As a general matter one cannot use short-term leach tests to infer anything about either rates of reaction or effluent concentrations for full-sized waste-management systems or natural watersheds. There are at least three significant issues:

• The leaching behavior is a function of the water: rock ratio. [And of course of the lixiviant - I assume we are talking about a simple lixiviant like DI water or a well-characterized rainwater analog.] In a laboratory leach test, you set that mass: mass ratio (or perhaps you defer to a standardized formula). However, some simple calculations (start with what is the solution: solid ratio by mass in a porous medium - guess 25%-30% porosity - that is fully saturated?) will show you that some of the "standard" solution: solid ratios are astonishing overestimates of real-world conditions.
• The leaching behavior is apt to be related much more strongly to the surface area of the solid than it is to the mass (what do you expect qualitatively for leachability if the mass is divided into twice the number of particles, into 10 times?). We need to be able to scale the outcomes we achieve under controlled conditions, including of surface area, to the system we actually care about in the field.
• For some materials, the geochemcial reactivity of the solids is evolutionary, not constant. The oxidation of sulfides is perhaps the best example. The results one gets from the short-term testing of a sample developed from fresh core may not in the leachates represent the behavior of that same material after 1-5-10-... years of reaction.

Are short-term tests then useless? No. Firstly, you may have to do them as a practical, regulatory matter. In this case, I think we should call them "classification" tests, not "characterization" tests. But that's just me. More scientifically, you may find that with some materials (already weathered tailing or waste rock), these simple tests indicate that the pH and TDS are apt to be problematical and that detectible levels of some trace elements would likely be present. So they can help you identify potential contaminants of concern, and (again) to classify your materials for further evaluation or for preemptive engineering controls.

But if you want to get at kinetics, you need to do kinetic experiments. And if you want to estimate effluent chemistry of a real-world system, you need to think about how to design and conduct your lab tests in such a way that you can scale them to the actual problem in some defensible manner.

I agree with you on this matter. Mineralogical controls on leaching (e.g. encapsulation) are often not evident as a result of the sample preparation techniques or only show themselves once we start to see weathering of the minerals. Selenium where it is a replacement element within the sulphide mineralization comes to mind.

If this is a site where you have access to older drill core then you may be able to do a bit of a pseudo-kinetic evaluation while needing to keep in mind core storage etc. I tend to use the short term leach tests to flag potential contaminants of concern that may be seen in the effluent as much as to have an idea of the initial flushing of metals that may occur during the first flushing of the rocks after being excavated.

You only mention lab tests but I think as much as anything there is great value in field kinetic testing programs provided you can get materials that are representative of what is being mined. There is a huge difference between crushed core and blasted waste. That said, the fundamental need is to properly design any study based on a good understanding of the site in question.

I agree cautioning on the use of leach tests or other laboratory based kinetic tests work as a predictor of what will happen in the field. Also on the observation that site specific conditions play a significant role in AMD; tailings will behave differently to waste rock and with co-disposal, well the situation becomes more complicated!

In the nineties we conducted scaled up leach tests in the field using run-off mine samples of waste rock and tailings in large leach tests using 220L barrels and two dump truck sized leach pads (with under-drainage for sampling) all under field conditions, that is waiting patiently for rain events followed by drying events.

One of the main observations I made was that iron/manganese oxides from higher levels in the barrel coat rock and appeared to prevent acid generation or neutralization in lower layers. There was a higher tendency for oxides to coat carbonates. These are mechanisms that lab tests do not capture.

There may be nearby relict waste dumps that can provide invaluable clues. Use laboratory test work for classification and back of the envelop calculations for early predictions but set up field testing very early in the program and make lots of careful field observations to revise your acid leachate models. Finally remember that most mines initially mine the oxidized material before they get to the sulphides, so make sure your tests are on the materials of concern.

This question has been discussed extensively in North America for the past two decades, as it goes to the heart of predicting mine drainage chemistry. It was addressed in three separate reports by MEND (Mine Environment Neutral Drainage, a Canadian organization) on the prediction of mine drainage. The most recent report on this topic can be downloaded here:www.abandoned-mines.org/pdfs/MENDPredictionManual-Jan05.pdf. It represents our most current knowledge on the subject.

The issues, specifically with respect to ARD, have been discussed not only in North America, but all over the world. Scientific and engineering work is very advanced in Australasia, southern Africa, Europe, and Latin America also. We are indebted also to the concepts of landscape geochemistry and the now extensive academic investigations of mechanisms and rates of mineral dissolution and precipitation.

The MEND 2005 document is well worth reviewing, and we should take full advantage of the documented work there and in other compilations and reviews. But, in my opinion, we do not want to take such compilations as if they were fixed authority, to be applied one-on-one to a new problem. Instead, what we need to do is to adopt a proper exploratory attitude. What, *for my site*, am I trying to determine? What materials (proto-lithologies, alteration sequences, minerals and mineral textures) do I have to consider? What climate/hydrology/hydrogeology? What are the mining and mineral processing methods that will apply here? What will count as an answer to the questions I have been asked (or I need to ask myself)? If this requires testing - as it indeed will - what is my conceptualization of the physical-chemical system, and how can I design experiments that are related to that conceptualization. Then, how shall I scale my results to the problem that actual interests the world?

