News & Articles
Complying with Water Quality Permit Limits:
The Role of Analytic Variability
Journal of Environmental Regulation, Spring, 1993
Current regulatory initiatives to set permit limits at “no detectable amount” or “no net increase” of toxic pollutants may cause false indications of noncompliance that may result from unavoidable analytic variation. Permit levels should be set at quantitation levels. Dischargers should know the difference between detection levels and quantitation levels and how the uncertainty interval of reported analytic data may work against them.
Procedures for calculating water-quality-based effluent limits, as required by the 1987 amendments to the federal Clean Water Act, are resulting in proposed permit limits that are below the levels that can be accurately measured by current analytical techniques. The concept of analytical error and the statistical nature of analytical results are not well understood by either dischargers or regulators. Regulators expect to impose permit limits at definite detection levels (levels at which a substance can be detected in samples). However, the degree of analytical variability or uncertainty at detection levels is unacceptably high. Chemists and EPA guidance recommend that limits be above quantitation levels (levels at which the concentration of a substance can be measured with a degree of confidence). Dischargers who understand how to calculate reliable quantitation levels may increase chances of permit compliance at quantitation levels and will be better able to defend against alleged violations that may be a result of laboratory error.
This article reviews the concept of unavoidable analytic variation, the terms used to describe detection and quantitation levels, and EPA’s attempts to account for analytic variation in the water toxics program. Two methods of approaching the issue of acceptable reliability of data near detection levels – relative uncertainty and confidence intervals from method precision equations – are illustrated. Understanding the concepts of analytical variation and quantitation has practical implications for negotiating permit limits and defending enforcement actions.
Unavoidable Analytic Variation
When analytical reports for the concentration of a chemical at one outfall show different results for different days, there are three possible explanations for the difference: the concentration did actually differ; the sampling process (collecting, preserving, shipping, and storing) introduced variation; or the difference is due to the inherent inaccuracy of analytical methods. The inherent inaccuracy, or unavoidable analytical variation, is significant near the detection levels.
Analytic variability refers to the inability of test methods to reliably and consistently measure the actual concentration of the substance being analyzed. The intrinsic imperfections associated with all analytic methods are referred to as method performance limitations. Laboratories cannot eliminate all of the imperfections in their performance capabilities, and these imperfections cause analytic variability. The American Chemical Society’s (ACS) Principles of Environmental Analysis1 states:
Analytical chemists must always emphasize to the public that the single most important characteristic of any result obtained from one or more analytical measurements is an adequate statement of its uncertainty interval.
Because of the legal implications of the National Pollutant Discharge Elimination System (NPDES) compliance monitoring, the uncertainty of the data is a legitimate concern. Determination of noncompliance, which could result in enforcement actions, must be based on data that are highly reliable. For example, if unavoidable analytical variability causes a true concentration of 4 to be measured as 6 and then compared with a permit limit of 5, that variability causes a false impression of noncompliance. Although the concepts of unavoidable analytic variability have been known for some time, the widespread use of permit limits at or near detection limits in the water program and the rapid expansion of demands on commercial laboratories have created problems.
There is general agreement among authoritative chemistry sources and EPA that quantitation levels should be used for legal compliance determinations. The ACS states unequivocally that detection levels should not be used for regulatory purposes:
Data measured at or near the limit of detection have two problems. The uncertainty can approach and even equal the reported value. Furthermore, confirmation of the species reported is virtually impossible; hence, the identification must depend solely on the selectivity of the methodology and knowledge of the absence of possible intereferents. These problems diminish when measurable amounts of analytes are present. Accordingly, quantitative interpretation, decision-making, and regulatory actions should be limited to data at or above the Limit of Quantitation.2
Although EPA has recently recommended that permit compliance levels be set at quantitation levels, EPA’s definition of minimum level for the permit compliance levels ignores current analytical practice. Chemists advise us to allow for interlaboratory variations and matrix interferences when interpreting results. In addition to inherent method performance limitations, differences in application of the methods result in a lack of reproducibility of data between laboratories analyzing the same sample using the same source method and perhaps the same instrumentation. Even with modern methods, different operators and conditions can cause different analytic results for the same sample analyzed in the same lab. Interlaboratory variation is greater than the variation resulting from repeated measurements in the same laboratory (intralaboratory variation) and can be substantial at the lower limits of the analytical method.
