Cap Before or After Compositing!?

TL;DR

Cap at the smallest level of equal support (read: same interval length) that makes sense geologically. Sometimes this means compositing multiple times!

Background

If you’re looking on why capping is applied and how to sort out a cutting level check out this great Geostatistics Lesson here!

Introduction

One question LMC members get very frequently is: “Hey, so do you cap values before or after compositing?” I don’t understand how, but it seems that if you have 2 or more people in a room, there is always a difference of opinion. This is likely because there isn’t any concrete research that shows what an ideal approach is…instead, people tend to pick a side and roll with it.

My Thoughts

I’m of the opinion that the question of compositing before or after depends primarily on:

• the nature of mineralization;
• if sampling was completed using regular or selective intervals; and
• the type of drilling.

Keeping in mind that most economic variables average linearly (diamonds being an exception), I don’t think capping can be considered until the subset of samples being evaluated are of equal support.

The Discussion

The concept of a subset of samples is something to focus on here. At this point you have estimation domains and all the samples within them are going to be treated the same during estimation. So making geological distinctions between the samples within the subset at this point is just going to make your brain hurt. You’re on the highway with no exit roads, just accept that the “phat” mineralised vein is getting mixed with garbage. If it’s not sitting well, then revisit your domaining and come back here after your geological heart is broken with an overwhelming sense of defeat.

Anyways, moving on with a hypothetical thought experiment. Let’s say you have a 10 ft interval of rock that consists of 7 ft of waste with sporadic veining and a 3 ft quartz vein with 5% pyrite and 2% chalcopyrite. The interval was initially drilled with an RC and later twinned with a diamond drill hole:

Drill Hole #1 - Reverse Circulation Drill Hole

• Single 10 ft interval (I get this is big for a vein-hosted gold deposit and that it’s RC just humor me…I have seen bigger 😅
• The interval is logged as 30% qtz vn, and 2% py 1% cpy, and all the VG you could ever dream of…just jokes, that got washed away in the cyclone 😭

Drill Hole #2 - Diamond Drill Hole

• Vein Material Sample Length - <= 2 ft
• Waste Material Sample Length - Nominally 2 ft (I get this is small…just humour me again haha)

In this scenario, only the diamond drill hole can distinguish between the vein and waste rock while the 10 ft RC sample is a mix of them. Let’s say you decided to cap the DDH samples before compositing and the sample that ran 22.064 ppm Au is identified as an outlier and requires capping. After it is capped, you composite the 6 DDH samples to the same 10 ft interval as the RC hole. But the resulting composite (3.032 ppm Au) for that interval is lower then the RC sample (3.880 ppm Au) from the same interval of rock?! This brings me to my point about needing samples within a subset to be the same sample length so you can compare apples to apples. Proper QA/QC protocol, particularly in zones identified to have high mineralization potential, will provide confidence that the smaller DDH vein samples are representative. In the case of gold, insert a high Au standard right after the vein sample and collect a field duplicate. Without this verification, you may need to be more conservative then you’d like.

So what I have identified as my own general rule of thumb is: if everything is the same length (or you consider everything is equal support regardless) before compositing, cap before. Otherwise, get the samples to equal support, then cap. Maybe this means you composite twice, first to a smaller length that you cap at to minimize the amount of averaging (e.g., 5 ft) while getting enough samples to the same support, then composite again to your final composite length (e.g. SMU scale or 10 ft). As with every rule of thumb, this isn’t a prescription, use professional judgment. Don’t use this rational as justification for including samples that are obviously outliers and/or not representative.