#44 The Duplicate Pushback

In Episode #30 The Duplicate, Gunnar famously declared that duplicate soil samples in heterogeneous fill are useless. Risk assessor Belle Casement of Terravale Consulting heard the episode… and decided Gunnar needed correcting. 😲

In this nerd-out episode, they discuss QA/QC, dodgy duplicates, lab uncertainty, conservative risk assessments, and whether the typical QC sections in contaminated land reports are actually science or just expensive pseudo scientific fluff we do to please regulators. 

Gunnar briefly ponders writing his own version of the ASC NEPM and re-introduces the term Quality Theatre in hope of getting it officially introduced into the Miriam Webster next year. 

They also get into lead, vapours, CRM samples, rinsates, trip blanks and upper confidence limits, with a completely unnecessary but somehow relevant detour into Henry VIII and Scientology. Nobody backs down, several sacred cows get kicked over, while Hammond tries hard to keep the two fighters apart and sometimes even dares to enter the ring himself. 

BTS Episode #30: The Duplicate - the OG duplicate episode

BTS Episode #11: he Sampling Evolution - Ross McFarland

Listen to Below the Surface and collect CPD points. We have confirmation that you can claim 0.5 points for every hour you spend with us under the Self Education Category

Contact: bts@4pillars.com.au

https://www.linkedin.com/company/below-the-surface-podcast

The necessary disclaimer: The views, thoughts, and opinions expressed in this Podcast are the speakers’ own. They do not necessarily represent the views, thoughts, and opinions of 4Pillars Environmental Consulting Pty Ltd or any Client, Supplier or other party related to 4Pillars or the speakers. 

(c) Gunnar Haid and James Hammond

Transcript

Transcripts are AI generated and may not be an accurate representation of what was said on the podcast.

Gunnar Haid (00:00)

Hello and welcome to yet another episode of the below the surface podcast. I'm your host Gunnar Heid and my co-host Hammond, James Hammond will join me shortly. Today's episode is a completely unscheduled one. A few weeks ago, I happened to talk to our guest today, Bill Casement, on a completely unrelated matter. And a few minutes into that conversation, I very quickly realized this was something

that shouldn't just be private between Belle and myself. So as per usual, Belle will introduce herself during the episode and there's also a little bit of a banter on how this episode actually came about. I would suggest we dive straight into it. ⁓ Belle, before we go into anything, you are a risk assessor. You work for Terravale and we use you... intensively. How many projects have we done in last six months, 10?

Belle Casement (01:01)

Yeah, about that.

Gunnar Haid (01:02)

but you're not here as a risk assessor at all. Introduce yourself as how you want to be seen, and then we can dive into it.

Belle Casement (01:11)

Sounds good.

I would not introduce myself as a risk assessor yet. I think I consider myself kind of doing a risk assessment kind of apprenticeship at the moment. I got a degree in applied science and it was a very chemistry heavy degree. After I finished that I went and got a job with one of the Enviro Labs in Melbourne. I spent about two years in one of the labs in metals and semi-volatiles and after I realised that the pay and conditions were pretty woeful and unlikely to ever really get any better.

up and I jumped the fence to contaminated land consulting and I've been doing that ever since really. I spent nearly 15 years at Senversa in their Contam Land team and in the last 18 months or so I started working with Terravale Consulting doing risk assessment work. I think of myself as more of a risk assessment support person at the moment. I can do some risk assessment stuff but it's still kind of early days in that stage of my career.

Gunnar Haid (02:09)

Have you listened to our risk assessment episode?

Belle Casement (02:12)

I did with Kylie, yeah.

Gunnar Haid (02:14)

Yes, how... was that any good? From a risk assessor's point of view.

Belle Casement (02:19)

I thought it was really good. Yeah. No, I thought it was good. I think that people don't really understand risk assessment all that well. It's kind of this mysterious.

Gunnar Haid (02:22)

It was really good.

Belle Casement (02:28)

black box that you give people like your DSI report and then somehow they magic your problems away. Like it's a pretty niche arm of a pretty niche industry so I can see why not everyone is across it but the more I learn about risk assessment the more I'm just like wow this is so useful and I wish I'd known some of this stuff 10 years ago I could have turned out some much better reports.

Gunnar Haid (02:50)

Hmm. Yeah. mean, Rod Harwood calls risk assessors, the new rock stars of our industry. You're still waiting. Here's how this all started. We spoke about, one of our risk assessment projects, because I'm one of these guys who writes DSIs and then I don't know what to do. I hand it to you and say, clear the site for me. Right. I'm one of those. I'm sure you can do something here.

Belle Casement (02:56)

Yes, that's good. I'll put that on LinkedIn.

Yeah.

Yeah.

James Hammod (03:14)

You

Belle Casement (03:18)

Yeah, I don't

want to dig this out. Please make this problem go away. No worries. I'd love to do that for you.

Gunnar Haid (03:21)

Yeah, that's...

Or as Kylie said, can you risk assess this away from me? that's exactly what it is. So we were speaking about this and then you threw in a throwaway comment where you said, by the way, you know, with your duplicates, I don't really... I liked the episode, but some of this stuff, I just don't agree. And I was like, shell shocked. First of all, correcting me never goes down well. Does it Hammond?

Belle Casement (03:27)

Yeah. Yeah.

Ha

James Hammod (03:49)

You gotta give it a little bit of time to sink in and then you come back around, Gunnar. But no, the initial reaction is typically not good.

Belle Casement (03:53)

Yeah.

Gunnar Haid (03:57)

And I immediately hooked onto this and said, okay, tell me more. And then we started talking and we started talking and I'm like, whoa, whoa, whoa, stop it. We got to put this, this is interesting, not only to you and me, this is interesting probably for more people than that. And that's how we came together and recorded this.

Belle Casement (04:02)

Yeah.

Yeah, I think as I've got a lab background, I'm really interested in how the labs work and that gap between what comes out of the lab and that data and how it gets used. And I think sometimes stuff falls down in there. And I think the soil duplicates in fill is kind of where it all, where the rubber hits the road for this one.

I think we could do better, but I think that you seemed willing to like throw everything out all at once and I got a bit alarmed listening to you. It sounded like you were ready to, you know, hurl away all replication in science forever until you got to the end of the episode and was like, well, actually, no, not quite that bad.

James Hammod (04:42)

You

Thank you, Bell, for being willing to come on and talk to us because I think we've said this before, but all jokes aside about, you know, about Gunnar’s openness to be told that he's wrong, or me for that matter, but this is exactly what we actually want. We want people who listen to what we say and what our guests say to think about it critically and if you have a different point of view, then we just want to talk about it.

Belle Casement (05:14)

Yeah.

James Hammod (05:15)

What we're trying to do with the podcast is really generate discussion, not to try and present what we say is the be all and end all of anything, but really to generate discussion. So anyone who might listen to this and has some feedback on something else that they've heard on an episode, we would encourage them to do the same.

Belle Casement (05:33)

Yeah, there was a few things that you guys talked about in the, in that podcast for duplicates. And I was like, that is not how I do things. And that is not how I've ever done things. So I don't know if it's a New South Wales versus Victoria thing, cause nearly most of my career has been in Victoria.

James Hammod (05:48)

Well, and Gunnar and I, I think we're on the same page of this, Gunnar, right? Like we definitely want everything that we say and what I guess say to be factually correct, because there's a difference, I guess, between like factually correct things and things where there's just, it's either a grey area of policy or a grey area of science that we don't quite fully grasp yet. Yeah, we certainly want the facts to be correct, but I think there's a lot of room for disagreement around between experts on other things. Definitely, definitely.

Belle Casement (06:13)

Yeah, reasonable minds can differ.

Gunnar Haid (06:17)

Come on. It's not like me that I say something on a podcast that is slightly exaggerated just to drive the point home. This is not me.

James Hammod (06:25)

I'm the non-expert here between you two so I'm just like, you know, Michael Jackson popcorn meme sitting here just watching.

Gunnar Haid (06:36)

bell of all the many warm-up questions we sent you. Which one did you pick?

Belle Casement (06:41)

was the who you have dinner with.

Gunnar Haid (06:43)

And you said you're a history ⁓ nerd, you like early European history. Can I take a guess of who you picked?

Belle Casement (06:47)

Yeah.

yeah, yep, please do.

Gunnar Haid (06:53)

Machiavelli.

Belle Casement (06:55)

No.

Gunnar Haid (06:56)

Yeah, I

wasn't ready for that turn of events.

Please go ahead.

James Hammod (07:03)

place.