Matrix effects or interferences have a great impact on the actual detection level of a particular sample. Interferences are substances producing an unwanted signal that distorts the result. For example, high concentrations of petroleum hydrocarbons may mask trace amounts of chlorinated solvents in some methods. The analytical methods are usually developed using the target chemical in distilled or reagent-grade water. Environmental samples typically contain a wide range of chemicals, which may cause interference.
Detection and Quantitation Levels
There is great confusion on the use of the terms detection level and quantitation level. Major reference works in chemistry do not use identical terminology. EPA’s recent attempts to clarify these definitions have not been helpful. This section reviews the regulatory concept of method detection limit (MDL), which should be used to determine the actual detection level achieved by your laboratory in your wastewater, and two definitions of quantitation level, or the level at which numeric results can be believed.
Detection level is a term in general usage, but neither chemists nor the government define this term identically. The ACS defines instrument detection level and limit of detection. (See Exhibit 1.) EPA regulations define detection limit as the method detection limit and provide a protocol for determining an MDL in the Code of Federal Regulations. Only the MDL has a regulatory definition. The detection level is the concentration at which a substance can be determined to be present in a sample matrix. Specificity, or proper identification, of the material is the key to the definition. Detection level is a qualitative concept; no reliable quantitative (or measurable) information is given at the detection level.
ACS Guidelines for Reporting Data
|Analyte Concentration (in units of )||Region of Reliability|
|<3||Region of questionable detection (and therefore unacceptable)|
|3||Limit of Detection (LOD)|
|3 to 10||Region of less-certain quantitation|
|10||Limit of Quantitation (LOQ)|
|>10||Region of quantitation|
Instrument detection limit (IDL) is the best performance of a method and instrument in reagent water; that is, where there is only one element being analyzed in a sample matrix. Because it is determined from measurements in a clean sample, an IDL represents the lowest level that can be theoretically be achieved in the absence of matrix or sample processing effects. Although many laboratories report IDLs as “detection levels” on their lab reports, these numbers are a fiction suggesting that the data are more accurate than is actually the case.
The limit of detection (LOD) as defined by the ACS is “the lowest concentration level that can be determined to be statistically different from a blank” and is operationally defined as three times the standard deviation of measurements around the blank.
EPA defines MDL as “the minimum concentration of a substance that can be measured and reported with 99 percent confidence that the analyte concentration is greater than zero and is determined from analysis of a sample in a given matrix containing the analyte.”3 EPA is not always consistent in using the term “MDL,” as EPA has published numeric MDLs for many substances as part of the approved method,4 but these concentrations were determined in reagent-grade water, so they do not include matrix effects. MDLs can be determined in specific wastewater and water samples following the procedures contained in 40 CFR Part 136, Appendix B.
The MDL procedure requires analysis of a minimum of seven replicates of the matrix that must be carried through all the sample processing and analysis steps. An MDL differs from an IDL or LOD in that it includes effects due to sample matrix and sample processing. Like the LOD, an MDL is also an intralaboratory value – i.e., it is determined within a single laboratory, and different laboratories will achieve different MDLs. Therefore, the MDL does not account for interlaboratory variability. However, the definition of MDL is flexible enough to allow MDLs to be determined in any matrix, including laboratory reagent-grade water, natural waters, and wastewaters. Although there are many definitions of “detection level,” the proper usage is the MDL as defined in federal water permit regulations.
The quantitation limit is the minimum concentration of analyte that must be present before a method is considered to provide reliable and reproducible quantitative results. The two major definitions now in use were developed by ACS and EPA. The limit of quantitation (LOQ) is defined by the ACS in Principles of Environmental Analysis as “the level above which quantitative results may be obtained with a specified degree of confidence.” The recommended value for the LOQ is t en times the standard deviation of measurements around the blank. Even at ten standard deviations, there is an uncertainty of plus or minus 30 percent in the measured value at the 99 percent confidence level. An LOQ developed by one laboratory using one matrix may not be applicable to another instrument or laboratory.
