We have just downloaded hundreds of pages of papers, theses, and blog posting on Uncertainty Analysis in groundwater analyses for mining projects.
At last I have found another word that is as slippery as “sustainable” in its application to mining. Basically the word “uncertainty” like the word “sustainable” can mean whatever you want it to mean. Like sleeping with Alice in Wonderland.
The problem arises in both cases when a simple old colloquial term is applied to a sophisticated scientific concept. In this case, as applied to the word “uncertain,” uncertainty is the absence of confidence or sacredness that a groundwater modelling condition or parameter has a definite, ascertainable, or fixed value. Uncertainty analysis is also an advancing field of statistical evaluation.
You can almost see how it happens in this story of uncertain certitude.
The old professor asks his student: “Are you certain that is correct?”
“Not really,” the student replies. “I am uncertain about how it happened and how it will turn out.”
“We cannot very well say we don’t know what we are doing,” says the professor. “Why not say we will do an uncertainty analysis?”
“What is that?” asks the student.
“I don’t know,” says the professor. “I am uncertain. But that does not mean we cannot get research money to formulate a new branch of statistics called Uncertainty Analysis.”
And the next thing you know commercial software vendors are selling Uncertainty Analysis codes and regulators are demanding UAs as part of the EAs. And more and more research is done using the slippery term to greater uncertain understanding by the taxpayers and other affected types. For example, here is the abstract of the most understandable paper just downloaded:
Digital spatial data always imply some kind of uncertainty. The source of this uncertainty can be found in their compilation as well as the conceptual design that causes a more or less exact abstraction of the real world, depending on the scale under consideration. Within the framework of hydrological modelling, in which numerous data sets from diverse sources of uneven quality are combined, the various uncertainties are accumulated.
In this study, the GROWA model is taken as an example to examine the effects of different types of uncertainties on the calculated groundwater recharge. Distributed input errors are determined for the parameters’ slope and aspect using a Monte Carlo approach. Landcover classification uncertainties are analysed by using the conditional probabilities of a remote sensing classification procedure. The uncertainties of data ensembles at different scales and study areas are discussed.
The present uncertainty analysis showed that the Gaussian error propagation method is a useful technique for analyzing the influence of input data on the simulated groundwater recharge. The uncertainties involved in the land use classification procedure and the digital elevation model can be significant in some parts of the study area. However, for the specific model used in this study it was shown that the precipitation uncertainties have the greatest impact on the total groundwater recharge error.
In brief, the more it rains, the more water infiltrates. We do not know how much, but have had fun with the stats thereof. Thus informed, I was able to write the following—edited for blogness.
Two types of uncertainty that may occur in mine modelling result from:
- Subjective uncertainty due to imperfect knowledge or ignorance of a complex system (simplification of the system) or process uncertainty.
- Stochastic uncertainty is due to randomness (small scale heterogeneity, future changes, etc.) of parameters.
Subjective uncertainty may arise because of limited or poor judgment. In addition, the modeler may simply be wrong either as a result of inexperience, inattention, pressure, or stupidity. This may, for example, result in a model that is incorrect and/or unrepresentative of actual conditions. I recommend documented peer review as a way to limit the possibility of there being significant uncertainty arising from subjective factors.
Examples of uncertainty arising from errors include:
- The model parameters may be inaccurate, incorrect, or inappropriate.
- Measured data may be incorrect as a result of installation and measuring errors.
I recommend that error-based uncertainty be limited by implementation of Quality Assurance and Quality Control programs and procedures.
At the end of the day, I recommend a Monte Carlo analysis. At least I can understand it, and it does quantify the range of things that may happen.
The good thing about the word “sustainable” is that it sounds good in any context. You can throw it around with abandon and still be admired and forgiven. Not so with the word “uncertainty.” It is negative and down. Can you imagine the public reception of an environmental analysis that includes a groundwater impact evaluation that includes an uncertainty analysis?
Perforce the writers of the groundwater uncertainty analysis will have to admit to uncertainty; they will have to quantify their uncertainty; they will have to explain away the fact that, in reality, they have no idea of what they are talking about.
I do not care if the EA writes of uncertainty as stemming from error, stupidity, bad modelling, or the reality of trying to characterize reality. I do not care if they dress things up in the most sophisticated Uncertainty Analysis. No amount of smart writing will elucidate Beyesian analysis or matirx inversion.
The public will seize with glee on the word “uncertain” and attack the mining project with sound-bites of “they know not what they do.”
Thus statisticians, mining proponents, their consultants, groundwater modellers everywhere will have to come up with a new name for a powerful idea. If they do not, it is, to me, incoceivable that we will ever see an uncertainty analysis in a public document that purports to justify a new mine.
We will have to fall back on sensitivity analyses which sound so pink. Or we will have to block things out in terms of best case analyses and upper bound analyses. This sound so reassuring and progressive.
For terminology counts in mining. Hence the incredible popularity of sustainable, responsible, license to mine, and so on in mining literature, academic circles, and public announcements.
Let me know your thoughts on this uncertain issue.