Mineral Deposit Tasting and Smelling
Avoiding the GIGO and arm-waving problems in the applied science of mineral exploration -- and not stumbling during the process
Proceeding from Basic Principles
That last post (the one subtitled “When it is really just a simple matter of ‘garbage in, garbage out’”) was derived from the following much longer piece I wrote in late 2019 about practical application of the scientific systems perspective to mineral exploration. To produce that most recent Substack post, I basically whittled down the older paper to the very barest and rawest expression of its basic principles, and then used those principles to briefly show how a couple of important modern scientific controversies have recently been (mis)handled.
The original paper replicated below should be of interest to mineral exploration geologists for obvious reasons, while to the non-geologist it might be of interest because it indicates in a bit more detail how and why the same GIGO problem surfaces again and again in less iconic but still important human matters.
Introduction
A fundamental truth that must be grasped by applied researchers is that a mineral system is not an exploration targeting system. In order to apply the mineral system to make decisions on ground selection at various scales, a practical targeting model must first be created. While the acceptance of the mineral systems approach is increasing, its application to exploration targeting often stumbles at this step.
–- McCuaig, Beresford, and Hronsky, 2010, p. 129
You can’t always get what you want — but if you try sometimes, you find you can get what you need.
— Rolling Stones, 1969, Let It Bleed
In a 2009 address, McCuaig et al., provided a practitioners’ critique of the mineral exploration industry. Their presentation reiterated the widely acknowledged major problems troubling the industry. Growing demand for mineral and energy resources by society is occurring at a time when resource depletion is outstripping replenishment because of the apparent exhaustion of the search space. This situation is consequently accompanied by declining exploration success, rising average exploration costs per discovery, declining average ore deposit size, and declining average ore quality. Added to these seemingly insuperable technical difficulties are social change requiring a reduction of the physical, social, and environmental footprint associated with mineral exploration and mining – and a somewhat understandable lack of financial support for mineral exploration work itself.
McCuaig et al.’s prescriptive policy recommendations for dealing with this set of problems are to return to an emphasis on greenfields mineral exploration, and to adopt a potentially more accurate, flexible, and productive mineral systems perspective to guide mineral exploration operations. The data show that the authors are correct on the first account (Figure 1), but the jury is still out on the practical utility of the various mineral systems and targeting models elsewhere created by promoters of the mineral systems perspective.
This paper first reviews the development of the mineral systems perspective in mineral exploration, compares examples of mineral system hypotheses to concrete petroleum system theory, and then explains and discusses a generalizable exploration geochemical workaround that avoids much of the native difficulty of practically applying systems theory to mineral deposit exploration.
What is the Mineral System Perspective?
If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.
--The Duck Test: the essence of the traditional deposit model approach to mineral exploration
…mineral systems are not as well understood as petroleum systems, and greater uncertainties must be attached to their analysis.
-- Hagemann, Lisitsin, and Huston, 2016, Mineral system analysis: quo vadis
Single -- but not necessarily valid or immediately helpful -- ideas episodically surge into and out of group thought, causing individual minds to school together for a time like small fish swimming in an ocean. For example, the initial geological ‘consensus’ was that Wegner was talking pure nonsense, but now continental drift theory colors the professional thinking of all geologists. Kahneman calls this usually temporary external control of individual minds by salient ideas, “priming”, and Richard Dawkins has similarly observed that such ideas (“memes” in Dawkins-speak) frequently act almost as self-perpetuating parasites of human minds.
Accordingly, just a few years after the petroleum systems approach to oil and gas exploration was publicly summarized in 1991 by “The petroleum system – from source to trap”, Wyborn, Heinrich, and Jaques (1994) proposed adoption of the same process-reconstructive perspective for mineral exploration targeting. Wyborn et al. generally define a “mineral system”, and describe its primary potential advantage for the purposes of mineral exploration targeting thusly:
A mineral system can therefore be defined as ‘all geological factors that control the generation and preservation of mineral deposits, and stress the processes that are involved in mobilizing ore components from a source, transporting and accumulating them in more concentrated form and then preserving them throughout the subsequent geological history.’ The mineral system concept emphasizes that for many ore deposit types, although economically viable mineralization may occur only at a scale of, say, hundreds of meters, the total system of fluid-rock interactions that led to ore formation can extend over a distance of tens to hundreds of kilometers around the deposit. When mapped out, the total mineral system provides a far larger exploration target than the actual target itself. Important geological factors defining the characteristics of any mineralizing system include:
