There a few groups that sell exploration procedures (http://is.gd/tm9got) but I found reading public reports for Mineral Resource Estimates a useful way of assess best practices. NI-43-101 reports on SEDAR (http://is.gd/bHwypK) are particularly useful as they can contain a lot of detail and an often prepared by industry experienced personnel. It can be a bit frustrating finding the actual technical reports you are seeking but they tend to be the big files to download. I also use ASX report particularly for 2012 JORC statements as there may be some useful detail in the Table 1 summaries.

I'm sure many in the group can also offer advice - probably helpful to state what type of mineralization you are dealing with as there are specific core sampling issues for some commodities.

I will look in the ASX report. Mineralization is copper sulfides (hypo and supergene) with gangue dominated by orthoclase, muscovite, chlorite, quartz. And pyrite also when material from swallow areas is considered.

I had scrutinized a few NI43-101 reports in the past weeks where sampling protocols are set by the off-site testing laboratories which provide interval selections to site personnel. Drill-cores selected are then shipped oversea to them for further compositing prior to testing.

I am looking at some procedures that can be done onsite.
The following paper describes pretty well what I mean: http://is.gd/mLv0jX at page 4, section “sampling drill core for metallurgical testing : the distribution method, 2012”

It might be worth checking CIM Best Practice Guidelines for Mineral Processing (http://is.gd/5hk3EK). You might find useful some suggestions regarding sampling for metallurgical studies.

Good question; and a hard one to answer.

I was at SGS for a number of years; working on the flotation analysis software (previously at JKMRC - involved in JKSimFloat and Geometallurgy).

SGS provides guidance on a sampling approach. Their approach is based on a clustering method which is quite logical.

Having said that I work with a Company called Corescan who uses hyperspectral analysis.
We are trying to move forward on 'texture modeling', which in your context means identify parts of the drill core that are 'similar'. The idea would be to identify samples that are 'different' with a view to focusing test work on these.

The other issue I would like to flag is that I am convinced that current flotation analysis software is inadequate.
This is because current method analyzes each sample independently. The samples need to be compared as part of the analysis system.

Basically I tried a number of years ago to identify whether any groups were interested in improved flotation analysis, and didn't get much interest. Many of the labs just do the flotation tests; no value adding.

So if you were interested in talking to me further please do so. My final comment is that be cautious about using 'Best Practice' methods. 'Best Method' really means 'conventional' and regretfully stifles new ideas. i.e. the very phrase 'Best Practice' presumes methods can't be improved.

Having said that it is always important to ensure that any method complies with base standards such as 'JORC' but sampling compliance is a separate issue to optimal sampling methods.

A few comments on the approach suggested.

I very much like the idea of checking that the composite histogram has the same degree of variation and mean grade as the domain being tested metallurgically.

Spatial weighting is important with respect the volume of the domain being tested - some nearest neighbor declustering weights attached to each datum will help here.

A Q-Q plot may be helpful in confirming the all and sample distributions have same shape and mean grade and are more transparent than using statistical tests - if two histograms have the same shape and mean then the Q-Q path will be on the X=Y line.

At Aminpro Labs we do quite a few geomet studies per year.  With help from geologists we first select ore domains by alteration and lithology and sometimes we take those and may divide them by low or highly enriched minerals (pyrite, or a contaminant). 

Hopefully your samples come in the form of drill core.  Drill core logs are available from the geologists giving you description, location and assays of each 2 meter core length.  They will be able to give you a good estimate of each domain's average grade and standard deviation.   The geologists are helpful in issuing an Excel document that has the list of all 2 m drill core logs of each ore domain showing core coordinates, assays (Au, Cu, Mo...etc) and other pertinent information.  These documents can have as much as 200,000 data lines per ore.   With this information, we filter the data to only account drill core segments with assays above the cut-off grade and we number each valid data line.   We make a histogram of the core elements and show the core average grade and STD deviations.  These are most likely going to be close to the value provided by the geologist. 

Then, by using a random number generator we select precise number of 2 m core intervals that gives us the sample weight needed for test work and replicates the average grade and standard deviations of the elements of importance (must have at least 2) as those in-situ.  It may take a few searches for the best representation.  Be assured that you will get an amazing sample every time you randomly select your cores.  Once selected, the task of finding the core shack starts.  We may do a second search for "variability" samples to test ore hardness and flotation variability.