration Data
One can see the lower limit of detection for the old mine data is higher (vertical line at 0.03 g/t) than the younger data set. However, between 1 g/t and 30 g/t the two curves are very similar. Given that this was an evaluation of a project which would be mined by underground methods, that this grade range is the heart of the “ore grade” portion of the deposit and that the resource would be classified as inferred, it was decided to use the historic mine data. Assays from weakly mineralized portions of the vein(s) would not contribute to the mineral resource estimate.
Older mine core is often only about 1 inch (25 mm) in diameter, although it is typically whole-core sampled. If significant core size differences from the various vintages of drilling were introducing biases it is assumed that these would show up in the graph comparisons. As long as the population distributions remain similar it is concluded that the core size differences are not causing a problem and the data may be used.
If newer assay data, with associated QA/QC results, are not available for the project then the new (red star) holes shown in Figure 2 will have to be drilled to validate old data. Drilling in this even pattern has the advantage of also confirming the deposit width and location, as does twin hole drilling. However, when laid out properly, it also allows for confirmation, by vintage, of all of the old assay results within a domain as well as the overall size of the deposit. Some of the new holes to be drilled could be twins, if desired, but, given short range grade variability, this may not provide much useful, additional information.
Verification of older assay data may be performed if QA/QC data are available for the most recent vintage of drilling and the data verification steps of univariate statistics calculation and graphing are prepared separately for each vintage of drilling.
Micon has only seen this technique used on gold and silver deposits but has no reason to believe that it can’t be employed on other metal deposits. It is not known if there is any absolute age cut-off for older assay data but the analysis described herein should detect any issues.
If you want to use it on something else than precious metals, you will have to use a probability-scaler plot. Using a log scale will shrink the curves one on top of the others, and reduce your capability to discriminate them. Using scalar scale will enable also to discriminate various populations and estimate their univariate statistics.
Interesting! Supposed, both data sets are from one and the same domain boundaries, then I would use Probability to Probability plot to check if the data sets are from one and the same deposit. If they are, the graph would be close to 45-degree line. Otherwise, either of the measurement sets could be questioned.
Would be happy to have the data sets available for study, if possible.
Terry, thanks this is a useful short paper. Whilst caution must be used, there are many projects or resources where good quality historic data was collected, and to ignore it is just as wrong as to simply use it. Your paper shows there are ways to reasonably assess historic data for considered use. I have come across projects where historic data is simply ignored by all involved…despite it having been done using the best standards of the day.
Being ‘picky’ about the example in general, I’m wondering — though useful in verifying the resource, when converting to extractable reserve, would you exclude the area of the crown pillar on the shaft and those holes or require new drilling in proximity to the shaft at all levels to pull those potential ore blocks back into reserves? I also have some concern on the accuracy of the ore boundaries between levels.
Similar populations, yes, but a lot can hide in plots of log data. Consider that 95% of the historical data is < 7.3 g/t whereas 95% of the client data is < 9.6 g/t, a 32% difference which demands explanation. Would anyone accept a resource estimate if that level of uncertainty existed in the average grade? In this case, one assumes that the newer data is better quality, but is there bias in the sampling? I'm not arguing against the use of the historical data in this case, but clearly there are questions that need to be answered.