Paving a Path to More Ethical Uranium Mining with Environmental Statistics


  • Author: Lillian Pierson P.E.
  • Date: 08 Apr 2014
  • Copyright: Image appears courtesy of iStock Photo

When one thinks of the mining industry, environmental good stewardship and the concept of “ethical mining” don’t generally come to mind. In other words, the concept of “ethical mining” doesn’t exist… or, does it? At Coles Hills, Mr. Walter Coles of Virginia Uranium, Inc. is setting a novel and noble precedent in the world of uranium mining.

Mr. Coles’ family has inhabited and farmed the Coles Hills property for over 200 years, but more recently they’ve discovered that the family land is home to a sizeable uranium deposit. When similar discoveries have been made elsewhere landowners generally opt to start mining land resources immediately, but this is not the case at Coles Hills. While the State of Virginia’s moratorium on uranium mining is an influencing factor, Mr. Coles is genuinely concerned about the environmental implications of his future decision-making. He wants to be certain that any potential mining activity would not have a long-term adverse effect on the environment around Coles Hills. Evidencing this concern, Mr. Coles has been investing significant company resources to fund long-term, baseline environmental studies of the site.

thumbnail image: Paving a Path to More Ethical Uranium Mining with Environmental Statistics

This type of long-term detailed baseline study has not been done elsewhere. Because of the depth of these studies, research scientists Dr. Robert Bodnar and Denise Levitan have ascertained volumes of information about the geomorphological processes that contributed to the deposit’s formation and the baseline chemistry of the surface water in and around the Coles Hills deposit. What they have discovered has been quite revealing.

Prior to mining, environmental regulatory agencies demand that geochemical baseline studies are conducted and that concentrations of toxic metals like uranium, lead, and arsenic are reported. Although regulators demand this, they don’t suggest or advise on what protocols should be employed. In fact, industry-wide, there is no consensus on protocols, so environmental consultants are left to their own devices in reasoning out methods for sample site selection, evaluation of temporal trends, approaches for statistical analysis, and reporting of baseline data.

Levitan sampled the surface waters at Coles Hills for about 2 years and was greatly challenged by the quantity of non-detect values in her sampling results. Non-detect values occur in samples where the parameter being tested is not detected, either because parameter concentrations are below the detection capabilities of the analytical equipment or because the parameter being tested is not present. In this study, Levitan was getting back a lot of non-detect values because the concentration of uranium in the samples was below the detection limit of the analytical equipment.

After some research she discovered that, often times, important data related to the geochemical constituents of surface waters is not being captured or reported due to the limitations of analytical detection methods and due to the over-simplified treatment of non-detect values that accompany these limitations. In order to capture an accurate picture of the environmental and geochemical baseline from a sparse dataset, Levitan was forced to employ advanced statistical methods. In doing so, she developed a ground-breaking set of “best practice” methods for the responsible treatment and reporting of non-detect data in mining site evaluations.

To capture an accurate picture of the environmental and geochemical baseline from a sparse dataset, Levitan was forced to employ advanced statistical methods. In doing so, she developed a ground-breaking set of “best practice” methods for the responsible treatment and reporting of non-detect data in mining site evaluations.

To help her in her quest to make sense of the Coles Hills non-detect data, Levitan turned to the author of the Wiley book Statistics for Censored Environmental Data Using Minitab and R, 2nd Edition, Dr. Dennis Helsel. Dr. Helsel has abundant experience in applying advanced statistics to solve problems in environmental science and he developed the NADA package in R. In the Coles Hills study, Levitan used this package as well as many of the non-detect statistical methods that were described by Dr. Helsel in his book. More specifically, Levitan utilized the following statistical methods against her non-detect data to define which method would provide the most accurate estimation of the Coles Hills non-detect data values.

• Regression on Order Statistics (ROS) Methods (for small datasets)
• Expectation-Maximization (EM) Imputation Method (for large datasets)
• Maximum Likelihood Estimation (MLE) Function Method (for large datasets)
• Substitution Method (for datasets with only a small fraction of below detection limit values)
• Survival Analysis Method (for datasets where less than 50% of the values are below detection limit)

In evaluating the suitability of each of these methods, Levitan compiled a set of best practices for future evaluations of this type. In an exclusive interview for Statistics Views website, when asked about these best practices, Levitan summarized them concisely by stating, “The first step is acknowledgment of the problems of non-detect data and the existence of methods to analyze them. The selection of a method is heavily dependent on the characteristics of the dataset and the ultimate questions that you are addressing with statistical analysis. In order to determine what techniques to use, you first need to evaluate your dataset: how many data points, how many variables, how many non-detects vs detects, how many detection limits per element, whether data follow any sort of pattern, etc. You then need to consider what information you want to determine from your data: summary statistics, comparisons between datasets, values for modeling, correlations, multivariate analyses, etc. Once you have determined what kind of data you have and what you want to do with it, you can compare the pros and cons of the different methods and make an informed decision on which (method) best suits your needs.”

When considering the proportion of censored data and the size of the data generated in the Coles Hills study, Levitan found that Regression on Order Statistics (ROS), Maximum Likelihood Estimation (MLE), and Expectation-Maximum (EM) were the most appropriate methods to use in handling the data. She discovered that the estimated statistics generated by each of these three methods was very similar, but concluded that the ROS method was the best method to use because the ROS medians were most similar to the actual median values in the dataset. Due to the moratorium on uranium mining in the State of Virginia, and the money provided by Mr. Coles to fund the environmental studies, there is still time for Dr. Bodnar, Denise Levitan, and others to continue their baseline studies around the unmined uranium deposit located at the Coles Hills site.


(1) Levitan, D.M., Schreiber, M.E., Seal, R.R. II, Bodnar, R.J., Aylor, J.G. Jr., Developing protocols for geochemical baseline studies: An example from the Coles Hill uranium deposit, Virginia, USA. Applied Geochemistry (2014), doi:
(2) Martín-Fernández, J.A., Hron, K., Templ, M., Filzmoser, P., Palarea-Albaladejo, J., 2012. Model-based replacement of rounded zeros in compositional data: Classical and robust approaches. Computational Statistics & Data Analysis, 56(9): 2688-2704.

Related Topics

Related Publications

Related Content

Site Footer


This website is provided by John Wiley & Sons Limited, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ (Company No: 00641132, VAT No: 376766987)

Published features on are checked for statistical accuracy by a panel from the European Network for Business and Industrial Statistics (ENBIS)   to whom Wiley and express their gratitude. This panel are: Ron Kenett, David Steinberg, Shirley Coleman, Irena Ograjenšek, Fabrizio Ruggeri, Rainer Göb, Philippe Castagliola, Xavier Tort-Martorell, Bart De Ketelaere, Antonio Pievatolo, Martina Vandebroek, Lance Mitchell, Gilbert Saporta, Helmut Waldl and Stelios Psarakis.