Tuesday, December 3, 2013

Pretivm grade-smears through their latest release

Here's some commentary on yesterday's NR:

Mining.com - Pretivm up on a down day. Where we read:

Pretium Resources (TSE:PVG) announced more high grade gold intersections at its Valley of the Kings, which pushed the stock 5.73% higher to $6.12.

Sucks to not be long PVG, I guess. So what are these great results?

Hole VU-365 intersected 328.91 grams of gold per tonne uncut over 20.50 meters, including 13,400 grams of gold per tonne uncut over 0.50 meters;

Now, I seem to subconsciously do grade subinterval calculations in my head, because this rang an alarm bell. So I went off to the corebox drill interval calculator and plugged these numbers in and got:

20.5m @ 328.91g/t minus 0.50m at 13400g/t = a residual of 20.00m @ 2.133g/t

and that residual is waste, which makes me ask why the hell they report it. Why not just report the short high-grade interval?

Hole VU-369 intersected 96.71 grams of gold per tonne uncut over 44.95 meters, including 7,700 grams of gold per tonne uncut over 0.50 meters;

The residual here is 44.45m @ 11.18g/t which is better. Though having took first-year physics I ask whether that 0.50m high-grade interval is exactly 0.50m, or if they simply round off as a matter of custom.

After all, if the process rounds everything off to the nearest half-metre, then a 0.545m interval at 7700g/t means a residual of 44.405m @ 3.392g/t, and again the residual is waste.

Hole VU-384 intersected 331.99 grams of gold per tonne uncut over 8.00 meters, including 5,210 grams of gold per tonne uncut over 0.50 meters;

Which means a residual grade of 7.5m @ 6.789g/t. Again, assuming exactly 0.5m, and assuming exact numbers instead of carrying through error.

Speaking of which, do P.Geos go to university? If so, do they take first-year physics? At all? Are they required to pass it? Cos when I went to U, we had it beaten into our skulls that no number can ever be reported unless you include the complete margin of error. Seriously. That was the only reason we did those stupid 3-hour labs every week: to learn that there's no such thing as exact measurement of anything, and indeed that reporting numbers as if they were exact is unethical and/or incompetent.

So in every situation, you have to calculate error, and report the maximum and minimum possible values.

I mean, simply splitting a core, mashing, and pulling a sample should introduce error. Why do these numbers come without error estimations? Any professional geologists want to explain why their profession gets a pass from the proper and conscientious reporting of data? Error seems to be a major caveat in high-grade narrow-vein data.

Hole VU-390 intersected 179.14 grams of gold per tonne uncut over 13.50 meters, including 4,780 grams of gold per tonne uncut over 0.50 meters;

Which means a residual of 2.184g/t over 13.5m. I.e. waste.

Are these larger intervals (8m, 13.5m, 20.5m, 44.95m) being reported simply because that is the minimum size of tunnel or vault that will need to be constructed to mine the vein at that location?

Because hey, if you're trying to tell me that mining that 4780g/t interval will require constructing a 3m wide by 4.5m high tunnel to follow that vein, then I understand: you're saying the total bulk around that vein will still come in at 179g/t, so it'll be profitable.

But this is supposed to be an open pit, no?

Anyway, if the P.Geo associations, or maybe the BCSC and OSC, would like to initiate a process of refining the NI 43-101 to force companies to include error calculations in drill results, that would be very nice of them. Because it's first-year science. And it would mean less misleading data. This is a perfect example.


  1. Couple of comments.

    Yep, I'd bet they are actually 0.50m samples. Fairly standard for the minimum sample size. Also, if you check some of their other numbers, they're down to the centimeter. So we can safely assume they're at least that accurate. If you've ever actually measured core, its basically impossible to be MORE accurate than a centimeter, though. If you measure the same box of core 3 times, you'll get 3 different answers. The magnitude of the difference will be dependent on the skill of your geotech, and the quality of the core.

    Which, brings me to my next point. Sample error, and 43-101 reporting. Yes, there should be some way of reporting error. No, I have no idea what it should be, or how you would even go about measuring it. Yes, 43-101 reporting needs some serious re-vamp; it stamps a hard number on something that is inherently nothing more than an (hopefully WELL) educated guess.

    That said, how exactly are you going to measure error along the way? No idea. The calculation would be the easy part, the measurement of error along the way would be the hard part.