1. Sources of the mineralizing fluids and ligands;
2. Sources of the metals and other ore components;
3. Migration pathway;
4. Thermal gradient;
5. Energy source;
6. A mechanical and structural focusing mechanism at the trap site; and,
7. Chemical and/or physical traps for ore precipitation.
With their minds seized by the invasive petroleum system paradigm, and evidently hoping that a mineral system approach could be eventually made to generate the same level of relatively consistent exploration success as achieved in oil and gas exploration, an increasing number of applied mineral exploration researchers now labor to produce mineral system exploration targeting models that might help the mineral exploration industry break away from its habitual behaviors that have led to chronic wealth destruction (Figure 2).
Why Hasn’t the Mineral System Approach to Mineral Exploration Lived up to Expectations (So Far)?
The use of insubstantial, unsupported, or sensational arguments or claims, especially in scientific discourse; an instance of this.
--The second definition of “arm-waving”, https://www.lexico.com/en/definition/arm-waving
Given the very large differences between the technical problems faced by geologists in the mineral exploration and hydrocarbon exploration industries, efforts of mineral exploration geologists to imitate the use of the petroleum system by oil and gas exploration geologists can arguably be likened to trying to fill a round hole with a square peg.
The greater expense of hydrocarbon exploration and greater average profitability of oil and gas production aside, the ‘playing field’ for oil and gas exploration geologists is appreciably more level than it is for most mineral exploration geologists searching for hydrothermal mineral deposits. Because sedimentary basins dominate oil and gas exploration and production, relatively simple layer-cake hydrocarbon host and source rock geology are largely preserved and accessible to hydrocarbon exploration geologists. It is more feasible to decode and map such critical system variables as fluid source, migration pathway, thermal gradient, and energy source when dealing with intact sedimentary basins than trying to do the same with the usually much more lithologically and structurally heterogeneous mineral deposit country rock facing mineral exploration geologists. 2D and 3D seismology, in particular (Figure 3), provides hydrocarbon exploration geologists reasonably accurate and precise means for identifying and locating likely oil and gas traps before drilling. Hydrocarbon reservoirs are also typically much larger exploration drilling targets than are mineral deposits – a fact that is particularly helpful to those first cousins of greenfields mineral exploration geologists, wildcat drillers. Some hydrocarbon reservoirs can be hundreds of kilometers across. Given these advantages, with proper preparation it is feasible for hydrocarbon exploration geologists to complete exploration drill holes with more than a random chance of discovery and eventual economic resource delineation success.
With the exceptions of some sedimentary rock-hosted hydrothermal ore deposits, however, relatively small, discontinuous, and often much older hydrothermal mineral deposits usually exist within much more structurally broken, lithologically-mosaic and confusing low-validity terrains. The physically complicated nature of these mostly observation-inaccessible host rocks compounds the difficulty of validating and then mapping critical elements of hypothesized mineral systems. For several example illustrations of consequently very complex and often entirely speculative mineral systems, see Figure 4.
Contrast these examples with the plain and simple plan and vertical cross-section diagrams of the basic petroleum system shown in Figures 5 and 6. Unlike in the cases of the Figure 4 proposed mineral systems, with enough effort, time, and budget, each element of Figure 5 and 6 is observable and measurable (i.e., is “‘mappable”’) by hydrocarbon exploration geologists, is objectively self-consistent and causally related to all other components of the system, and is identifiable in nearly all other petroleum systems on the planet. Self-consistency and causal interconnection of many of the various elements of the proposed mineral systems shown on Figure 4, on the other hand, are uncertain and are generally not at all readily observable or measurable by mineral exploration geologists. In many cases on Figure 4, the relevance of some of the hypothesized process elements to particular types of mineralization is completely uncertain.
Kahneman (2011) uses a term that points to the general source of the greater level of difficulty of the exploration problems faced by mineral exploration geologists, including those geologists seeking to generate valid targeting models from the mineral systems perspective. Kahneman would classify the professional operating environment of mineral exploration geologists as an especially “low-validity environment”.1 Low-validity environments are those domains of professional problem-solving that entail “…a significant degree of uncertainty and unpredictability” (ibid., p. 223). It is, according to Kahneman (and others), extremely difficult to impossible for humans, highly experienced and educated or not, to make accurate predictions in low-validity environments by just referring to their experience or intuition; i.e., by ‘eyeballing’ or by ‘flying by the seat of their pants’.