Belle Casement (07:03)

I'm

quite into early modern European history. listen to quite a lot of podcasts, read quite a lot of books. So I think I'd like to have dinner with Henry VIII. I reckon he would put on a really good spread. Like he changed the course of European history. I reckon talking to him about what his thoughts on like religion and governance and stuff like that I think would really interesting.

Gunnar Haid (07:13)

⁓ That's a-

James Hammod (07:15)

Okay.

Gunnar Haid (07:30)

Mmm, divorce.

Belle Casement (07:30)

And because this is entirely theoretical, I'd like to have dinner with him at the start of the rain and at the end, because I also think that's so much change in that amount of time that I want both ends of it. I want to bookend it.

Gunnar Haid (07:36)

You're gonna have some bodyguards with you?

Belle Casement (07:44)

Aww, he only killed two of his wives. I think I'll be fine. I think dinner guests get away unscathed.

James Hammod (07:53)

That's it.

Gunnar Haid (07:54)

How does a risk assessor have dinner anyway? You're constantly worried about your intake pathway.

Belle Casement (08:00)

Nah, like I said, I think you'd put on a pretty good spread. Be good fun.

James Hammod (08:06)

If you're organic.

Gunnar Haid (08:08)

Henry the... Come on, Henry the Eighth. Really? Oh, yes.

Belle Casement (08:11)

Yeah!

Why not? Think about what that guy did. He made a whole new religion. Like, he must have been an absolute food of force in terms of charisma.

Gunnar Haid (08:20)

Yeah, but so

did L. Ron Hubbard and we still live with the kooks in Scientology, so that's hardly an achievement there.

Belle Casement (08:29)

I don't know, it was back then. It's easy to splinter off and make your own religion these days. He made it so. That's how everything splintered off from the Catholic Church in the first place.

Gunnar Haid (08:35)

you

James Hammod (08:38)

You

Gunnar Haid (08:38)

I

think it was actually quite a cool feat that he had there. He goes to the Pope and says, hey, this wife of mine, this ain't gonna last. Can you divorce us? And Pope says, no way. Just read section so and so and section so and so of our constitution here. And Henry VIII says, come on, get me divorced. The Pope says, no. And he says, okay, screw you. I'm gonna form my own religion. This is the way to go.

Belle Casement (08:55)

Yeah.

I'll make my own club then. Fun.

Gunnar Haid (09:10)

Maybe I should start writing my own nip in

Belle Casement (09:13)

Hahaha!

James Hammod (09:15)

Don't give me any ideas, pal.

Gunnar Haid (09:17)

So this whole thing we already discussed is how it started. Can I recap very quickly what's the one and a half hour episode of the duplicate actually, what I was trying to say. So see, we could have done it much, much shorter. So basically if you haven't listened to it, what's a duplicate for starters, a duplicate really is a sample that we take twice and we submit it to the lab or the laboratories as the same sample. And we hope.

Belle Casement (09:23)

Yeah.

Gunnar Haid (09:47)

that the lab gives us the same result for those two sample jars. Let's talk soil only. Let's not get into water. Let's not get into sediments. Let's just talk soil, make it easier. Why are we doing this? The point is the National Environmental Protection Measure, the NEPEM in schedule B3 in section 3.5.1 says we do that to provide a check on the repeatability of the laboratory analysis. And then the kind of...

say something different in 5.3.4 in schedule B2, they say it is to determine the repeatability of the field sampling and laboratory analysis program. There's already a slight ⁓ discrepancy. Yeah. Yes. And that's, look, let's not get into that, but basically we're trying to come to, to, to find out, Hey, how repeatable.

Belle Casement (10:32)

They're hedging their bets a bit, aren't they?

Gunnar Haid (10:46)

is our result. give you the same jar, the laboratory, but you better come up with the same or close to the same results for these analysis. That's what duplicates are and that's what they are supposed to do. My whole one and a half hour rant was for this to make any sense. Those two jars that I give the lab, they better be identical. They better be the same stuff.

And I mean really the same, not just approximately the same, because if you give the lab two different jars and you expect the same result, of course, this is going to cause a problem. And my point was when I go out into the field and I sample either fill material, mixed soils that contain all sorts of stuff, or I sample clay material or silty material.

then it is for all intents and purposes impossible for me to create two identical jars. As a result of that, I concluded that duplicates, soil duplicates taken in the field of these kinds of materials and submitted to the lab to check on the lab methods is a futile exercise. That was my whole point. Are we in agreement, Bill, on this one?

Or am I wrong? ⁓

Belle Casement (12:19)

I disagree. I don't think they're useless. ⁓

and I don't think we should scrap them. I think they do have value. We could give them a lot more value if we were better at data interpretation and spent more time actually looking at these results, these bad duplicates, and which part of them duped badly and thinking about it in more detail. You're right that the jars aren't the same. They don't have the same stuff in them. But duplication and replication is never duplicating and

the exact same thing in science. It's always a different population size, a different sample group, a different whatever, and the ability to replicate your results still matters.

Gunnar Haid (13:06)

I agree with that. Of course repeatability is hugely important, but testing it with two different jars doesn't make any sense.

Belle Casement (13:09)

Yeah.

What if we took them out of the same jar? Like is your issue with the jar or the fact that the soil is so non-homogenous that the results are likely to be so different?

Gunnar Haid (13:27)

The

second, yeah, it's the same. I cannot give the lab identical samples, whether it's in one or in two jars. In the episode, I also mentioned that when I have duplicates that have very high relative percentage differences, or when I have a sample that is just weird, I often go back and analyse some of, reanalyse some of my samples. So that's not even a different jar. That's another sub sample out of the same jar.

And the results I get so often out of the same jar are day and night. They're not even close in particular with some of our most common contaminants being the, benzopyrene that we're constantly after and the heavy metals. And there's good reasons for that. So, so it's the second, the whole concept of checking on the laboratory, which is highly controlled with a sample that is uncontrolled.

Belle Casement (14:20)

Yeah.

Yeah.

Yeah. Okay. So we're narrowing the problem down here a bit. So it's not just duplicates, it's the duplicates in the fill and it's the badly homogenized samples. So things like sand generally do pretty well. So we're really talking about the kind of the clays because it's much easier to homogenize a sand rather and in both the field and when the labs get the sample as well. So you're probably likely to have

Gunnar Haid (14:25)

is futile in my eyes.

Mm.

Belle Casement (14:51)

my experience anyway, your sand replicas will be better than something that's really clayey or gravelly or stuff like that.

Gunnar Haid (14:55)

Hmm?

⁓ yeah, Clay, gravelly, completely useless. I'm talking for me here, completely useless. Sand, even with sandy material, I would argue that a proper duplication is extremely difficult to achieve. When I analyse for benzopyrene and I analyse for heavy metals, the setup is already wrong because those contaminants attach to different particle sizes.

Once the sand is in my jar, I have no control over what happens to the finer fractions of my sand that will settle further down in the jar. And what happens to the coarser ones that settle further up. And when I take that to the lab, I have no control over which part of that sample they take out. So even with sandy material, I would argue that the duplication that we do is

Belle Casement (15:36)

Yeah.

Yes.

Gunnar Haid (15:55)

Close to useless. I tell you why. Because the moment you have a high RPD, the immediate fallback is that had to be heterogeneous. The tough one is to call up the lab and say, what the hell happened here? Because the lab will immediately say, check out our QC program. There's nothing we can do. It must have, have, it was definitely not us. Right? And they're probably right about this.

Belle Casement (16:06)

Yeah.

Yeah. Yeah.

Gunnar Haid (16:20)

You talk to the field tech and say, Hey, what the hell happened with this sample? And they say, well, what can I say? I did everything right in accordance with our standard operating procedures, whatever they are. You can't pinpoint it. The moment you have a weak point, the weak point being the sample could have actually been heterogeneous. And from that moment on, you have a weak point that everybody will jump on. And we do look, I mean, I have yet to read a report where someone's saying, well, our, our QC in the first round did not add up. So we went back out.

Belle Casement (16:34)

Yeah.

Yes.

Gunnar Haid (16:51)

Don't get me wrong, because the quality control sample is extremely important. It should tell you, wow, something's not right with my data set here. And I gotta be careful with interpreting it. But since our, our contamination is so unequally distributed in the jar, the duplication isn't a way of checking on my sampling quality. Yeah. There's always the weak point where I say, you know, it was heterogeneous and it most, most, most likely was.

Belle Casement (16:56)

Yeah.

Yes.

Yeah.

Gunnar Haid (17:21)

And as long as there is this weak point in that chain, it's all futile.

Belle Casement (17:21)

Yeah.