    1. I was thinking about it and realized the cores would be in half-metre blocks.

      But the "residual" being reported underneath the smear must be the little bit of the vein that shows up in the neighbouring block, right? It's still not proper to say "179.14 grams of gold per tonne uncut over 13.50 meters, including 4,780 grams of gold per tonne uncut over 0.50 meters"; they should report the high-grade block, and then the block next to it that has the remainder, and clarify that the other 12.5m is barren.

      But in that case, what does the "13.50m" represent? Is it intended to tell us that that is the width of the host rock of the vein system? I.e., that if they moved the hold 10m to the west, they'd expect to hit another high-grade vein somewhere within the same 13.5m (or more or less) interval of that type of host rock?

      Determining the error introduced by each process should be simple. You'd do experiments against a control to get an error result. In fact, I betcha academic Geos already do work in this field, it's just that so many people whine about how hard Intro Stats courses are that nobody working in the real world wants to even hear about it.

      Maybe the hard part is determining the level of error in extreme lognormal deposits, but then again, if Geos are using statistical methods to produce a pit for PVG then the stats techniques must already exist.

  2. Yeah, I mean they've got grade in the hanging wall and footwall (above and below the veins, respectively. How MUCH grade is basically what you're questioning.

    What's important to remember here though is that its all going through the mill. While it may be 'smearing', its probably more representative of what the head grade at the mill will actually be. And 200-300g/t is just fine. It would cost more money to sort it all out than it would to just mill the extra rock (not always the case, but somewhere out there there's an engineer or three who have thought about this. And they're smarter about it than I am.)

    Your second paragraph; basically yes. That's the width of the alteration/vein 'system', and in theory it would be correlative with holes all around it, in some orientation or other.

    Its not the stats that are the hard part here. Internet around a little bit and look at some core boxes; each row will have a different amount of whole core, broken rock, etc. Each one of these areas is going to have a different measurement error associated with it; different TYPES of error, and different MAGNITUDES of error. Everything from your length measurement error (which will also vary with the individual doing the measuring) to the rock density will change through these areas.

    The other issue is just the inherent variability with rock masses. Lets just think density for a minute. Every rock type starts with a different density (and even the same rock type will change density); add on to this alteration, veining, fracturing, leaching, etc, etc, and it starts to get complicated. There's no way to take enough density measurements, or accurately estimate density to account for this. You take as many density measurements as you can, and work with that.

    So I agree with you and disagree with you; its a BIG issue, but not easy to solve. And its not just a stats problem, its partially difficulty in actually measuring things to even estimate the error, and partially difficulty in the ridiculous variability within the rockmass.

    You've hit the nail on the head with 'extreme lognormal deposits', w.r.t. VoK, though. That's why they have many of the problems they do with their resource estimate. Keep in mind that the extreme lognormal bit is really only the Au grade, though.

  3. Interesting discussion here. I did some similar residual grade calculations on this here: http://www.geologyforinvestors.com/pretium-pops-on-more-drill-results-more-bonanza-grades/

    There are lots of good examples of grade smearing out there, but based on the nature of the VOK deposit I don't think this qualifies. Why report the barren intervals? Because they are not barren, although the grades are low.

    It could be argued that the reported grades are below their resource cut-off of 5 g/t, but with exploration drilling in an underground environment I think all the grades should be reported. Especially with some of the model block sizes being used (up to 10m x 10m)

    Regarding the margin of error issue....
    When I worked in oil and gas I would go to the engineer with my proposal for a oil or gas play. We would sit down and discuss the size of the resource, factor in porosity and permeability, deduct water content and come up volume of oil that could be considered a resource. Then he would ask, "What's the chance of success?" If you used too a high number they would call you a liar. If you used too low a number then you didn't have confidence in your work. Generally 80-90% was the sweet spot. The engineer then multiplied the resource size by the chance of success and voila! A bookable resource (or reserve) was born.

    My point is this: In geology, margins of error come with the territory. Its as much art as science. When it comes to real statistics and hard numbers, we blame the engineers.

  4. The sentence between your two last paragraphs: "But this is supposed to be an open pit, no?"

    Answer: No- Pretivm's Snowfield project (which is near Seabridge's KSM Project) nearby is planned to be an open pit operation. Brucejack (including VoK deposit) is to be an underground operation. This is a longhole open stoping, highly selective underground mine plan, as per Pretivm's feas study page 1-10