Kahneman (ibid., p. 224-225) maintains that there are two empirical reasons that members of professions working in low-validity environments find it difficult or impossible to make accurate predictions and come up with good solutions:
They strive to be clever and think outside the box, and therefore consider complex combinations of features in making their predictions (e.g., Figure 4); and,
They (and humans in general) are incorrigibly inconsistent in making summary judgments (and measurements) of complex information, and unreliable and inconsistent judgments (and measurements) cannot be valid predictors of anything.
The thread tying these two observations together can be found by considering the concepts of “necessary” and “sufficient” as they concern cause-analysis, decision-making, and problem-solving in science and systems theory. Boskey (2019) defines and distinguishes these two important concepts in this manner:
If someone says that A causes B:
1. If A is necessary for B (necessary cause) that means you will never have B if you don't have A. In other words, if one thing is a necessary cause of another, then that means that the outcome can never happen without the cause. However, sometimes the cause occurs without the outcome.
2. If A is sufficient for B (sufficient cause), that means that if you have A, you will ALWAYS have B. In other words, if something is a sufficient cause, then every time it happens the outcome will follow. The outcome always follows the cause. However, the outcome may occur without the cause.
3. If A is neither necessary nor sufficient for B then sometimes when A happens B will happen. B can also happen without A. The cause sometimes leads to the outcome, and sometimes the outcome can happen without the cause.
4. If A is both sufficient and necessary for B, B will never happen without A. Furthermore, B will ALWAYS happen after A. The cause always leads to the outcome, and the outcome never happens without the cause.
Boskey’s alternative #3 is where disorder, error, and inaccuracy pours into the human decision-making and problem-solving process. If factors or features of a system are not necessary and/or sufficient to make cause and effect-related predictions about that system, then subsequent predictions dependent on those features or factors will -- of course -- not be accurate. Too, if the measurements or observations of necessary and/or sufficient factors or features determining the outcome of a system are themselves unreliable and inconsistent, then predictions based on these measurements or observations will also be unreliable and inconsistent.
Because of the extreme difficulty of identifying, validating, and mapping the hypothesized critical elements — i.e., necessary and/or sufficient features of mineral systems — because of the low-validity environment problem, it should not be surprising that the mineral system paradigm has yet to offer material solutions to the current difficulties of the mineral exploration industry. A remaining question to consider, however, is whether exploration targeting approaches derived from the mineral systems perspective would be significantly more effective if ALL of the speculative and/or unverifiable processes and attributes of mineral systems were modestly (and rigorously) discarded and disregarded when seeking to construct practical exploration targeting models.2 If only those mappable (measurable) mineral system attributes that are clearly necessary and/or sufficient to the formation of hydrothermal ore deposits are considered, perhaps then some effective, practical mineral systems exploration targeting models can be created that would, in fact, rival the effectiveness of the petroleum system and the related targeting models utilized in hydrocarbon exploration.
Is There a General Workaround to the Problems of Making Accurate Predictions in Low-Validity Environments like those of Mineral Exploration Geology?
The provision of a winning solution was positively related to increasing distance between the solver’s field of technical expertise and the focal field of the problem.
-- Jeppesen and Lakhani, 2009, Marginality and problem solving effectiveness in broadcast search
Kahneman (ibid., Chapter 21) discusses several examples of a numerical bookkeeping method that minimizes the fallibility of human decision-making in low-validity environments by employing simple prediction models (e.g., targeting models) that only use necessary and/or sufficient mappable/measurable system attributes. Such strict modeling discipline consequently, according to Kahneman’s review of the matter, increases the reliability, accuracy, and consistency of judgments made with these models.
Kahneman discusses an example of such a predictive procedure from Dawes (1979, p. 572) which provides a two-variable mathematical model that has proven to be a very reliable means of accurately predicting human marital happiness and consequent marriage stability:
Marital happiness index = the frequency of lovemaking minus the frequency of fighting
Certainly, many, many more separate processes and complex interactions — mappable and unmappable — go on within a marriage besides lovemaking and argument, but a very simple combination of these two necessary and critical elements within the ‘marriage system’, as retrospective calibrations and tests reported by Dawes and others show, are sufficient to accurately gauge the overall condition of any one couple’s marriage system. Using this very simple numerical accounting approach, in any given marriage case, there is no need to painstakingly and laboriously try to consider and take into account all of the many mechanisms and processes occurring within the ‘marriage system’ in order to accurately and objectively gauge its current state and predict its likely future.