There's so much variability for environmental sampling that it is really hard to work out exactly where any discrepancy is coming from. This bit of sample is not the same as that bit and which...

and you transport and then you take it to the lab and the labs have their own procedures and processes that add more uncertainty and more variables in at each time. I still think that when you get that duplicate though and you look at the data that there is value in it.

if you can do the interpretation correctly. You're right in that you won't be able to pull back and say, well, this has definitely come from a bad extract in the lab or poor homogenization there, or I got a bit of metal in my sample or whatever. So it's all over the place. Sometimes you can pick it apart a little bit more, but I still think that if people were better at looking at that data, then they would find more value in it.

But QAQC is like the unloved child of the many, many appendices that we have to put together for these reports and it doesn't get the time and attention that it deserves. Which is bad on us because like the whole, like all of our reports rest on data and data quality and you don't do the data quality section until, you know, an hour before the reports due out.

Of course you have to conclude that everything is fine, you've not left yourself anywhere else to go.

Gunnar Haid (18:51)

Correct.

James Hammod (18:51)

Yeah.

Well that's it, it should be kind of flipped around shouldn't it, Bell I assume, like we really should be looking at the QAQC much earlier before we I guess get to the interpretation of the actual data set right?

Belle Casement (19:07)

Yeah, when I work with junior staff, I'm like really keen to push that as soon as data comes in, you should be looking at it.

James Hammod (19:13)

Mmm. Yeah.

Belle Casement (19:15)

Just look at the COAs, the certificate of analysis, the PDF that the lab sends you. If that's the best you can do, then just do that. Better to have it all tabulated and have those RPDs all calculated so you can start thinking about it. And any problems with that data should be caught much earlier so that you can go back and ask the labs to repeat things or re-extract things within the holding times and within a timeframe that you're actually able to influence

the data interpretation more.

James Hammod (19:46)

I just want to reiterate again that we're not talking about gases or liquids really and certainly not well mixed liquids. So we're just parking that to the side and we're mainly focusing on like you said Bill, soils and particularly fill soils that are not well mixed, at least in their normal state how we find them. You said that RPDs or heterogeneous materials that have different RPDs.

Gunnar Haid (20:12)

Relative

percentage differences.

James Hammod (20:14)

⁓ can

still be useful but it comes down to how we interpret that information. Yeah, well just to drill a bit deeper on that one, what do you actually mean by how we interpret that?

Belle Casement (20:19)

Yeah. ⁓

think people just look at what RPD is above 30 and go well, it's because they're heterogeneity but I think there's benefit in looking which metal's dripped badly.

Are these metals that you see across the rest of the site? You should be thinking about the RPDs and why one result is much different to the other in terms of the conceptual site model as well. You need to be looking at where the location was collected from. Is it near the source? Is it a long way from the source? Are you expecting variability at this point? Like thinking about...

which components have duped well then and which haven't. So like if you look at a sample that had lots of zinc and lots of copper in bad duplicates where you could be like all right well zinc and copper make brass so maybe omnia, plumbing or something like that. I think there's still value to be had that doesn't mean all the metal starter is junk ⁓ but like yeah zinc and copper

Okay, maybe there's a problem there and then this is when you go back and you get your repeat sample to see if you did get a little fleck of something in there. ⁓ But interpreting it with what you know about the site, I think can shed light on why some things look worse than others.

James Hammod (21:42)

So ⁓ we're talking about reasons other than heterogeneity there that may have impacted the set of results, but would you suggest we're looking into those things after we've considered or done everything we can to assess whether it was from heterogeneity and then if that's a no then we move on to other sources of difference?

Belle Casement (22:03)

Well, it's still heterogeneity.

It's just trying to pin down why. It's really hard to do. Like, I'm not saying that if you look at your data for five minutes, like, an explanation will jump right out at you. It's also really hard if you've only got one set of dupes to work with because that doesn't really tell you anything about the variability in the fill across the whole site.

James Hammod (22:14)

Yeah.

Belle Casement (22:25)

⁓ If you've got a huge data set and you've got a whole whack of dupes and all of them are just like all up and down everywhere we can be like, wow, this feel is crazy bizarre. It's all over the place. We need to take the highest result everywhere to make sure that we're capturing everything. I know, I know we'll get to that. We'll get to that.

James Hammod (22:33)

Mm.

Gunnar Haid (22:42)

Whoa, whoa, whoa, whoa, whoa, whoa. Like,

James Hammod (22:42)

I'm sure we'll come- we'll- I'm sure we'll come back to that.

Gunnar Haid (22:46)

they're declaring war here. This is...

Belle Casement (22:49)

I know, I'm gonna fight

you on that one too.

Gunnar Haid (22:53)

I'm gonna call Henry the 18th in to solve that problem here.

James Hammod (22:56)

That's like the fire alarm lever, we can't pull that too early in the conversation.

Belle Casement (22:59)

Yeah, no,

yeah, no, let's come back to that one.

Gunnar Haid (23:03)

Guys, I'm sorry, I need to pull you back because what you are missing here is that we are not looking for reasons why my field is heterogeneous. That's that's that's different or why I have a high result in my, in my, in my sample. That's part of the interpretation of do I have a contaminated site? The point of an, of a duplicate is to check on the repeatability of my sampling process and the repeatability.

of the lab, depending on which section of the map you read. And when, when I sample something that is clearly from the onset, not duplicatable, it doesn't really matter why it is not duplicatable. It just isn't. So my experimental setup is flawed from the beginning because the experimental setup required me to provide identical samples.

Otherwise, duplication isn't duplication. But if we have a data point and the duplicate should tell me is the methodology that I, how I sampled and how I analysed in the lab, is that repeatable? And when I have high differences between these two, there's two conclusions. Number one, everything was right, but I screwed up in the field or the lab screwed up. Let's assume that we're not checking on the, on the

that we're only checking on the lab here because this is the primary reason why we are supposed to do this. Did the lab screw up or is what I gave the lab not suitable for this kind of test? Regardless of whether I'm near piping or not, but what I give the lab is not suitable, is not fit for the purpose of duplication. In fill, in clay, in soil. Let's just put it that way in anything. I agree.

Belle Casement (24:49)

Yeah.

In anything though.

Gunnar Haid (25:02)

I completely

agree. would even go further that it's hard to duplicate water samples, but let's not open that can of worms. Because when you go to parts per billion sometimes, right? In the field, in clay material, in fill, there is no way you give the lab two identical jars. So let's stop doing it. Because the experimental setup from the start,

Belle Casement (25:18)

Yeah.

Gunnar Haid (25:31)

requires that these jars are identical. Otherwise, what are we doing?

Belle Casement (25:34)

You're right.

Look, when you frame it that way, that is, unfortunately, I really want it to not say, but you are right if you frame it that way. Environmental sampling is so inherently variable that you can't give the lab two samples that are the same to test. Shall we wrap up?

Gunnar Haid (25:51)

Excellent, so we are in agreement. No, no, no, no, but...

James Hammod (25:54)

So, ⁓ can I jump

in on that? Does it come back to the fact that we were saying from the NEPM, the purpose of field duplicates, that we're trying to achieve two different things? We're checking on the repeatability of the field sampling methodology and at the same time the laboratory's analytical process, right? So, Bill, is it pointless trying to do both things at once? Is that part of the problem here?

Belle Casement (26:19)

I mean it would be better if you could split them, but how could you possibly split those two pieces up?

Gunnar Haid (26:26)

Thank you. This is another one of those things you can't check for two things at the same time.

Belle Casement (26:29)

Yeah.

James Hammod (26:29)

So

can I throw an example at you, Bill? Say we used a field duplicate to measure or assess the sampling methodology, the repeatability of that, and we used, say, a certified reference material that's not marked and submitted as part of the sample set to check on the laboratory's repeatability. Or I guess it would have to, well...

Gunnar Haid (26:52)

Submit it once or twice.

James Hammod (26:54)

Well, I don't know. I mean, we know what the values are meant to be for a certified reference material. So you could just slip in at once, I guess, and compare the results against the known reference values.

Belle Casement (27:01)

Mm.

Gunnar Haid (27:08)

You're going for precision here.

Belle Casement (27:09)

The labs are running

their own certified reference materials. I feel like the labs do their own QA, QC type stuff. They don't give us data that didn't pass all their quality checks. I don't really think it's about, I don't consider duplicates to be, I don't consider them to be primarily a test of the lab's repeatability. ⁓ There is...

James Hammod (27:17)

Hmm.

Yeah.

Hmm.

Gunnar Haid (27:33)

No, it would be

a test of the labs accuracy. How accurate are they? And not the repeatability, not the precision. So we're not on repeatability.