The practical quantitation limit (PQL) is defined by EPA as “the lowest concentration of analytes…that can be reliably determined by the indicated methods under routine laboratory operating conditions.”5 Unlike the LOQ, which is specific to an individual laboratory, the PQL is meant to be an interlaboratory concept.
EPA uses PQLs to set realistic standards for contaminants in drinking water and derives PQLs from interlaboratory performance evaluation studies or from extrapolation from the observed MDL of a sample: “EPA believes that inter-laboratory studies, whether method validation or performance evaluation, are useful in establishing the PQL, but also believes that a multiplier of five to ten times the MDL is an effective way to establish the PQL.” EPA further explained that “the five to ten times the MDL rule…has been corroborated by evaluations of water supply performance data for the five inorganic contaminants in today’s rule.”6 EPA uses the PQL in a variety of contexts, including hazardous waste listings and delistings; land disposal restrictions; groundwater rules; and drinking water standards.7 However, EPA’s water toxics program distinguishes all of the currently available defined quantitation levels and has attempted to introduce a new term, the minimum level.
EPA’s Efforts to Account for Analytic Variability
EPA has not adequately addressed analytic variability when setting discharge standards, in establishing permit limits, or in enforcement proceedings. In Chemical Manufacturers Association v. U.S. EPA, the court recognized that analytical methods can be unrealistic at low concentrations and said that EPA must take analytical variability into account when setting limits:
The analytical methods for measuring pollutants become unreliable at the low concentrations the EPA has established in the limitations. Two different laboratories, each using acceptable methods, may measure the pollutant in a given sample and reach different results, yet neither lab may be demonstrably wrong. This has been referred to above as “analytical variability.”8
EPA responded that the statistical method used to calculate the limits included a variability factor, and the court upheld the technology-based limits because CMA had not met its burden of proof.
EPA has refused to consider analytic variability in the National Toxics Rule but stated that analytical uncertainty should be considered in permit negotiations or enforcement proceedings:
Analytical detection limits have never been an acceptable basis for setting standards since they are not related to actual environmental impacts. The environmental impact of a pollutant is based on a scientific determination, not an arbitrary measuring technique which is subject to change…The Agency does not believe it is appropriate to promulgate insufficiently protective criteria [e.g., criteria equal to the current analytical detection limits].
EPA does believe, however, that the use of analytical detection limits are appropriate for determining compliance with NPDES permit limits.9
Permit development is the critical time to consider analytical variability. EPA’s 1991 technical support document (TSD) for water-quality-based effluent limitations states that analytic variation should be considered when the concentration of the water-quality-based limit is below the analytical detection level for the pollutant of concern:
For most NPDES permitting situations, EPA recommends that the compliance level be defined in the permit as the minimum level (ML). The ML is the level at which the entire analytical system gives recognizable mass spectra and acceptable calibration points. This level corresponds to the lowest point at which the calibration curve is determined based on analyses for the pollutant of concern in a reagent water. The ML has been applied in determinations of pollutant measurements by has chromatography combined with mass spectrometry….
The minimum level is not equivalent to the method detection level,…. EPA is not recommending use of the method detection level because quantitation at the method detection level is not as precise as at the ML. It is not similar to the PQL, which is typically set as a specific (and sometimes arbitrary) multiple of the MDL. Because the PQL has no one definition, EPA is not recommending its use in NPDES permitting. Nor is it similar to other terms such as the limit of detection, limit of quantitation, estimated quantitation limit, or instrument detection limit.
The permitting authority may choose to specify another level at which compliance determinations are made. Where the permitting authority so chooses, the authority must be assured that the level is quantifiable, defensible, and close as possible to the permit level.10
EPA’s definition of minimum level in the 1991 TSD is not applicable in most contexts because the language is specific to organic chemicals determined by a gas chromatography/mass spectrometry methodology and is not applicable to other organics, metals, and inorganics.