The wine-loving economist, Orley Ashenfelter, uses a qualitatively similar, but more mathematically rigorous approach to guiding purchases of red Bordeaux wines (1990 and 2008). Because yearly vintages of some red Bordeaux wines improve markedly in taste over time, there is considerable initial uncertainty when any of these wines are young and astringent whether or not they will develop poor, mediocre, or excellent taste after being stored for a decade or two. Linear regression of the historical records of only four variables, however – age of wine, average growing season summer temperature, amount of rain during harvest time, and total rainfall in the winter preceding that harvest – against records of subsequent selling prices, permits Ashenfelter to accurately predict magnitude3 of future prices (and associated palatability) of ensuing red Bordeaux vintages years and decades before their market sale. About 80% (adjusted-r squared) of the variance in log10 future wine price is predictable using Ashenfelter’s linear regression of price records against just the simple determinants or critical elements of age of wine, summer air temperature, and seasonal precipitation data. See Figure 7.
In contrast to this linear regression use of necessary system elements located at the beginning of the winemaking system like growing season temperature, Ashenfelter and Jones (2013, Table 1) showed that later obtained pooled wine expert taste ratings,4 plus the age of the wine at time of future sale, can also be employed to statistically predict palatability and consequent log10 price of maturing bottles of any given vintage. Accuracy achieved with this alternative limited set of necessary elements of the red Bordeaux wine system is very close to that achieved using objective early-in-the-system’s-history temperature and precipitation data (77% of variance in log10 price predicted: Figure 1 of ibid. ).
The fact that the outcome of the winemaking system can be predicted from either early primary causes (air temperature and precipitation), or later developing features (taste and smell) of the system and its product, shows that the two different, simple prediction models both employ mappable/measurable critical elements of the same winemaking system that are necessary and/or sufficient to that system. Neither model uses factors that are unmeasurable (unmappable) or causally-disconnected from the winemaking system. Ashenfelter and Jones (ibid.) in fact recommend that, in cases where primary air temperature and precipitation data are not mappable/measurable because of lack of weather records for a particular wine-producing region or estate, the secondary wine expert tasting data should instead be used to predict future wine prices (and wine palatability) from there.
The same sort of circumstance, as you may have already recognized, is analogous to the usual case in mineral exploration where the primary critical elements of a mineral system mentioned above from Wyborn et al. (ibid.) are largely unknown, uncertain, or are not materially mappable, while – at the same time -- some necessary system elements (those long familiar traditional model ore body ‘duck characteristics’) like local trace element geochemistry, rock alteration, or geophysical signatures characteristic of that same system are instead actually quite observable and very mappable.
Two Linear Regression Workaround Examples Relevant to the Target Modelling Problems Plaguing the Mineral Systems Perspective: “Mineral Deposit-Tasting and Smelling” for Exploration Geologists
While you guys are busy focusing on the rock, petroleum exploration geochemistry focuses on the stuff moving through, and accumulating within, the interconnected cracks and voids in those rocks.
-- author’s recall of an explanation provided to a petroleum exploration geologist by David Richers concerning the difference between petroleum exploration geochemistry and petroleum exploration geology, 1980.
So, the goal is to find something in the way of the mineral exploration equivalent of Dawe’s very simple Marital Happiness Index where “Happiness” = the frequency of lovemaking minus the frequency of fighting. Here are two examples of just such a thing from the world of mineral exploration:
1. An example from a mineral exploration environment with a relatively mild case of low-validity: northern Arizona high grade sedimentary rock-hosted breccia pipe uranium deposits
Northern Arizona’s uranium-mineralized collapse breccia pipe province, currently closed to uranium exploration for political reasons, contains most of the uranium resource of the United States. Located in flat-lying sedimentary rocks of the southwestern Colorado Plateau, the geology of this exploration and mining province is quite regular and predictable, and is very similar to the layer cake geology of the typical oil and gas exploration region. The high grade uranium deposits themselves, most of which are deeply concealed by cover of up to 1350 feet thick, share many attributes of Mississippi Valley type lead-zinc deposits, as well as most of the characteristics of sediment-hosted roll front uranium deposits. Turner and Turner (2016) explain and describe the relatively simple mineral system of the deposit type, and provide an exploration geochemical target model capable of predicting almost 98% of the variance in log10 uranium resource among individual prospects within the mining and exploration province from bulk rock chip surface sample chemical analyses alone.