Belle Casement (27:38)

Yeah, yeah, yeah, yeah, because it's a certified reference.

James Hammod (27:42)

Right.

Belle Casement (27:44)

Yeah.

James Hammod (27:45)

But

if you submitted two samples that were the same certified reference material and you compared those results, that would be assessing repeatability of the laboratory's analytical program. So let's use that as example then. So you submit two jars unmarked or unidentified that are CRM materials and that's to check on the laboratory's analytical program, the repeatability of that. And then you have a field duplicate that is used to assess the sampling program. Am I completely off track here or would that make sense?

Belle Casement (28:15)

Yeah, I think that would give you like a bit of an inside check on what the lab was doing. I also think you'll still get two different numbers back, ⁓ even if you're testing a CRM.

James Hammod (28:27)

Yeah, well even on the CRM documentation they give you a plus minus ⁓ on the reference value. I mean, but yeah, and let's not even get into that because we don't even talk about plus minus on lab reports.

Belle Casement (28:33)

Yes.

I would

love to see a plus minus on lab reports. think that people make a big mistake. They see their data and they see 100 PPM and they're like, right, locked in 100. We know it for sure. If they thought about it more in the plus or minus 20 to 30%, then it changes interpretation and thinking a little bit sometimes.

James Hammod (28:43)

Yeah.

Gunnar Haid (28:44)

Gosh

yes

James Hammod (28:48)

Mm-hmm.

Gunnar Haid (29:02)

Completely, totally, completely agree. I mean, yes.

Belle Casement (29:06)

But people just

see the lab as this magic box, you know, you put soil in, you get a report back, bing, everything is correct, you know, they wouldn't possibly give me anything that wasn't 100 % accurate, so this is how much lead is in this jar. Like, the numbers that we get out of the lab have a pretty high uncertainty attached to them.

Gunnar Haid (29:26)

Plus minus 30 % at least. I, if I, so you get your 100, 100 PPM of lead, you may want to be careful that this is not 120 in reality, but look, different, different discussion. Hammond, I interrupted you unusually.

Belle Casement (29:30)

Yeah, I would say so.

Yeah.

Yeah.

James Hammod (29:45)

No, no, I was just going to add to what you saying, Bell, and just say that I would argue that, we do the minimum level of sample preparation as well, both us consultants in the field and commercial laboratories. And we've seen that.

Belle Casement (29:56)

Yeah.

I did a project

at uni where I did bunch of sampling to test for metals in a field that had biosolids applied to it for quite a few years. I took a kilo of soil from each sample and I dried it and I crushed it and I sieved it. Each of that got extracted three separate times and each those extracts got analysed three different times on machines. Each analyte came with error bars and whisker plots. This is how much I trust each analyte for

James Hammod (30:11)

Yeah.

Belle Casement (30:27)

location sort of thing and then I went into Contaminated Land Consulting and I was like my god we only take one sample from here whoa this is what we're basing it on like yeah

Gunnar Haid (30:37)

Yes. Yeah. My, I

went to a mining university and my, my guys who work in, in mine exploration, they figured out, I tell you what, once you spend a couple of million bucks on developing a goldfield or a copper field, you know, once you spend a lot of money, you better make sure that your sampling is correct. These guys know how to sample soils. We in our industry don't. Okay.

Belle Casement (31:02)

Yeah, I bet.

Gunnar Haid (31:06)

because the problem starts with our sampling technique. I refer everybody back to the episode we've done with Ross McFarland. So then to use a sampling technique that's outlined in our Australian standards, that is utterly inadequate. And then to try and check whether the lab analyses our inadequately taken samples with a duplicate, it's a farce. It's not quality control, it's quality theatre.

It's a joke, what we're doing here.

Belle Casement (31:38)

I don't want push it that far. I think maybe it's just because I've spent hundreds of hours of my life reviewing these bloody things that I'm like, don't put it all in the bin,

Gunnar Haid (31:42)

You

James Hammod (31:51)

So could you give us, so Belle, you give us, I guess, overview as to noting all the ⁓ imperfections or issues that we've talked about in what we do. So how do we get the most we possibly can out of field duplicates right now? So that's your take, Gunnar. I want Belle's take.

Gunnar Haid (32:05)

Don't take them. Save yourself the money and save your client the money

Belle Casement (32:07)

⁓ it's hot, it's hot.

Yeah.

Gunnar Haid (32:13)

and save yourself the time of having to explain this nonsense. ⁓

James Hammod (32:18)

Nah,

I wanna hear what Bell says as to how we get the most out of them.

Belle Casement (32:22)

You need to look at your results in a timely fashion, which we talked about. You need to have someone with like more than two years experience looking at some of this data. think that I know I'm asking way too much here already.

Gunnar Haid (32:32)

my God, you're, that's another sacred cow in our industry. Keep going. Yeah. It's like

you want someone who actually knows what they're doing to look at data? out of here.

Belle Casement (32:41)

Yeah, that's very unfortunate. People need to know what kind of things can go wrong so that they kind of know what they're looking for. If you just teach people to write off their RPDs as non-homogenous fill, like, then they never really learn about the other things that could go wrong. Notionally, this is sort of checking for, like, is it carryover or cross-contamination within the instruments, within our instruments in the field? Stuff like that needs to be taken into

account but a lot of people don't know how to check for that because like who teaches them that of course there's not really any time for that.

James Hammod (33:18)

Yeah. And

do you have any views on the consultant's role, the sampler's role in sample prep? Like, do you think there needs to be more preparation and homogenization of samples? I guess obviously for non-volatiles, but do you think there should be more of that happening?

Belle Casement (33:28)

Yeah.

Yeah. I

did not know what a riffle box was. You guys made sure you the last thing. We were like, we should put it through this spot. And I was like, what the hell is that thing? 20 years I've been doing this work. I've never heard of that before. Yeah. Yeah.

James Hammod (33:43)

The magic box.

Gunnar Haid (33:43)

How many years at Saint-Versa? Okay, this puts things into perspective, because I have never used one either. And everybody always

says, we are sampling in accordance with Australian standard bullshit dot bullshit, which we don't! Nobody does! Anyway, keep going.

Belle Casement (33:56)

Yeah. Yeah. Yeah.

James Hammod (34:03)

So that's a yes, we should really be looking at the sampler doing more in the way of sample prep.

Belle Casement (34:10)

Well, I don't know where to focus the efforts on, like, because we don't really know where all of these problems are best fixed and best tidied up. I think probably it's that drying and crushing and well mixing that you're not getting ⁓ in the kind of work that we do where a lot of the variability comes into for the sampling of fill. But like, how could we do that?

James Hammod (34:19)

Yeah.

Yeah, so is that on the lap side?

Belle Casement (34:37)

I think it's maybe even before it gets to the lab. I think it's more of the field scientist's job to be doing that rather than the lab.

James Hammod (34:40)

Mm.

Yeah,

when I've raised that, particularly with regulators, there seems to be this concern that a consultant is somehow fiddling with the sample when it goes through a process of sample prep that is controlled by the sampler or the sampler's organization. Do you have a view on that?

Belle Casement (34:58)

Yeah.

I wouldn't have thought that at all. mean, you don't want people to just like take their soil out of the push tube and put it in a jar in like a little block. That's not... yeah, it is fast and it's easy. And that's why it's happened because you're the poor field scientist is, you know, three logs behind and they're sprinting past the drill rig trying to keep up all the time. I mean...

James Hammod (35:10)

Hmm. That's so often what happens.

Yeah.

Gunnar Haid (35:25)

And it's

5pm and it's getting dark and the day was hot.

Belle Casement (35:28)

Yeah, it's a big ask for them to be crushing and homogenising out in the field.

James Hammod (35:33)

I'm

a big believer in that there should be more sample prep and this is one of my little wagons I like to push but every time, well not every time, but a lot of the time when I've raised it there's concerns raised around this kind of chain of custody and that certain things need to be done at arm's length and yeah that's one of them that seems to come up particularly with regulators.

Belle Casement (35:53)

Yeah, okay.

I mean, it's pretty tough when you're doing any volatiles analysis. That just throws another huge spanner in the works. You obviously, in that case, it might be better to just put your push tube plug in and close the lid as fast as possible. then when that gets to sample prep in the lab, then they're like, are you kidding me? Just a tube, guys? And then they have to get it out and it's gone all crusty and dry and they're like hacking it up. It's not great science. Yeah.