The definition lacks operation specificity and does not allow for matrix effects. Minimum levels for most chemicals, including all metals, have not been defined or published by EPA. Although EPA distinguishes most of the known definitions of detection limit and quantitation level and criticizes the widely used PQL, EPA also fails to define minimum level on an operational basis. However, the introduction of the minimum level concept in the 1991 TSD is important, because it states that NPDES permit limit compliance should be based on measurements that are quantifiable and clearly recognizes that the MDL is inappropriate for permitting because of the lack of reliability. EPA recognizes analytic variation and specifies quantitation levels as the appropriate standard of compliance.
Previously, the government had successfully countered dischargers’ arguments that analytical variability is a defense to an enforcement proceeding by contending that the dischargers had waived the argument by verifying that discharge monitoring reports (DMRs) were accurate. These holdings are from older cases, for which the impact of analytical variation was not as great. However, this adverse precedent can be set aside if DMR results are reported with a statement of the analytical uncertainty.
Measuring Uncertainty in Analytical Results
A number of methods can be used to describe analytical variation, all of which depend on basic probability and statistics. For nonparametric measures, such as presence or detection of a chemical, the probability of a false positive or negative result is helpful in interpreting the data. However, permit limits and enforcement actions rely on precise numeric standards. The purpose of defining quantitation levels is to increase confidence that the reported concentration is accurate. All analytical data have some degree of unavoidable analytical variability, or uncertainty (i.e., a reported analytical value is only an estimate of the true value, which is not known except in laboratory standard solutions). If replicates of a sample are analyzed, that uncertainty can be quantified, usually in terms of the probability of the true value occurring within a range around the mean value.
Because analytical results are averages, the probability of a result at a certain value can be calculated. EPA-approved methods are developed through a method validation process, considering the bias, precision and accuracy, or recovery of each method. The published precision and accuracy equations can be used to develop the confidence intervals for a given analytic result. Another measure of analytic variability is relative uncertainty, based on the coefficient of variation and basic probability theory.
If a measurement is repeated many times under identical conditions, the results will be distributed randomly about a mean value. If an infinite number of such measurements were accumulated, the individual values would be distributed as the normal distribution. The standard deviation fixes the width of the normal distribution, and also includes a fixed fraction of the values making up the curve. About 95 percent of the measurements lie within plus or minus two standard deviations of the mean, and 99 percent will be within plus or minus three standard deviations of the mean. When values are assigned to the plus or minus standard deviation multiples, they are confidence limits. For example, ten plus or minus four indicates that values from six to fourteen represent the confidence interval.
Detection limit is defined in 40 CFR Section 136.2(f) as the “minimum concentration of an analyte (substance) that can be measured and reported with a 99 percent confidence that the analyte concentration is greater than zero….” The 99 percent confidence limit is three standard deviations from the mean.
False Positives or False Negatives
One way of quantifying analytical variability is the probability of a false positive or false negative result. A false positive, “Type I” or “alpha” error, is the probability of deciding a constituent is present when it actually is absent. A false negative, “Type II” or “beta” error is the probability of not detecting a constituent when it actually is present.
As noted above, detection limits are defined to minimize the occurrence of false positive results. By defining the MDL or LOQ as three standard deviations from the mean, EPA and ACS define a detection limit that has less than a 1 percent probability of producing a false positive (i.e., less than 1 percent of measurements will report the analyte as “present” when it is not). An EPA committee is considering proposing the reliable detection limit (RDL), defined as two times the MDL, or six standard deviations from the mean at the 99 percent confidence level.11 At the RDL there is less than a 1 percent probability of false positives and less than a 1 percent probability of false negatives. The reliable quantitation level is two times the RDL.
The Probability of a Result
The detection level does not provide any quantitative information. Because an analytic result at any number is a statistical estimate of the true value, one can argue that a positive test result at the MDL indicates only a 50 percent chance that the analyte is present at or above that number (MDL); that is, 50 percent of the time that an analytical result us reported at the MDL, the true concentration is less than the reported value, if not zero. A positive analytical result at the MDL does not provide a reasonable degree of certainty that the analyte is present at or above the MDL, or other level of regulatory significance.