Three main critical elements acting early in this mineral system have been determined to govern the extent of uranium and metal sulfide mineralization in collapse breccia pipes during the breccia pipe mineralization process. These synchronous factors are: (1) existence of bacterial feedstock (oil) within pipe breccia; (2) upwelling, metal-rich brines passing through the pipe; and (3) consequent and simultaneous generation of two proximal geochemical reduction barriers capable of precipitating uraninite and metal sulfides from upwelling mineralizing fluids. See Figures 8, 9, and 10.
Unfortunately, unlike in the case of growing conditions discussed winemaking system, it is not readily possible to map or measure records of the early variations in these three main critical elements for each of the thousands of northern Arizona collapse breccia pipes in order to predict which pipes are most apt to contain economic quantities of uranium. At the local scale of a mineralizing collapse breccia pipe, however, the three named critical factors came together to generate two geochemical barriers, a reduction barrier and a sulfide barrier within each breccia pipe. In turn, these two primary system geochemical barriers not only precipitated the high grade uranium mineralization and capping massive sulfide mineralization characteristic of mineralized northern Arizona breccia pipes, but also created and attenuated the vertical and horizontal metal zoning of the much weaker metal leakage surrounding the ore grade material existing at the core of the local system. Turner and Turner (ibid.) show that this zoning of primary system metal leakage, as represented by variations in the four geochemical parameters of Figure 10, functions just like the taste and smell of wine in the winemaking system when carrying out linear regressions against breccia pipe uranium endowment. The estimated equation, a means of measuring the “taste and smell” of any given uranium-mineralized mineralized breccia pipe allows accurate pre-drilling estimation of the subsurface uranium resource of any given northern Arizona breccia pipe.
The necessary “taste and smell” measurements of the metal leakage of any given breccia pipe can be obtained by quantitative chemical analysis of simple bulk chip surface samples. See Figure 11 and ibid. for more detail.
2. An example from a mineral exploration environment with a relatively strong case of low-validity: the Carlin-type gold deposits of northern Nevada
The Basin-and-Range geologic environment in northern Nevada is more representative of the difficult technical working conditions usually facing mineral exploration geologists. Rather than evidencing the regular, mostly structurally intact, layer-cake geology of the neighboring Colorado Plateau containing uranium-mineralized collapse breccia pipes, the northern Nevada region containing Carlin-type gold deposits is an exploration region of considerable structural complication and lithological variation.
One of the fortunate fruits of the U.S. National Uranium Resource Evaluation (“NURE”) program of 1973-1984, however, is a large reconnaissance-scale collection of drainage sediment samples obtained from most of the northern half of Nevada known to host Carlin-type gold mineralization. Fortuitously, nearly all of these samples were collected BEFORE most of the surface-disturbing exploration and mining of Carlin-type gold deposits took place in Nevada.
The original public domain geochemical analyses of the Nevada NURE drainage samples were analytically very inconsistent and often incomplete being provided by several different, independently functioning federal laboratories that lost funding before analyses of the NURE sample backlog could be completed. Nonetheless, starting in the 1990s, re-analyses of the original NURE samples, and analyses of additional soil and sediment samples taken later by the USGS and BLM, were carried out using much better standardized, more modern analytical methods (USGS 2004, pp. 65-69).
Employing the same linear regression geochemical target modeling geochemical data approach used in northern Arizona for predicting the uranium resource endowment of uranium-mineralized breccia pipes,5 it was determined that the modern chemical analyses of the Nevada reconnaissance drainage sediment samples are extremely predictive of resource and location of known Carlin-type gold deposits. See Figure 12. The tight Figure 12 goodness-of-fit corresponds to prediction of 89% of the variance in log10 gold resource among Nevada’s known Carlin-type gold deposits on the regression basis of metal leakage “mineral deposit taste and smell” of only seven different geochemical parameters.