Gunnar Haid (36:21)

So we talked

about CRM, certified reference materials. When you buy them, you get a report with it. On that report, they tell you, look, there is X parts per million of zinc in this particular crushed up ore. And we have measured it with everything that we can possibly come up with in the world. The most spectacular laboratory methods. have homogenized it to the

Belle Casement (36:27)

Yep.

Gunnar Haid (36:48)

best of our ability under laboratory conditions. And here is your range, not the number of, here is not the number of how much zinc is in there. Here's your range. And that goes from 80 to 120 parts per million of zinc or something along those lines. I don't know exact numbers for zinc. That's the best we can do with all the methodologies we have with all different lab methods.

Belle Casement (37:08)

Yeah.

Yes.

Gunnar Haid (37:17)

And we cannot determine how much dam zinc is in this highly, highly homogenized and analysed sample. And then we go out in the field and I take my, you say push tubes, very often I even sample, Hey, everybody does it. It's not just me. We sample from the auger. And we fill that stuff into our sample jar. We wipe it off. You put that jar in the esky off to the lab.

Belle Casement (37:37)

Yeah.

Gunnar Haid (37:46)

This method, we are kidding ourselves in getting anything of value out of that kind of sampling. But that's a, we already done the, we have done the episode. It's with Ross McFarlane. Our sampling method as outlined in the Australian standard is inadequate. That is however, a holy cow in our industry that nobody wants to, what do you do with a holy cow? Slaughter. That's not, know, anyway.

Belle Casement (38:01)

Yeah.

Gunnar Haid (38:16)

Can we, let's move on. There is a couple of other things you mentioned in your, in your ⁓ emails to me. Do we want to talk Rinsate and Trip Blanks? I mean, I'm happy to.

Belle Casement (38:28)

I mean,

I'm happy to. I think they can be quite useful. ⁓ I don't know if you've got a controversial opinion on them that you'd like to get out there and start airing to people, but I think they're okay. Like, if used and interpreted correctly, they can add value.

Gunnar Haid (38:45)

Okay, let's for non geeks, so let's talk about trip blanks for starters. I assume we're talking water here, correct? Okay.

Belle Casement (38:52)

Yeah.

Gunnar Haid (38:58)

Because we don't trust soils in the first place in a sample jar. I'm just throwing this out, but we're talking waters here. Why would I take a trip blank? Because on a trip, a lot of things happen. You have water samples and they are usually done for, or often done for highly volatile substances. And highly volatile substances have...

Belle Casement (39:05)

Yeah.

Gunnar Haid (39:25)

a problem because they're volatile and they get into all sorts of they move around in your esky they move around on your service station where you take your sample they move around in the environment and particularly when you put samples into an esky and it gets particularly nasty when and it happens one of your sample jars breaks I so you end up with a soup of of water dirty soil jars and It that is not a good thing in any case

Belle Casement (39:27)

Game of rounds.

Yeah.

Yeah.

Gunnar Haid (39:54)

That's a bad thing to happen. So if you then have in your esky, you have a known or a water container where you know there is nothing in it and it floats around in your soup that you accidentally created, at least it gives you an indication when you analyse this, it gives you an indication if that particular sample jar or water container was affected by the mistake I made. And I get it.

Belle Casement (40:21)

Yeah,

think field trips like that are pretty good and can be pretty useful.

tend to be analysed for actually the right things. Like people at a minimum tend to do C6 to 10 analysis on them and if you've got the cache people do a full VOC screen on them. So you actually get some useful data back that says like say something has gone wrong and you get a little bit of TCE in there then you should be looking you know with a slightly raised eyebrow at some of your very low TCE hits in other samples and be like well maybe that's not a real hit maybe we need to think about whether or not the plume is really out here.

are.

Gunnar Haid (41:01)

I agree. I don't

mind trip blanks. It's the same with almost the same with ⁓ trip spikes. A trip spike is now the same container, but in this case, it's not a blank. It's not free of contamination. It actually has a known amount of a volatile in it. Why would we do that? You know, sometimes it's hard. You're out near Birk and you collect highly volatile samples and you need to make sure that they are cooled.

All the time and that while in the, during your three or four days of sampling for they actually managed to get to the lab that you don't lose some of your analytes in the sample jar. That's where trip spikes come in handy. And I do use them quite regularly and religiously on these kinds of sites. Mind you. Yes. Yes, they are. Yes, they are. In particular, when you come back and.

Belle Casement (41:50)

Yeah, I think they're one of the better samples for QAQC.

Gunnar Haid (41:59)

from a trip like this and you see the lab results and you think like, really? That result is awfully low. Right? It has happened. And you then go into thinking overdrive and think what could have possibly happened? One of the things that you then immediately look at is your trip spike. At least you can say, all right, it's at least, I think, and I'm confident.

Belle Casement (42:06)

Yeah.

Gunnar Haid (42:24)

that my sampling procedure, not the sampling procedure, but my sampling holding procedure was all right. Mind you, something that people forget is that the trip spike makes two trips. It makes a trip from the lab to my office and out to the site. And the control samples, the samples that I take on the site only take one trip. So there is again a factor in that.

Belle Casement (42:39)

Yeah.

Yeah.

Gunnar Haid (42:51)

If your trip spike comes back low, so you have lost some volatiles in your trip spike, the question then is, have you lost it on the way there or on the way back? Not something that gets ever discussed and is a clear weak point of trip spikes.

Belle Casement (43:06)

So if you had a trip spike that came back and it was 20 % less than what the lab had spiked it with initially, what would you do with your data? What are you doing? Are you writing a little QA, QC thing? Are you putting little caveats around your data set? Are you saying we think we've lost maybe up to half of what should have been in there?

Gunnar Haid (43:11)

Mm.

⁓ something like this would definitely trigger me to contact the lab and say, guys, can we have a look at this again for starters? Because the likelihood that I screwed something up with a closed BTEX vial on the way out to Burke and back is it's possible. But, ⁓ let's put it this way. At least it's questionable. Now ⁓ I can look at the, was it me?

Belle Casement (43:33)

Yeah.

Yeah.

Yeah.

Gunnar Haid (43:55)

Was it the lab or, did it affect the rest of my results? I then have to go and look, even if I can't get to the bottom of it with the lab, I will have to go and look at my other samples and say, are they in line of what I know about this site? If I have been there before or aren't they? That's when a little bit of experience and strong thinking comes in. But if it's, if it's a site where you've been for the first time,

Belle Casement (44:14)

Yeah.

Yeah.

Gunnar Haid (44:25)

⁓ I would definitely in my QC section, not only in my QC section, talk about it. And then you have to look, what's my site. How critical of how critical is this data? Am I signing off on something highly volatile that I'm discharging into a, into a creek? And it's really critical that I, that I know. And in this case, I would not hesitate to say the data is to be taken with a grain of salt.

I do that, fact. Thank God it doesn't happen too often, but I am not saying that quality control is all nonsense. No, no, in a case like this, it is vital. You need to think about what happened here.

Belle Casement (44:55)

Yeah. Okay.

Yeah, okay. That's good. I think people are really reluctant after they've done their data review to...

chuck out any data or even reduce the reliability of certain subsets of data despite what QAQC interpretation sort of tells them, which I think is a bit of a shame. But I think it would be more accurate and true sometimes if we got to the end of the report and was like, look, we're really happy with all of this, but you know, the B-TEX data is, we think we're probably under-reporting the next round. You know, we need to do it faster or change something or to try and make sure

that we don't do that again. You hardly ever see it in reports though.

Gunnar Haid (45:52)

The good thing about trip spikes is that in their water, so you are on a site where you take water samples, very rarely are you on a site that is contaminated where you're only there once. So it is actually a very good flag. And you say, don't have to throw this data set out now, but let's look at it in the overall picture and have an asterisk on this particular set. It doesn't mean that it's thrown out, but it means it has to be looked at in the overall

Belle Casement (46:02)

Yeah.

Yeah.

Gunnar Haid (46:21)

site assessment. So I have no problem with all of this and it should be absolutely should be done and needs to be done, but it is not something that can be done in a spreadsheet. That's something that needs to be done by someone who knows the site and who knows data and data analysis and who knows how to read a lab report and who has experience on that site. No question. Rinse aid samples. Sorry, I haven't.

James Hammod (46:41)

And... and...

I was just gonna say, just on that, I mean, yeah, you talk about people's reluctance, I guess, to...

question the outcome or the conclusions from a report because of concerns around data quality. But part of that I'm sure is about not wanting to remobilize and not wanting to upset the client or take additional time or whatever it is. But there are lots of real situations that we see where someone might take the easy route at that point and accept the data as being sufficient quality and then draw their conclusions from it and save some time or headaches there. But it can create massive headaches down the track.

a project that comes to mind where we had an investigation by another firm, an upstream groundwater well with very high lead, very, very high lead, then downgradient wells with no, like below detection for lead, with a basement that was going to intercept that. And the conclusion was made that that upgradient well, that the groundwater did in fact contain very high lead that was of concern to human health, and therefore there was going to be an LTEMP was going to be required because it was a drain basement.