Precision and Accuracy Equations
Bias, precision, accuracy, and recovery are terms used by chemists to define the quality of analytic data. Bias is a measure of systematic error, or a result consistently higher or lower than the true concentration. Precision is a measure of the closeness of a set of results. Accuracy is sometimes used as a combination of precision and bias, but is more correctly used to compare analytic results with a known or “spiked” concentration. EPA’s Section 304(h) Report explains:
“Accuracy is a measure of the closeness of an individual measurement to the true concentration. It includes both precision and bias. It is determined by analyzing a sample containing a known native or spiked concentration of the target pollutant. Accuracy is usually expressed as a percent recovery (R) or as a percent bias (R-100)….12
Each EPA-approved analytical method should have a statement of the systematic error or the method as part of the method validation process. The precision equations are published in the appendices to 40 CFR Part 136 and should indicate to the user what kind of performance to expect from a specific method.13 For example, the precision and accuracy equation for lead analyzed by EPA Method 239.2 is14:
X = 0.9430(C) – 0.504
S = 0.2224(X) + 0.507
SR = 0.1931(X) – 0.378
C = true value for the concentration, mg/L
X = mean recovery, mg/L
S = multilaboratory standard deviation, mg/L
SR = single-analyst standard deviation, mg/L
The precision and accuracy equation gives the standard deviation of the analytical requirements about the true value of the sample concentration and is derived from a comparison of the replicate results achieved by a number of government and research laboratories. They can be used to calculate the range of expected measurements at any desired concentration level by simply multiplying by the number of standard deviations from the true concentration for a desired probability level. The range of measurements that could be expected to occur 95 percent of the time for a specified true concentration (or regulatory limit) is estimated by multiplying the precision estimated from the applicable regression equation by 1.96 and adding or subtracting the resulting value from the mean recovery of the true concentration.
One statistical basis for determining reliability is the relative standard deviation (RSD, also known as the coefficient of variation (CV), s/x), which is a measure of precision. The RSD, or CV, is defined as the standard deviation divided by the mean. This statistic normalizes the standard deviation and facilitates making comparisons of analyses that have a wide range of concentrations. It is often then multiplied by 100 and expressed as a percentage. As defined, the RSD describes the unavoidable analytical variability at approximately a 67 percent confidence level (one standard deviation). Relative uncertainty is the RSD at the 99 percent confidence interval. At an increased confidence level of 99 percent (three standard deviations), the relative uncertainty becomes three standard deviations of the mean divided by the mean.15
Case Study: What is Acceptable Reliability?
There are a number of practical implications of imprecise use of “detection level” and misunderstanding of analytical variation. For permit limits written as “no detectable amount,” the difference between the detection level and quantitation level is apparent, because the quantitation level is five to ten times the method detection level. Analytical variation must also be considered in negotiating “no-net-increase” permit limits. To illustrate the concept of relative uncertainty and reliable or believable data, consider the data from a recent project.
The published detection level for analysis of lead by the graphite furnace atomic absorption technique, Method 239.2, is one part per billion (ppb). A discharger agreed to a no-net-increase-of-lead permit condition after collecting analytical data that indicated the net limit was not a problem. These data were analyzed using a method that could not provide the desired sensitivity. However, the laboratory reported data to significant figures of .1 ppb, far better than the theoretical capability of the analytic method used. The numbers reported on the lab reports were meaningless.
Relative Uncertainty for Lead
|Level||Definition||Relative Uncertainty||Reported Concentration (ppb)||Actual Uncertainty (ppb)|
|MDL||3s||100% (3s/3s x 100)||5||5±5|
|LOQ||10s||30% (3s/10s x 100)||17||17±5|
|PQL||5 x MDL = 15s||20% (3s/15s x 100)||25||25±5|
|10 x MDL = 30s||10% (3s/30s x 100)||50||50±5|
A year into the new permit, it was apparent that the discharger could not comply with the permit limit. The company performed MDL analysis on its effluent and found that the actual MDL ranged from 1 ppb to more than 10 ppb, with an average MDL of 5 ppb. The formula used to calculate permit compliance assumed a sensitivity of 1 ppb and could not work when numbers below the MDL were used. The problem would have been avoided if the company had understood the inherent uncertainty of the methods used or had estimated the confidence intervals at the analytical level needed using the precision and accuracy equations.