I am, of course, all in favor of field trials. I've done many, and over a long time. But as valuable as they are, they cannot be one's entire program, or we will never be able to provide timely advice to our Principals. And in any case, they still will be of a scale smaller than most actual facilities. And some facilities, like open cut mines, we have no way to test via a field trial, yet we will have to address that system, too.

So, I suggest, what we need is a program that makes active use of experimental design - where we make conscious choices: what we shall test; at what particle-size distribution; at what water: solid ratio; for how long; over what periods will sample the test system; how will we know when to stop? That's just to get started. What will our approach to QAQ/QC be? How is our test protocol to be related to the site climate and hydrology? How much data is enough?

I started in mine geochemistry in the summer of 1970. There are certainly similarities, of some kinds and some strengths of association, but I think it is entirely fair to say that no two mines I have worked at in 40 years are so close to identical that *exactly* the same approach could be applied to any two. Between the difference sin materials, climates, and issues that need attention, every single project needs our focused attention to produce the work necessary and sufficient (which is to say cost effective) for its specific case. Sure, we can (and ought) learn from the thinking of others (in other circumstances), but we should be very careful about taking a "standard" approach, or thinking that all the questions already have been answered.

Not sure if you have come across this Master’s thesis but it may have some useful information in it.http://eprints.jcu.edu.au/3220/2/02whole.pdf

I think you have all raised some good points. I always liken ARD/ML prediction to being a detective. You piece together information here and there from the range of imperfect tests that we carry out and try and come out with a reasoned, defensible prediction of what the longer term results will be during operations and on into closure. Field cells are great, but early on in a program you may be scratching around to get enough cores for static tests, let alone finding enough to fill a barrel or something larger.

The issue of QA/QC is something which in my experience tends to get overlooked, in part because clients on the whole still seem reluctant to fully invest in geochem programs that don't advance their bottom line. I would be interested to hear others experiences in getting clients to commit to QA/QC testing e.g. for humidity cells?

For the first time I have heard the comment about scaling and this is a problem which cannot be overcome.. Hence when I thought about setting up Boojum business plan, i decided experiments need to be carried out in the field first i.e. on the tailings and on the waste rock pile and then fitting science around it. Have a look at the Kinetics of water -rock interaction..SL. Brantley, J.D. kubicki and A. F. White, springer 2008..It explains a lot, why predictions are impossible - given the scale at which we can make observation.

Glad to hear these comments

http://www.springer.com/earth+sciences+and+geography/geochemistry/book/978-0-387-73562-7

I do not believe that scaling is impossible, only difficult, or that meaningful predictions are not possible.

But we need to distinguish what we mean by prediction. It seems to me that there are two senses, a strong one and a weak one. A strong prediction might be something like predicting the trajectory of an artillery shell, or the next return of Halley’s Comet. This sort of prediction is strong in the sense that we expect a high degree of accuracy and precision in a specific quantitative analysis. This possible for some systems, especially if they are quite simple and readily characterized with accuracy and precision.

In contrast, one may envisage a "weak prediction" as one that is suitable for decision-making under uncertainty when sound decisions can be made while tolerating some range of accuracy and/or precision. For example, imagine I were able to estimate a future water quality outcome within classes of outcomes, such as:

• pH > 6; 4.5 – 6; 3 – 3.5; < 3
• SO4 5000 (all as mg/L)
• Metals/metalloids present at 1 mg/L

And so on. I offer the ranges only as guidance for discussion; specific classes of outcomes should be based on site-specific conditions and concerns.

It seems to me that if we could predict water-quality outcomes in such ways we would have almost all the information that anyone would need for decision-making. If that were so, we would understand where our projected water quality lay with respect to water-quality criteria, and we would be able to make sound recommendations for water-treatment alternatives. We would know how to advise our engineering colleagues on designs for waste-management. It would inform our evaluation of the need to define the potential outcomes further (i.e. the need for additional testing and its cost-effectiveness). If I thought the effluent was likely to have a pH between 3 and 4.5, would my understanding of the system be different if the “true” pH was 3.4 or 3.8? Almost surely not. We would need to be accurate as between classes, but we would not need to have great precision within classes.

It seems to me very likely that with thought and due effort we can find our way to predictions of this second, “weaker” kind. What *exactly* shall be the chemical conditions of let’s say 100 (or 1000) M tonne of waste rock? I’m not sure what that even means. But with good materials characterization and kinetic testing that can be applied forward through a well-founded scaling approach, I believe we can make useful predictions that are very relevant to Project decision-making, with respect to the concern of both the miners and the protection of human health and the environment.

I am not a geochemist, but have worked with good ones in several projects. Predictions about expected water chemistry during operations or at closure were always qualified. They would bracket scenarios as "base case" and "conservative best judgment", etc. Metal release rates were obtained from "mean case" and "worse case" kinetic leach tests and used to determine metal leach rate for different rock units. Adjustments would be made for degree of fracturing, secondary mineralization, etc, eventually resulting in predictions of water chemistry sometime in the future.

