For more information on the Final Rule, please contact Adrian Hanley (hanley.adrian@epa.gov) or 202-564-1564. Documents relating to this rule are included in the EPO list under regulations.gov. The file number is EPA-HQ-OW-2014–0797. The acquisition of data points at this maximum concentration leads to a distribution of the measured concentrations. As an example, Figure 2 shows the resulting distribution (histogram) of measured chlorobenzene concentrations obtained from 50 injections with a peak of 0.05 micrograms per litre (μg/L) according to USGS Method O-4127-96 (Appendix 2020, Connor et al., 1998). Questions and comments on the proposed methodology update rule for 2015 and the EPO`s responses. In 1992, the NWQL began using the EPA (U.S. Environmental Protection Agency, 1984, 1997 [Note: Revision 1.11 has remained unchanged since 1984]) to determine DMLs for two methods developed under the National Water Quality Assessment Program. These methods were developed for pesticides in water by solid phase extraction C-18 with gas chromatography/mass spectrometry analysis (analysis plans NWQL 2001-2010; Zaugg et al., 1995) and pesticides in water using carbopak-B solid phase extraction with high-performance liquid chromatography analysis with diode array detection (Appendices 2050-2051; Werner et al., 1996). For these methods, the reporting level equivalent to the MDL determined using the USEPA method for all analytes in Annexes 2001/2010 (Zaugg et al., 1995, Table 9) and for 20 of the 41 analytes in Annexes 2050/2051 (U.S. Geological Survey National Water Quality Laboratory Technical Memorandum 98.03A, 1997) was prepared. The reporting conventions used for these methods were described in detail in U.S. Geological Survey National Water Quality Laboratory Technical Memorandum 94.12 (1994).

The usepa method was chosen because it was considered by the NWQL to be a generally accepted method for determining MDLs. It is required by the USEPA as an acceptable performance test for detectability of analytical methods for laboratories using UEPA methods. In addition, the USEPA MDL method is a statistical approach that can be applied consistently to all NWQL methods. The MDL method is designed as a simple technique for estimating the detection limit for a variety of physical and chemical methods. The previously published version of the MDL Method (Revision 1.11) states: “The method detection limit (MDL) is defined as the minimum concentration of a substance that can be measured with 99% certainty and that the concentration of the analyte is greater than zero and is determined from the analysis of a sample in a given matrix containing the analyte.” The limitations of the USEPA MDL method include the normal distribution and constant standard deviation assumptions over the low concentration range from peak concentration used to determine MDL concentration to zero concentration. These and other limitations and considerations have been discussed at length elsewhere, e.g. Keith, 1992; Eaton, 1993; Gibbons, 1996; Gibbons et al., 1997a; Hall and Mills, 1997. MDL is generally determined using a minimum number of peak replicas (n > 7) measured over a short period of time. This provides only a narrow estimate of the overall variation of the method and thus of the standard deviation(s). If the USEPA procedure is performed in this manner, it is particularly unsuitable for production laboratories where there are multiple sources of method variation, including multiple instruments, instrument calibrations, and equipment operators; And for methods that require preparatory steps, there are several preparatory and employee events. Sometimes a lab can calculate a USEPA MDL that is so weak that it produces a signal during routine analysis that cannot be reliably distinguished from instrument noise. This is most often done for single-device, calibration, and/or single-operator tests that result in low standard deviation estimates.

In this case, the actual probability of false positives to the USEPA MDL is likely to exceed the desired maximum of 1%. Therefore, MDL must be determined using routine conditions and procedures. This means that the determination is carried out over a longer period of time using all methodological tools and as many variables as possible to obtain a more accurate and realistic measure of the standard deviation near the MDL. Another important assumption of the USEPA MDL calculation is that the frequency distribution of the sequentially lower replication peaks of the concentration and thus the standard deviation of the distribution (not the percentage of relative standard deviation) becomes constant at a low concentration and remains constant until zero concentration. Standard deviations calculated from low concentrations become similar because small differences in the decreasing signal cannot be adequately measured. A graph representing standard deviations for different peak concentrations is shown in Figure 4. The USEPA method recommends reducing the peak concentration to successively lower concentrations in an iterative process to ensure that the virtually constant standard deviation range near the MDL has been reached. The frequency distribution of concentrations measured at the low concentration peak used to determine MDL is assumed to be normally distributed in the USEPA method and represented by the bell-shaped curve in Figure 3. In addition, one or more standard deviations of this distribution are displayed.

Reporting a detection when no substance is present is called a false positive. The USEPA MDL is designed to control false positive results at a 99% confidence level in an ideal matrix. Reporting of detection of a substance at MDL concentration in a blank sample or a sample that does not contain the analyte should be rare (less than or equal to 1%). Therefore, it is unlikely that a signal representing the presence of a substance in a sample at the MDL concentration would be erroneous. The amendments include revised methodologies published by CEPOL and voluntary consensus standardisation bodiesVoluntary consensus standardisation organisations national or international organisations that plan, develop, define or coordinate voluntary consensus standards using agreed procedures. These bodies are defined by openness, balance of interests, due process, the existence of an appeal process and a consensus process, such as ASTM International and the Standard Practice Committee. EPA added to 40 CFR Part 136 certain methods reviewed under the Alternate Test Procedures (ATP) program and clarified the agency`s approval procedures for national and restricted ATPs. In addition, the EPA revised the method for determining the method detection limit (MDL). In summary, taking into account the assumptions of constant standard deviation at low concentration, normal distribution, and minimal matrix interference, the EPA`s MDL is the concentration at which no more than 1% of blank measurements result in false positive detection (Fig. 6). Therefore, at concentrations greater than or equal to MDL concentrations, the evidence must be true 99% of the time.

The LFS MDL is calculated in equation 1: The 2016 revision of the MDL procedure (Revision 2) differs from the revision 1.11 of the MDL procedure in three main respects. The U.S. Environmental Protection Agency (USEPA) Limit of Detection (MDL) method is described as the minimum concentration of a substance that can be measured with 99% certainty and has been reported to have a concentration of the analyte greater than zero (U.S. Environmental Protection Agency, 1997) and is based on the approach of Glaser et al. (1981). MDL protects against misreporting the presence of a compound at low concentrations when the noise and the actual signal of the analyte are indistinguishable. The concentration of MDL does not imply the accuracy or precision of the quantitative measurement. Assuming a constant standard deviation between the low concentration peak and zero, the frequency distribution of repetitive and low-concentration peaks of zero concentration (analyte not present in the samples) is superimposed (Fig. 5). This is the distribution one would expect when a signal from actual instrumental noise or undoped analyte, or both, is measured in a blank sample series. Since it is generally not practical to measure noise in repeated blank samples, the frequency distribution of the low-concentration peak is used as a hypothetical blind frequency distribution.

Hypothetical blind measurements are used to calculate the concentration at which no more than 1% of blind measurements result in a false positive being reported.