The relative uncertainty calculations for lead at the 99 percent confidence level for each of the detection and quantitation levels discussed above are shown in Exhibit 2.
The concept is illustrated in the last two columns above for an analysis of a water sample of an MDL of 5 ppb. From that MDL, the other detection and quantitation levels and the uncertainty associated with each can be calculated. The last column above shows that, although the “absolute” certainty (5 ppb) is constant for all detection and quantitation levels, the “relative” unavoidable analytical variability decreases from 100 percent at the MDL to 20 percent or 10 percent at the PQL.
The expected precision of the method is another way of examining the reliability of the results. Applying the equations given in 40 CFR Part 136, Appendix D (1991) for analysis of lead by GFAA, Method 239.2 for reagent water and EPA method Study 3116 for an effluent water, the published precision and accuracy equations give the expected performance at a 95 percent confidence level illustrated at Exhibit 3.
Expected Precision of Method 239.2
|Concentration (ppb)||Reagent Water||Effluent Water|
|Result and Accuracy||Range||Result and Accuracy||Range|
|3||2.8 ± 81%||0.5 to 5.1||3.73 ± 167%||-2.57 to 10.03|
|5||4.2 ± 69%||1.3 to 7.1||5.13 ± 140%||-2.1 to 12.33|
|7||6.1 ± 61%||2.4 to 9.8||6.53 ± 125%||-1.67 to 14.73|
|15||13.6 ± 52%||6.5 to 20.7||12.13 ± 98%||0.24 to 24.02|
|26||24 ± 49%||12.3 to 35.7||19.8 ± 86%||2.77 to 36.11|
|65||60.8 ± 46%||32.7 to 88.9||47.13 ± 76%||11.33 to 82.93|
The comparison of predicted confidence limits at various constituent concentrations shows that the expected accuracy is far worse than suggested by the relative uncertainty method.17 Exhibit 3 illustrates the impact of matrix effects or interferences on low-level analysis, because the confidence intervals for effluent water at a given concentration are almost double the confidence interval for reagent water.
As the concentration of the analyte increases, the accuracy of the method improves, but much more slowly than suggested by the relative uncertainty method. The term “believability” has been suggested to describe the situation in which the numeric value of the result is greater than the uncertainty interval. In this example, for effluent water, results above 15 ppb show less than 100 percent uncertainty, and results above 15 ppb would not include a negative or zero concentration in the 95 percent confidence interval.
Many of the definitions we have reviewed suggest that an uncertainty of about 30 percent is acceptable. For example, ACS states uncertainty of about 30 percent is expected from an LOQ at ten standard deviations from the mean. EPA has defined the PQL as providing an uncertainty factor of 30 percent. Some states define the compliance level as a level giving uncertainty of 20 percent at a 99 percent confidence level. However, use of the precision and accuracy equations suggests that an uncertainty of plus or minus 20 percent cannot be achieved until the analyte is 50 to 100 times above the detection level.
State water quality standards are calculated, and in many cases are imposed, without consideration of unavoidable analytical variability. In the case described above, the state water quality standard was 3 ppb and the quantitation level was at least 15 ppb.
When permit limits for some toxic chemicals are calculated directly from numeric water quality standards, it is possible for the calculated limit to be lower than the level at which the chemical can be measured reliably following routine, standardized analytical techniques. The American Chemical Society differentiates between detection level and quantitation level and recommends that permit limits be set at quantitation levels. At the detection level, a chemical can be reliably said to be present, (i.e., concentration not zero), but there is a large uncertainty as to what the actual concentration is. At the quantitation level, the concentration can be measured reliably with an acceptably low level of uncertainty.