Gunnar Haid (47:48)

Sorry, ill temp.

James Hammod (47:48)

long-term environmental management plan. So the decision not to drill into that potential cross-contamination or whatever it was that caused that, that resulted in that data point, not scrutinizing that enough at the DSI stage then led to the potential for a long-term environmental management plan and all the headache and cost and issues that come with that throughout the entire life of that building.

Belle Casement (48:14)

Environmental sampling is too variable to ever rely on one data point to make judgements on the outcomes for assessments. It's just too much question marks on everything. That's why we do things with multiple rounds of everything. There's just too much variability in it.

James Hammod (48:26)

Yeah.

that's it we should we got it you we should

click that bill that's like a does a very very succinct excellent point

Gunnar Haid (48:34)

It's TikTok worthy.

Belle Casement (48:34)

Yeah.

Yeah. If we must, like, I don't love them. I think they're okay for groundwater sampling ⁓ if you do the right analysis suite, but I think putting running water over a metal hand auger and testing it just for metals and then asking yourself whether or not the solid matrix was clever. Yeah, nah, waste of time.

Gunnar Haid (48:39)

⁓ Rinse states.

I'm not a fan.

Well, yeah, that's.

James Hammod (49:06)

I

find it hard to catch the water coming off the of the tool or the implement. I don't know how people do it.

Belle Casement (49:09)

Yes.

Gunnar Haid (49:10)

⁓

It's completely useless, it doesn't show anything.

Belle Casement (49:15)

Yeah,

I don't mind them off the groundwater pumps, but I think they're pretty useless when you're switching matrices like that.

Gunnar Haid (49:23)

I can live with yes on soil augers forget it

Belle Casement (49:24)

Yeah. But again, they people are,

they're too cheap to test for anything except metals. So you get back a few tiny little hits of metals and then you're right. The pump is made of metal. See you later. It's, it's the groundwater version of non-homogenous soil. Let's move on.

Gunnar Haid (49:40)

Well, okay, so

what's the RIN state supposed to tell us? The RIN state is supposed to tell us whether our sampling equipment was properly decontaminated between sampling points so that I don't carry contamination from point A to point B. I get it. Now, if you take a RIN state and you do it in accordance with the Australian standard, you should take, is it one in 10 or one in 20? See, I don't know because I never take them. Whatever it is. Is it one in 20? ⁓

Belle Casement (49:51)

Yes.

James Hammod (50:05)

one in 20 I think. Yeah.

Gunnar Haid (50:09)

So one in 20. All right. Science goes out the window because do you think that the rinse aid that I take or that my decontamination of my auger on the hole where I take the rinse aid is going to be the same as on the other ones where I'm not tested on. Come on, let's not kid ourselves here. This is, this is all nonsense. Rinse aid samples need to go for, especially for soil sampling. Water and pumps. Yes.

Belle Casement (50:28)

Yeah. Yeah.

Gunnar Haid (50:38)

I am with you. Okay. Let's go to the real hot button.

Belle Casement (50:46)

I think we need to talk about taking the highest result.

James Hammod (50:50)

Yes!

Gunnar Haid (50:51)

So you have a duplicate, you go to the field, you take two samples, you submit them to the lab and the results are quite significantly different. Your lead has 180 and 420 parts per million. In the first episode of the duplicate, I argued that this relative percentage difference

It should trigger something, right? You can't just ignore it, that's for sure. What it shouldn't trigger is, hey, don't worry about it, that's heterogeneous material, like that's what we always say, and it's all good. That's quality theatre, we all agree on that.

Belle Casement (51:36)

Yeah, but if

we've done our proper data checks and there's no clumps of lead in the soil and we've, you know, someone who's a good professional's had a look at it all and been like, no, we think these are both two like representative results of the soil that we submitted to the lab, then you should use the highest result.

Gunnar Haid (51:56)

Absolutely not. No, no, no, no. No,

no. Why, why, why the higher one? Why not the lower one?

Belle Casement (52:02)

Because we're

assessing threshold compounds or analytes that can have a toxicity response over a certain level. And if we think that it's a real data point and is representative of what the soil looked like in that spot, then we should use the highest one to protect site users.

Gunnar Haid (52:08)

Mm-hmm.

Mm-hmm. Mm-hmm.

Well, number one, it's not representative because you have two samples, right? So one of them isn't wrong. One of them isn't right. We have two different samples. So the representativeness is already out the window because it ain't representative. It's representative of what's in that jar, but aren't we trying to find out what's representative across the site? That's the most important thing. Go ahead.

James Hammod (52:48)

Can I just, what you just said then,

Godan, so you said two samples, but isn't what we're saying that we have two data points that are essentially the same sample, right? That's what we're saying. Yeah.

Gunnar Haid (52:58)

That's what a duplicate is.

Is it represent... I can guarantee you, or in most cases, yes, what the lab took out of there, the result is very representative of those three grams or five grams that they took out. All right. Now to take the bigger one here is you are picking and choosing data. You are for some reason, you say, I like this one better than the other one.

You can't do that in science. That's not how this works.

Belle Casement (53:31)

But I can do that because I'm doing a risk assessment on it and I want to make sure that this site isn't going to make people sick and we know that they get sick above this level so I'm concerned about results above this level and you got a result above that level that you think is real so why would you throw it out?

Gunnar Haid (53:50)

I'm not- ⁓ never said anything about throwing it out.

Belle Casement (53:52)

Sorry, why would you

not include it in the data interpretation?

Gunnar Haid (53:56)

I include it in my data quality analysis. I include it in my quality analysis. I do not include it as a data point because are you sure that when you say you are risk-assessed, are you sure that that sample, the higher one is in fact the one that's representative? Are you sure about that?

Belle Casement (54:01)

Yeah.

No, we take the worst case scenario to be protective of people's health. To be protective.

Gunnar Haid (54:22)

Why?

No,

no, no, no, no, no, no. You are supposed to be representative. You're not supposed to pick and choose and say, I'm going to be conservative. That's not, that's not what we're here for because this has, this has ramifications. If say aside, we take 50 samples and we run every sample twice. And I, I take every time I take the higher of the two, or I take the lower of the two. There's going to be a

significant difference between the two of them. We can not definitively say which one of the two is more representative of the site. The most representative one, if you ask me, the most representative one is if you take the average of those two, then we are getting at least closer to what is most likely at the site, which we will never know what is at the site.

Belle Casement (54:56)

Yes.

Yes.

Gunnar Haid (55:23)

Hence we do all these, these statistical calculations, but you cannot go and pick one of them out and say this one, I take the bigger one. Just think about when you calculate your upper confidence levels. Okay. Again, my site, I have 50 samples. have two duplicates in there and both duplicates and everybody and including Bell. Now you need to think with me. Okay. You need to picture the Bell curve and my duplicates are

Belle Casement (55:23)

Yes.

Mmm.

Gunnar Haid (55:51)

all on the left-hand side of my bell curve. So they are all, they're below the mean and they all fall below there. The upper confidence level takes into account the number of samples, the standard deviation and the mean. Now, if I now take of my two duplicates that are both on the left-hand side on the bell curve, if I take both of them, I take the higher one.

Belle Casement (56:00)

Yeah.

Gunnar Haid (56:21)

There's two things I do. I increase the mean, which is, I guess, what you want to do because you want to be, I'm drawing here quotes here, conservative. Right? So you're increasing the mean and that's already questionable for me because you're already, okay, so you're increasing the mean. But what you're also doing is you're moving those data points closer to the mean and you are reducing the standard deviation.

Belle Casement (56:30)

Yeah.

Yeah.

Yes.

Gunnar Haid (56:47)

So you got to be really careful with what you're doing here because you may end up with a lower upper confidence level. Not the kind of fiddling I agree with.

Belle Casement (56:55)

I think it depends what you're doing with your dataset though. Like this sort of calculating the mean and the...

upper confidence limits and stuff. Like you see that all the time for waste classification. I think the risk assessment response to this is that we know that somewhere out there on that site is a piece of lead that is above the HILA and it is incumbent upon us to protect future site users of that kid that walks along and eats that bit of soil. ⁓

James Hammod (57:26)

Can I ask a question on that, Bill?

