Sept 2018

Triangle NoTES


September 2018

I have been discussing what can only be described as perceptions of analyses based on assumptions. It is hard to not succumb to some of the assumptions based only on “common sense” but there are more things to consider when quality is involved.


One of the biggest points of confusion is that between precision and accuracy. A precise analysis is not always an accurate analysis, and an accurate analysis may not be that precise. Precision is usually the basis for the desire to look for the lowest detection limit possible. It seems logical that the lower the detection limit the better the analysis, but that is not always the case. If you are dealing with low levels near the detection limit, this can be a good move, but if you are looking for higher levels it may not be so. The fact is a detector can become saturated by a compound or compounds, which may require an additional dilution for the analysis to be completed. This means the detection limit for the sample is not the same as the limit at the analyzer. That analyzer limit is what you may be getting when you ask about the detection limit. This is especially true for multiple compounds reported using only one compound as the calibration or in the case of a method such as TO-15 which actually is set to condense larger sample volumes prior to injection.


Method 25 and 25-C make the detection limit very difficult to determine because using just the detection limit for propane ignores the other peaks which may elute. This is why we just use a very low calibration point. To put it into perspective, if you were have an analysis of C1 – C6 hydrocarbons using just one compound for the calibration, the detection limit for one peak would be very low, but the multiplication of that single peak limit by every other normally reported peak not detected would make it much larger. Thus, if you only had one compound detected you would be unable to determine how many other compounds may have been present below that detection limit. It is common to assume there were none, but that is an assumption that can easily be incorrect. Such a situation would impact the accuracy of the results even though the precision may be great.


The accuracy question is why audits are so important. If the results can be checked against known concentrations, the general accuracy can be determined. It does not mean that some of the analyses are without problems, but it is a means to confirm the general accuracy of the laboratory and its procedures. I believe everyone knows the required accuracy for Method 25 is +20% from the known in order to pass, which is also the requirement used for the few Method 25-C type audits. To compare this with the TO-15 methodology, that requirement is +30% at ambient levels. There is no requirement for higher concentrations which are often where the method is used. In the few instances of a Method 25 audit being used in a Method 25-C type analysis with us, I have not been told of any failures, but I do know of two which passed the requirements. I also know that we passed a full Method 25 audit which was slightly above the 50 ppmC limit.

Another aspect of accuracy is whether the results are reproducible, for which a TO-15 analysis duplicate analysis performed from the same canister must indicate a maximum of a 25% variation between the two analyses. As Method 25/25-C reports the average of triplicate injections from the same canister and there is a requirement for a < 5% relative standard deviation for the analysis to be valid, there is no real comparison between the requirements.



The various other types of analyses utilizing single injection from a whole gas sample probably will not have any similar requirements as does TO-15. The ASTM and other non-governmental types of methods seem to be more flexible than the EPA or other governmental reference methods, such as the SCAQMD methods, for example. Even in the case of the relatively common C1-C6 hydrocarbon analysis, there are no audits available through the TNI program for which a basic accuracy check may be performed.


Another aspect of the assumption regarding the use of most methods is the fact that there will be unknown compounds. In many cases what is in the sample stream does not correspond with the target compound list for the method. Even in the C1-C6 type analyses there are differences between the various carbon compounds which cannot be determined unless the compounds are individually identified. For example, early on in the Method 25-C program, we would have what we called a “fingerprint analysis” performed. This was simply a GC/MS scan of the sample to determine the top 10 – 20 compounds present. This resulted in the majority of compounds being reported as merely a C11, C12, or even C13 compound, with no information on the structure or specifics on the compound. Of course, when the analysis included the TO-14 or TO-15 target list, there were compounds reported, but in much smaller concentrations than the major constituents of the “fingerprint analysis”. If you look at the AP-42 for landfill gas, the TO-15 target list is clearly represented, but not that the constituents are identified as it really only represents the hits from the target compound lists used for the analyses.


To put all of this into perspective, there was a large project in the early 1990s where, in addition to the Method 25 analysis on some 15 sources at a facility, it was requested that we determine the exact constituents for each source. Working from the MSDSs to determine what was potentially being emitted, we set up a sampling plan to cover those compounds and any additional compounds which could be added to the analysis. The TGNMOC program analytical cost at the time was about $10,000.00 but the additional program to identify the species of the carbon was in excess of $250,000.00. The costs to run several sampling trains for the different compounds for each of the sources hopefully was not that magnitude of difference, but it was still several times more given the increase in testing crews needed. The final question came down to whether that additional cost really gave proportionally more usable data because there was still the possibility something was missed, whether the concentrations were comparably accurate, and whether the non-detected compounds were properly included since many of them were not expected to be in the samples. Even though this was over a quarter of a century ago, some of the same issues are still present today in sampling programs. Thus, the perceptions of quality information may not give any significant amount of usable data and may just act to increase the costs.


Wayne Stollings

Triangle Environmental Services, Inc.


Wstollings@aol.com


P.O. Box 13294 122 US Hwy 70 E

Research Triangle Park, NC 27709 Hillsborough, NC 27278

(919) 361-2890 (800) 367-4862 Fax: (919) 361-3474