Compositing

If you are new to Wisdom Wednesday with CW, welcome -- I'm so glad you are here to read along to my crazy brain, knowledge that I have stored - some useful, most useless (we are our worst critics), some right and some wrong (I mean all models are technically wrong ha ha) as well as networking with me, with each other, ask questions, if you don't feel comfortable commenting for all to see, send me a message. I normally answer within a day.

I wanted to talk about Compositing. Take a minute and think about why you would composite. Did you do it because you were told to? Did you do it because the software won't let you move on until you do it? Did you composite the data because "everyone else is"? Did you composite because "that's what the SOP said to do." Before moving on, put it in the comments. Just kidding since sometimes you'll get the shamers hitting up the feeds. They have nothing better to do LOL! Don't take it personally, they just want to be famous...as we all do (or maybe I don't cause then you can't be a hermit and live in peace). If you said any of these things, you are in the right place.

We don't normally sample all on the same length. Especially if you are working on a project that has went through different companies, different time periods, different knowledge, different ideas, etc. then they probably did not keep status quo and keep doing things the same way. In fact, knowledge has changed so much that you probably don't want to be doing thing the same way today as it was done 30...40...100 years ago.

If the samples were collected with different lengths, then you'll need to composite. This is to stop one form of bias entering the resource modeling workflow. Imagine if you wanted quick results to determine if the hole you planned actually hit the ore you thought and you couldn't tell from logging, it wasn't visible ore control. Then some geologists will take a "core chip" or "RC chips" or a quick type of sample in the part that they believe is the ore they were targeting (usually high grade) and they send it off to the lab, it comes back and now that small itty bitty sample that is a very biased selection, now represents the entire interval if you don't go and do a proper assessment for that entire interval. Then you know what happens... no one checks it, people forget about it, it ends up in estimation after estimation and only becomes a problem when reconciliation busts or during an audit. Other times it's when we composite different lengths when it's the metal that is of value versus the deleterious variables that we didn't care about much before. For example, gold versus sulfide. We put a lot of effort into gold sampling and assaying, trying to get the resolution that we need to make educated estimates to then plan off of and mine versus sulfide which causes havoc environmentally and in the processing stream. Sometimes the deleterious are being sampled at a scale of 2-4 times larger than the metal that we want.

What is Compositing?

Compositing is the process of combining individual sample intervals from drill hole data into longer, uniform-length intervals (called composites) while preserving key statistical and spatial properties of the original dataset. This is to ensure consistency for geological modeling, resource estimation and other analysis that may be performed. Geostatistical techniques such as variography, kriging and simulation require consistent support (interval length) to avoid bias (don't want those high-grade core chips influencing these because they will). Also, compositing better represents the deposit's spatial distribution and prevent over or under representation of intervals. Lastly, it helps simplify calculations of grade, resource estimation and block model development. Alot of software require some form of compositing to continue on in the process.

Let's say these are the interval lengths we collected down the hole. How would you choose what size of composite to use?

|--0.5--|---1.0---|-----5.0-----|-----5.0-----|-----5.0-----|-----3.0-----|--------8.0--------|

1.8 units 2.2 units 1.5 units 1.7 units 1.6 units 1.9 units 1.3 units

You probably don't have this few of data, you probably have 100s if not 1,000s of drillholes. One would probably create a histogram of the sample lengths and then choose the nominal length to use for compositing.