I think where I'm getting a little bit confused here is my understanding has always been that unless we're dealing with something that is like a very, very clear point source, very localized issue, when we're assessing soil, call it a fill layer or whatever it is, or, you know, bedding sands around a tank or whatever it is, like we look at things on a decision unit basis, which is like a mass of soil that covers a certain amount of volume, right? Or area. Is there a difference in what you're saying when you're talking about one sample location versus

a decision unit that will have multiple samples in it, ⁓ one of which may have a duplicate associated with it, but we're running statistical analysis for that whole data set to cover that whole decision unit. Are you saying that from a risk assessment point of view that you pay more attention to those individual results and individual exceedances? Is that kind of the process?

Belle Casement (58:19)

Yes,

the risk assessment will take the highest concentrations to model the risk.

Gunnar Haid (58:25)

Okay. You got to be kidding. That shouldn't be the case.

Belle Casement (58:27)

It

just does. That's how it works. They take the highest.

Gunnar Haid (58:31)

That doesn't mean it's right. Why do we even take UC upper confidence level calculations? Why don't we just go out and we take 50 samples and we take the highest and say, whoa, this is a contaminated site. That's not how it works.

Belle Casement (58:41)

I don't know, I just follow the guidance that I'm given and all the risk assessment stuff and the practice within the industry is to be protective of everything and you're right, I'm trying not to say it because you're going to jump at me, but it's to be conservative and to protect everyone, we take the highest one.

James Hammod (58:43)

Well, are we?

Hmm.

I might be misunderstanding, but think we're talking about, are we talking about two different things here? Cause we're talking about on one hand, a data set and statistical analysis to compare a value, which is the 95 % UCL against the HILs or the EILs, whatever it is, as part of a tier one screening, risk screening process. And then as we've talked about with Kylie, following that, if there's exceedances, you then go through to a tier two, tier three, that sort of, that risk assessment work that you do.

Are we talking about two different things here? Are we talking about the fact that we use 95 % UCLs all the time for decision units to make a tier 1 assessment, but you're saying that you don't use 95 % UCLs for a tier 2-3 risk assessment, you rather use the highest data point for the site or for that particular decision unit. Is that the way that works? Right.

Belle Casement (59:52)

Yes, we would take the highest concentration. Like

if you are looking at soil vapour data and there's four bores and one of them will have the highest concentration, we would take that highest concentration to model the potential risk. And if there was a higher value in a duplicate sample, then we would take that.

James Hammod (1:00:05)

Right. Is that?

Yeah.

Okay, so that, I get that and it sounds like that is articulated in risk assessment guidance or standard practice somewhere.

Belle Casement (1:00:22)

⁓ I tried to find something in some of guidance documents. I couldn't find anything particular, but certainly all the risk assessments that I've been involved with work with maximum concentrations to model risk.

James Hammod (1:00:33)

Okay, so there's a practice at least in risk assessment. So Guna, does that help answer the question? think we're talking about two different things here. So, Belle, would you... So if we're about, say, what we're doing, which is we're using a 95 % UCL that provides a single value that represents an entire data set, or, know...

Belle Casement (1:00:49)

Yeah.

James Hammod (1:00:51)

and we're using that value to assess against the tier 1 criteria, are you suggesting that in that process we should also be incorporating the highest value of a duplicate into that data set or are you saying that just for your work in risk assessment you need to use it?

Belle Casement (1:01:07)

It sounds more like someone who is better at maths should probably answer that question because I feel like by the time you're doing calculating upper confidence limits you're already doing something to the data so I can see where you're coming from that you're a bit low to pick and choose the data that you put in the data set for that.

James Hammod (1:01:26)

Yeah.

Well, yeah,

that's where we're coming from. And sorry, just to add to what I was saying is that ⁓ another issue that we've had come up when we talk about using... So if we have 10 samples and one of those samples has a duplicate associated with it, your data set is then not 10, it's 11. And one of the issues that I've had, again, with money speaking with regulators around this stuff is that I'm not an expert in this, but I've heard that referred to a pseudo replication. So you've got multiple results for a individual sample.

Gunnar Haid (1:01:29)

For starters, yeah.

James Hammod (1:01:57)

sample but you don't have that same number of results for the rest of the samples and so you're pseudo replicating one and then therefore you're weighting that sample more heavily. So like...

Belle Casement (1:02:04)

Yeah. Yeah. So you would drop

one of those, like you generally don't have two, you generally have three, you have the duplicate and the split, the inter and the inter-lab sample. So you've often got three results to choose from. And if I was doing averages or UCLs or whatever, then I would take the highest to put in that data set. And that was the way I've been taught to do it. And that's the way every DSI I reckon I've ever done has, has used that method. ⁓ and I don't think.

James Hammod (1:02:13)

Yeah. Yeah.

Okay.

Interesting. Interesting.

Gunnar Haid (1:02:32)

You're fiddling with your data.

You're fiddling with your data. You are not... Not the point!

Belle Casement (1:02:36)

I'm being conservative. I

just said that to make you cry.

Gunnar Haid (1:02:42)

⁓ Again, to risk assess a site based on the highest value is preposterously stupid. I'm sorry. That is nonsense.

Belle Casement (1:02:50)

Take, you should

take that up with someone who's got more experience with risk assessment than me, but I'm sure, like, this is what everyone's doing.

James Hammod (1:02:57)

But I can understand it for a risk assessment, Gwyn, I'm sorry, just to be, I don't know. ⁓

Belle Casement (1:03:02)

Yeah, because if the kid could

eat that clump of lead, you might have the old ground floor apartment that's sitting on top of the old tank. Like, you want the worst case scenario modelled.

Gunnar Haid (1:03:11)

Yeah,

so freaking what? Over how many years is that gonna happen? The child is not gonna go out there every day and eat that piece of soil there. That's the whole concept of...

Belle Casement (1:03:20)

I guess works better for vapour

than the dirt eating, but yeah.

Gunnar Haid (1:03:23)

of prolonged exposure. Vapours, even with vapours, I come on, it's going to get into your basement based on the highest concentration. That is just so far away from what is actually happening in reality. And being conservative is one thing, but being ridiculously unrealistic is another one.

Belle Casement (1:03:27)

Yeah.

Yeah.

Yeah.

risk assessments

know they're conservative. They know the whole way through that everything they do aims to be conservative. And it's only if they get to the end and go, there's a potential risk here, then the modelling gets more granular. And then you might need more data, or then you might need more repeats. And then you might be like, well, okay, that super high result that we had six months ago, we don't think it's representative of site conditions. Actually, we're going to try and do something else now. So it's not that they're just taking the highest money in every case forever, but you

You start there and then you move forward and then you funnel down to try to get to a point.

Gunnar Haid (1:04:20)

The super high result you had in the first place isn't representative for starters. It's not a representative sample, it's the maximum. It's not representative. It's like you're going into a room and you want to know how tall are the people in that room and you go in and you deliberately pick the tallest one and you say, this is a room of tall people. It's just nonsense. This just doesn't make any sense.

Belle Casement (1:04:38)

Yeah, but that tall

person, like if they're above a certain height, that tall person isn't making you sick. We're trying to protect people from getting unwell on the sites that we tell them that it's safe to be on.

Gunnar Haid (1:04:45)

No no

Again, nobody will inhale the maximum all the time. It's just not happening. It's just not happening in reality. Anyway, I okay, I'll leave you risk assessors to assess risk, but jeez.

Belle Casement (1:04:56)

No other words. Yeah.

James Hammod (1:05:00)

I think... But can I... can I... ⁓

Belle Casement (1:05:01)

Yeah.

You still bringing me for help then?

James Hammod (1:05:12)

Can I

just make another comment on this point, which is that I kind of do find myself attracted to this idea of having a very clear distinction between what is a primary sample and what is a duplicate and what are the roles of I find that a neat thing to put around it. So is it not the case that if we had the two results, if we had the lead at 180 in the primary and 420 ppm in the duplicate.

Shouldn't it be rather than say, okay, we're going to stop there and that's our data set and we should pick the higher one to be part of that data set. Shouldn't it be that we're saying, clearly something's going on here, there's a big discrepancy between these two results. Shouldn't we be saying our data set isn't ready to be used for analysis, that we need to do further work to drill into why is there such a big gap here?

Belle Casement (1:06:03)

Yeah, probably. Yeah, and I think we do get to that point relatively well as an industry. No one is having one small HALA exceedance and being like, that's it, dig it out boys, you know, it's no good. We are, if you had just like one exceedance and it was in a duplicate, I would still hope to see that discussed in a report and mentioned. ⁓ I think that would be a reasonable approach to take.