EPA recommends using quantitation levels rather than detection levels for determining permit compliance limits. EPA suggests that the “minimum level” is the appropriate quantitation level for NPDES permitting. Other federal and state environmental programs, including hazardous waste site remediation and drinking water programs, use the practical quantitation limit for determining compliance.
NPDES permittees must negotiate limits that they can measure reliably at quantitation levels. Although the government recognizes the inherent difficulties posed by limits set below detection level, unavoidable analytic variability may not be a defense to liability in an enforcement proceeding. Permittees should recognize that the MDL is the only legal definition of detection level in water regulation, and should periodically have their laboratories calculate the actual MDLs achieved by their analyst on their wastewater. If effluent concentrations are within fifty to one hundred times the published detection limits, dischargers should be familiar with the precision and accuracy equations to calculate the expected performance of required methods.
1. 55 Anal. Chem. 2217 (1983).
3. 40 CFR Part 136, Appendix B.
4. Id., Appendix A.
5. 52 Fed. Reg. 25942, 25952 (July 9, 1987).
6. 57 Fed. Reg. 31800, 31806 (July 17, 1992).
7. Hazardous waste listings and delistings (55 Fed. Reg. 46354, 46365 (November 2, 1990), 38090, 38098 (September 17, 1990)); land disposal restrictions (55 Fed. Reg. 22535 and 22540 (June 1, 1990)); groundwater rules (52 Fed. Reg. 25942, 25944-25945 (July 9, 1987)); and drinking water standards (56 Fed. Reg. 26509, 26512 (June 7, 1991) and 57 Fed. Reg. 31776 (July 17, 1992)).
8. 870 F.2d 177, 230 (5th Cir. 1989).
9. 57 Fed. Reg. 60870 (December 22, 1992).
10. U.S. EPA, Technical Support Document for Water Quality-Based Toxics Control. Report No. EPA/505/2-90-001 (1991), pp. 111-112.
11. Keith, L.H., Environmental Sampling and Analysis: A Practical Guide (1992); Revising Definitions: Low-level Analysis, Environmental Lab (June/July 1992), pp. 58-61.
12. U.S. EPA Report to Congress, Availability, Adequacy, and Comparability of Testing Procedures for the Analysis of Pollutants Established Under Section 304(h) of the Federal Water Pollution Control Act at 3-3 (1988).
13. 52 Fed. Reg. 33547 (September 3, 1987).
14. 40 CFR Part 136, Appendix D (1991): “An interlaboratory study on the metal analyses by this method was conducted by the Quality Assurance Branch (QAB) of the Environmental Monitoring Systems Laboratory-Cincinnati (EMSL-CI). Synthetic concentrates containing various levels of this element were added to reagent water, surface water, drinking water, and three effluents. These samples were digested by the total digestion procedure, 4.1.3 in this manual. Results for the reagent water are given below. Results for other water types and study details are found in ‘EPA Method Study 31, Trace Metals by Atomic Absorption (Furnace Techniques),’ National Technical Information Service, 5285 Port Royal Road, Springfield, VA 22161 Order No. PB 86-121 704/AS, by Copeland, F.R. and Maney, J.P., January 1986.”
15. Taylor, J.K., Quality Assurance of Chemical Measurements (1987).
16. EPA Method Study 31: Trace Metals by Atomic Absorption (October 1985).
17. These precision and accuracy equations are based on interlaboratory studies by the best laboratories in the country, and the performance of commercial laboratories should be worse! Some of the equations published in 40 CFR Part 136, Appendix D, are known to have typographical or transcription errors, and you should review the method performance studies to find the correct equations.
At the time of publication, Harry F. Klodowski, Jr. was director of the Environmental Department at the Pittsburgh, Pennsylvania office of the law firm Doepken, Keevican, Weiss & Medved, P.C. He has negotiated water permits and resolved enforcement litigation for steel and chemical plants in Pennsylvania, Ohio, and Michigan, and has managed environmental assessments for major acquisitions and refinancings on behalf of borrowers and lenders.