Up until 2015, in open pit, I would have composited to the whatever the mining height was and using downhole composites. Most of the holes were drilled from topographic surface downward into the Earth since not all projects I worked on had a tag-a-long underground deposit. For instance, if we were doing 20-ft benches (flitches), I would have composited to 20-ft because well that's the selectivity and we will be mining downward. The 20-ft would add in some dilution and put it on a constant support while the downward motion would be used to mimic the "benches" or "flitches". This was only because the software I used, it would take much longer to composite using lithologic contacts ore grade envelope contacts. We didn't have use of a software that we could do this in. Enter software. It had this option now, to not only composite at different lengths, while composting up to contacts but you could composite uphole, downhole, bench, length weighted, etc. But depending on whether you want to clip to your geologic model, that is up to you. Some of the deposits were very wide spaced drilling as well as in these areas it was a guess at what was going on and normally drawn in very straight lines (since you don't have any data), then once infilling occurs you start seeing undulation, fold thrusts and all sorts of things that make the line transform into something that looks like a tangled mess. To choose these options, it's up to you and what you see on your deposit, how you want to model it...what area do you care about.

But for now, let's say we just composite to 5.0 since that is the nominal length shown above:

|-----5.0-----|-----5.0-----|-----5.0-----|-----5.0-----|

1.87 g/t 1.7 g/t 1.78 g/t 1.3 g/t

In summary on this, before choosing a compositing scheme you need to consider geologic boundaries or the story, should you composite across boundaries, or will this bias it one way or the other? Understand the variability in the deposit and variable you are considering. Review statistics before and after compositing to ensure no bias was introduced (did you do too big or too small). Test multiple different compositing schemes to find the one that most represents your deposit.

Types of Compositing:

Fixed Length Compositing: one of the simplest forms, combines the intervals into uniform, fixed lengths (i.e., you choose 5-ft then it does 5-ft, no breaking on geologic boundary. Can be done uphole or downhole). Disadvantage: may lose fine-scale variability if fixed length is too long and can't split intervals across geologic or mineralization boundaries.

Geological or Zone-Based Compositing: Compositing occurs within a specifical geological, structural or mineralization zone to ensure honoring of boundaries is met. Disadvantage: Requires high-quality geological interpretation and domain definition and it's a bit more complex although most software has it as point, click, fill-in so not as hard as it used to be.

Weighted Compositing: Compositing takes into consideration weighting factors usually in the form of sample length or density. This ensures that longer intervals contribute proportionally to the composited grade to reduce bias from variable interval lengths, and some say accurately represents the "true" average grade. Disadvantage: it may over-smooth data in deposits with high variability, especially high-grade.

Length-Proportional Compositing: Each composite is proportional to the original sample length. Some say to use this when there is irregularly sampled data. It avoids overrepresentation of smaller or larger intervals. Disadvantage: Same as before with the smoothing.

Nested Compositing: Compositing on multi-levels such as primary intervals in lithologic units and then across larger zones usually to address variability at different scales in order to maintain geologic context. Told that it is used in hierarchical structures. Disadvantage: Significant computation and interpretation are necessary.

There are soooo many different types of compositing and advancements such as machine learning, could lead to advancements in this space. Does anyone have any recent papers they have read or used to test out newer techniques, innovations in this space? It is very interesting. I've mostly just status quo ran downhole fixed length or broken on geologic/min contacts depending on project scope and where it sat within the project pipeline, if it was surface or underground.

Remember how I said I downhole composite to 20-ft intervals in open pit. I did try comparing many scenarios and using the nominal 5-ft composites downhole. To calibrate to production, since at the time I was using production as a metric (blasthole model), I saw that I had to add 5-ft composites to get the same dilution as a 20-ft. It was interesting, got similar results just more resolution in the composites to feed the statistics, variography and estimate. For open pit, it didn't make a huge difference breaking on contact at least for the test I did. I did find it funny at times when geologists would want me to break on 1-ft intervals when we don't really mine at that level and it all dilutes/washes out especially in open pit. But we all have to see it for ourselves and learn those lessons.

If you want a good reference, go pick up Mineral Resource Estimation by Rossi/Deutsch - this is like what I call the go to reference because it's very relatable, matches a lot of the plots you see in software and its written so that we can understand. That's coming from a non-geo, non-mining person.

Thanks for listening!

#WisdomWednesdayWithCW #Compositing #Geostatistics #ResourceGeology #Mining #Geology