But I don't think you should just pop all of those duplicate data numbers in your QAQC appendix and not talk about them anywhere else. I think they still should be, unless you've got a reason to exclude them as erroneous. you know, we thought there was, you know, chunks of lead. We tested it four more times, no lead came back. So we're saying that that super high hit is junk and we're junking it.

James Hammod (1:06:33)

Yeah.

Gunnar Haid (1:06:50)

Mm-hmm. See...

James Hammod (1:06:51)

Hmm.

Yeah. Now that's

Gunnar Haid (1:06:55)

Now

James Hammod (1:06:57)

an interesting point there. Yeah.

Gunnar Haid (1:06:58)

you're talking to me, we tested it four more times. That's the correct response.

Belle Casement (1:07:02)

Yeah. Yes.

James Hammod (1:07:04)

I think it's

one of the approaches you could take.

Gunnar Haid (1:07:08)

Yeah. So you run the sample five times. What do you do with these five results? Sorry, one sample, one sample jar five times. What do you, you know what I do, right? I average them out. I have no problem doing that. can, I can justify doing that. What do you suggest you do?

Belle Casement (1:07:14)

Yeah.

I would take the second highest. If you've gone through and done all of this rework...

Gunnar Haid (1:07:28)

The second highest?

Belle Casement (1:07:30)

Yeah, well would take, it's the same thing, but multiply it up. So you know, if you've got three results, one that's sky high because of a fragment of metal in it, you've gone back, you've re-extracted, you've re-analysed a bunch of times, and everything else is kind of within an acceptable RPD limit, then I would take the next time's one.

Gunnar Haid (1:07:48)

Mmm, okay

Yeah, okay, that's actually a very good interpretation. If you have one that is at 1500 and the other sit at 200 or something, I agree probably that the one is a freak. But again, you cannot ignore it. That's the thing. You cannot say, well, know, that was a medal because what if your little poor kitty goes out and eats that piece of medal, right?

Belle Casement (1:08:01)

Yeah.

Yeah.

Gunnar Haid (1:08:22)

I would argue when you take a sample and you present the sample to the lab, what the ultimate goal is, I want to know what is the concentration of lead in this sample, not in five grams of sub sample. If we had the ideal scenario where I say, I submit this 250 millilitre jar, they have 500 grams, we can run this sample a hundred times until we run out of matrix.

Belle Casement (1:08:34)

Yeah.

Yeah.

Yeah.

Gunnar Haid (1:08:51)

we get 100 results. Wouldn't you think that the most accurate and most representative analysis and most representative concentration would be the average of those 100?

Belle Casement (1:09:06)

Yeah, is cool, but that's not how works, that's not the data set we have, so...

Gunnar Haid (1:09:11)

No, I understand,

James Hammod (1:09:11)

Mm.

Gunnar Haid (1:09:12)

but

James Hammod (1:09:12)

Mm.

Gunnar Haid (1:09:12)

you have your own of 100, you have five and not take the average. hear you if you have, say you have, ⁓ okay, I hear you when you have one that shoots through the roof, but you still, can't take the second highest one because you're basically in this case, you're ignoring four of the five. You're not taking them into account. There are valid points. are information that you have gained about this site that you with some

Belle Casement (1:09:15)

Yeah.

Yeah.

Yeah.

Gunnar Haid (1:09:40)

arbitrary decision that you make privately throw out. Doesn't work that way. You can't do that.

Belle Casement (1:09:49)

Yeah, sometimes it doesn't matter at all and sometimes you have to spend a lot of time and money and effort drilling down to find exactly what's going on in that location. If all of your samples, even the sky high one, is well below, say it's zinc and not lead, well, who cares? It's zinc! I mean, never mind the poor aquatic ecosystem that's probably nearby, but you know...

Gunnar Haid (1:09:59)

Mm-hmm.

I will then. We don't have to worry about it. Yeah. Yeah, we don't have to worry about it.

James Hammod (1:10:11)

Yeah.

Gunnar Haid (1:10:12)

Hehehehehe

Belle Casement (1:10:14)

that shout out

to the oncologist. from a human health perspective, if that's what you're worried about and it's zinc and everything's way below everything, then you can just move forward. So sometimes it doesn't matter and then sometimes you have to spend, sometimes this sort of bad duplication and variability within the sample is the trigger for more work and more analysis.

Gunnar Haid (1:10:36)

Yeah. Yes,

see, you nailed it. No, we're not. We are not. But you nailed it. You nailed it. Exactly right. What needs to happen is someone needs to sit down and think about what does all this tell us? And that is not easily done in, not easily put into an Australian standard. That is definitely not easily done in a guideline. A good friend of mine always says,

Belle Casement (1:10:40)

We're not that far apart, Kuna,

Yeah.

Yeah.

Yeah.

Gunnar Haid (1:11:06)

standards and guidelines for bad engineers. I'm not going that far, but you know, it's so difficult to, even now on this podcast, we don't know exactly how do you interpret all this, except for that we all, of us agree, you know what? We need to, someone needs to really, really think. And one of the things of really, really thinking is not only looking at the data in the spreadsheet, but it is also looking at...

How far away are we from our thresholds in the first place? That would be an interesting one. What contaminant are we talking about? And there is so much that goes into this. Am I near a sewer ⁓ line? Am I near a water pipe? All these things, and they are so difficult to put in a standard. They are so difficult to train on juniors. But you are absolutely right.

Belle Casement (1:11:55)

Yeah. Yes.

Gunnar Haid (1:12:02)

We are way closer together than it sounds here. Really. Hammond, you meant to say something? mean, by the experts we're talking?

Belle Casement (1:12:05)

Yes.

James Hammod (1:12:11)

Yeah, I'm just trying to keep up guys, sorry. ⁓

No, I was just going to add, really my last thought on all this is that...

Belle Casement (1:12:17)

Thank

James Hammod (1:12:19)

Like clearly there's differences of opinion and this is one of those things where I think that whether you want to take that conservative approach or whether you're trying to sort of get into the main, into the middle is, I would describe it as a difference of opinion and it comes down to policy and the outcomes we're trying to get. And for me it's just like I would just love it if it was articulated somewhere so that we could all, even if we didn't all necessarily agree, at least we all know that we're all, this is what we're doing, right? As an industry, this is what we're doing. Because I think that some of the challenges that we come across is where

Belle Casement (1:12:42)

Yeah.

James Hammod (1:12:49)

⁓ These kinds of differences of opinions come out when you're working with say an auditor or somebody else But they come out when a project is well advanced and it's kind of like you think that I think this ⁓ And then it can cause all sorts of problems and and then we're you know We're in this world where there's commercial realities and our clients are trying to get certainty on stuff early for something like this ⁓ Geez, it be great if it would just be that somebody makes a call You know New South Wales EPA makes a call and says this is how the industry is gonna do it right and I guess

Belle Casement (1:13:00)

Yeah.

Gunnar Haid (1:13:02)

Hmm.

Hahaha

Belle Casement (1:13:15)

Yeah.

James Hammod (1:13:19)

what you're saying Gunnar that we don't always want everything articulated to the nth degree in guidelines because it hams everybody and it stops any kind of free thought or real analysis but we're in a world of commercial realities where people are trying to get certainty and I feel like something like this could be cleared up.

Belle Casement (1:13:37)

Yeah,

I agree. The NEPM makes us do some QAQC sample collection and analysis that produces some pretty low value data. I think that's inarguable.

Gunnar Haid (1:13:51)

Yeah.

And poor auditors and I need to say it here, poor auditors. I have so many similar discussions like I had with you guys here now with auditors and so often, not necessarily about data quality, but about whatever it is. And ⁓ these are all highly experienced people. And of course they agree with a lot of the stuff that we just...

just mentioned here, but these poor guys and girls have handcuffs on because someone thought it was a good idea to put the NEPM as a de-facto law, as a regulatory instrument.

Belle Casement (1:14:26)

Yeah.

well.

feel that sorry for the auditors. Plus it's the audit support staff doing all the work anyway, so.

Gunnar Haid (1:14:35)

No, it's okay. It's okay.

Hammond has to go. Bill, it's been, it's been, this has to be one of the most entertaining and most pleasant podcasts we have recorded. I loved it. I loved everything about it.

Belle Casement (1:14:46)

Hahaha

So good, I'm so pleased.

James Hammod (1:14:54)

Definitely.

Gunnar Haid (1:14:55)

Thanks so much for being with us and have a good one you say hi to Victoria please.

Belle Casement (1:15:00)

I will do.

James Hammod (1:15:01)

I'm sure we'll cross paths again soon.

Belle Casement (1:15:03)

Thank you so much, guys.