Aside from providing a direct, relatively accurate means for ranking the Carlin-type gold mineralization potential of most current mining claims and exploration projects in northern Nevada, the mathematical relationship constituting the “mineral deposit-taste and smell” targeting model for Nevada Carlin deposits also located and ranked almost 150 separate additional, unstaked greenfields target areas having apparent gold endowments on the order of 10^6.5 ounces or more within the northern half of Nevada sampled by the NURE reconnaissance stream sediment sample set.
Very interestingly, this modeling and data analysis work also quite accurately detected and predicted the gold resources of both Nevada’s Marigold Mine (known historical gold resource = 10^6.7 ounces) and the proximal Lone Tree/Stonehouse deposit (known historical gold resource = 10^7.1 ounces). This is significant as neither of these “orphan” sedimentary rock-hosted gold deposits was used in the Carlin-type gold deposit log10 gold resource linear regression target modeling work because neither deposit is considered to be, strictly speaking, a Carlin-type gold deposit. Nevertheless, equation-predicted gold resource values from chemical analyses of the various drainage sediment samples sourced by the 10^6.7 ounce Marigold Mine area very closely matched the known gold resource of the mine, varying from a predicted 10^6.4 to a predicted 10^6.9 ounces. Predicted gold resource values from chemical analyses of separate drainage sediment samples sourced by the 10^7.1 ounce Lone Tree/Stonehouse deposit area also closely matched the known resource of this orphan gold deposit, varying from 10^6.4 to 10^6.6 ounces.
This unexpected result of the “mineral deposit-tasting and smelling” method for exploring for Carlin-type gold deposits strongly suggests that the “mineral deposit-taste” targeting model method is intrinsically flexible enough to find and accurately assess the mineral resource potential of ore deposits that exhibit a “style” moderately departing from the standard attributes of a given deposit model. As McCuaig et al., 2010, anticipate (p. 129), this sort of targeting model behavior would be especially useful in greenfields exploration programs where changes in rock type and other critical elements of a mineral system would be expected to create some otherwise unanticipated changes in mineralization style.
Conclusion
The mineral exploration approach described in the two examples above provides “…a conceptual framework to help translate the mineral system understanding to practical application” (ibid.). Indeed, the technical development of the explained exploration geochemical approach discussed here independently evolved along the exact path generically proposed by McCuaig, Beresford, and Hronsky (ibid.):
A four-step process is proposed for linking the conceptual mineral system with data available to support exploration targeting. These steps include translation from (1) critical processes of the mineral system, to (2) constituent processes of the mineral system, to (3) targeting elements reflected in geology, and (4) targeting criteria used to detect the targeting elements directly or by proxy.
In contrast to the rather complex and cumbersome exploration guiding procedure envisioned and promoted by McCuaig et al., however, “critical processes”, “constituent processes”, “targeting elements”, and “targeting criteria” in the mineral systems target modeling approach discussed here are all resolved by geochemical sampling, analysis, and linear regression into the single resultant “mineral deposit-taste and smell” (i.e., the estimated metal resource) from the primary system metal leakages surrounding sought ore deposits.
Finally, by deploying this exploration geochemical method with due technical respect to the long-recognized differing behaviors of trace metals in the primary and secondary geochemical environments, it is not difficult at all to use any defined regression equation-based geochemical targeting model to guide and manage all scales of exploration work.
This is not the same as stating that the hydrocarbon exploration environment is not a low-validity environment. Rather, it is to say that mineral exploration geologists contend with substantially more exploration uncertainty and unpredictability than hydrocarbon exploration geologists. For a recent summary review of the technical woes of hydrocarbon exploration geologists that they themselves, see https://www.sciencedirect.com/science/article/abs/pii/S0012825215300301.
Indeed, this is a major strength of the older, less flexible, but currently yet practical, traditional ore deposit models.
Ashenfelter’s regression equation predicts log10 future price of red Bordeaux vintages, hence the reference to “magnitude” here.
Based on such mouth- and nose-perceptible qualities as “balance”, “length”, “depth”, “complexity”, “finish”, and “typicity” (https://www.dummies.com/food-drink/drinks/wine/how-to-discern-wine-quality/).
Plus the utilization of some additional statistical, noise-reducing Bayesian legerdemain to compensate for the changing influence of the variations in Nevada’s mosaic country rock lithology on regression coefficients.
A reader wrote me privately and humorously this morning to ask, "But what do you do when you don't have any sense of taste?" My answer to him was that you then need to find a new analytical chemistry laboratory to use.