My point is that predicted water chemistry was a required deliverable, and geochemists made those predictions with the tools available at hand, qualifying their results as best they could. It may be part Science, part wizardry, but we've had some pretty good wizards in this field.

You raise some interesting questions with respect to QA/QC. Let's try to take the technical bit first.

If I have a rock or mineral sample and I wish to investigate it via say XRD or whole-rock chemistry, I can structure a sampling approach that permits replicate testing. It may or may not be easy, but I can imagine at the extreme crushing the entire sample, then using a standard approach, like riffle-splitting (or some other approaches) to producing truly replicate sub-samples. Then I can run my replicate samples and analyze the outcomes using standard (and maybe also non-standard) statistical techniques. This would allow me to rigorously assess the central tendency and the dispersion of the outcomes, and to offer a well-qualified answer as to the composition. If I were diligent, I also would include some standard reference materials to help me understand the accuracy, and I would also include in my experimental design duplicate testing across analysts and perhaps across analytical laboratories to understand other possible sources of divergence in outcomes.

But for the sorts of kinetic testing we most often would imagine in a mining environmental study, this kind of attention to QA/QC is not possible. Firstly, the sizes of the particles that we wish to analyze are such that formulating replicate sample sis not practicable even with the smaller end of our experimental range. Perhaps more importantly, for most rock masses that we may be testing, we are typically not at all confident that we have collected random samples in the sense that is typically imagined for statistical analyses. Imagine: we wish to address the environmental fate over time of a rock pile (or pit walls) containing millions to hundreds of millions (or for some mines billions) of metric tonnes. We have available to us, if it is a prospective study, some number of cores, the total mass of which amounts to a very small fraction of 1 % of the total mass of the rock in question. And from these cores, we will select a few kg of sample at some intervals (that we need to justify). I suggest that we cannot imagine that we are engaged in the sort of replicated experimental designs for which the standard statistics of sample analysis and inference to population characteristics is well suited. So we have a QA/QC problem.

Does this mean that it is OK to just grab samples and do more or less whatever we will with them, trusting to wizardry or whatever for the outcomes? Not at all, of course. We need to exert as much qualitative control on the experimental design and the experimental process as we can. And we should consider as much QA/QC testing as the system allows. Then what?

We need to be suitable humble in what we believe we can say, and use predictions in what I have suggested as the “weak sense.” If we hold ourselves to a “strong sense” view of what we are doing, I think the problems outlined above are clear, and we are probably fooling ourselves about the quantitative rigor of our work product. The “weak sense” predictions may, if done carefully, be entirely sufficient to the requirements, and we should be happy to put them forward, provided we are clear about what we have done.

The issue of support from clients is a somewhat different one. I don’t think their problem typically is that they are a bunch of cheapskates or philistines. It is that they need to be convinced that the additional costs that are proposed are necessary and sufficient to the task. That is the charitable interpretation of what being bottom-line conscious really means. So it becomes our duty – and I do mean duty: this is a matter of fiduciary responsibility – to present the work plan that is necessary and sufficient.

Your thoughts?

If I can throw in my two cents - I think it is very important to consider the geo-environmental characteristics of the system that one is trying represent in a geochemical simulation of water quality no matter what scaling factors are used; which is a concept I believe you know well from your published work. Observed data from known systems provide a very good guide of the ranges of chemistry that should be expected from a simulation model for particular type of geology-(bio) geochemistry-mineralogical-climate setting. There can be a wide range of possible concentrations, such as for metals and metalloids, for any given setting. But, if a model produces results in conflict with what might be expected for a geo-environmental setting, then, perhaps, the assumptions inherent in the model calculations should be very carefully considered and possibly revised. With respect to scaling, I think that geochemical models should incorporate the ranges of uncertainty in their scaling calculations and produce ranges of uncertainty in outputs; something may not always be feasible and sometimes difficult to explain, but more realistic when considering the variability of natural systems. Thanks for the discussion. Scaling up from laboratory to field is always a thought-provoking process no matter how many times I do it!

It is interesting that in all discussions the climate factors and the microbial contribution to the weathering processes have not been mentioned.. Look at the book I suggested. It will be interesting.

Thank you for your detailed reply.

Doing QA/QC on whole rock analyses is fine in itself, but in my mind, ultimately, it's the humidity cell tests which are the focal point of most programs and even if I can afford to do adequate QA/QC testing on whole rock analyses at $25 a go, it still leaves a little bit of a leap in faith to assume that the humidity cell results are representative, unless you have replicate humidity cells too, which is rarely done, as we have discussed.

I fully agree that bracketing water quality predictions and applying suitable caveats to the results is the way to go. I think retaining a healthy degree of skepticism to modeled water quality predictions is appropriate. At times the grayness of the voodoo that we do does get to me 🙂

We agree completely. I see no way that even column tests can be replicatedas is usually the case in process-control studies. So we need to think veryhard about the confidence with which we express our results. That isexactly why I think we should be thinking and planning around prediction in

the "weak sense".

You made such an effort in replying - but QAQC has little to do with scaling and the processes are different at different scales and the interact between the scales.