There are standards—and there is THE standard

Kim H. Esbensen^a and Claas Wagner^b,*

^aKHE Consulting, www.kheconsult.com
^bSampling Consultant—Specialist in Feed, Food and Fuel QA/QC. E-mail: [email protected]

This series of columns has come to a natural half-way stage, at which it is worth reflecting a little. The basic principles for sampling of heterogeneous stationary lots, materials and systems have been covered—it is time start thinking of sampling from moving lots, dynamic systems and processes. But first: the Theory of Sampling (TOS) is proclaimed to be the only complete theory with which to address all the world’s many types of materials with a view of guaranteeing representative samples, and the column makes an effort, hopefully appreciated and easy to follow, to explain all the elements and their relationships in this endeavour. However, many standards, guidelines and norm-giving documents (CEN, ISO) already exist, which include elements of prescriptions for “proper sampling”, such as have been agreed upon by numerous task forces, committees etc. as being fit-for-purpose within the relevant scientific, trade and technological contexts addressed. There have been many such fits and starts towards a recommended sampling practice, but always in a partial sense only, indeed none cover the full breadth of all that is necessary to master representative sampling (with two spectacular exceptions: the iron ore and cement industry sectors). With so many partial recommendations available, when not in compliance with respect to TOS, there are objective, serious contradictions. Que faire?

The situation

The publication of DS 3077¹ represented the world’s first standard dedicated exclusively to representative sampling. Hardly any other standard is in full compliance with the appropriate TOS requirements laid out here, although partial elements can be found in many places, e.g. see the bibliography in DS 3077.¹ Two notable exceptions exist, however, the cement and the iron ore industries, which have been well serviced with excellent standards in this context for many years.

Non-compliance issues regarding such standards, guidelines, good practices as well as regulatory and legal requirements must be handled with insight and patience. Where found not to comply with TOS’ stipulations, it will be necessary to start a process of revision or updating of the relevant standards or norm-giving documents—which may be a lengthy process, and one that requires quite some logistical and organisational drive. While this is taking place, or when dictated by documented sampling variances that are too high (a key issue in quality control and assurance, QC/QA), it is always an option to employ more stringent quality criteria from a TOS-based approach than what is specified in today’s imperfect standards. As there are serious economic and societal consequences of non-representative sampling, simply staying with “following the book” is never a sound strategy, scientifically as well as regarding the economic outcome of decisions which will then in reality be based on inferior, non-representative data. DS 3077 has the overall objective of establishing a comprehensive motivation and competence for taking the stand relying only on fully TOS-compliant sampling procedures and equipment irrespective of the theoretical, practical, technological, industrial or societal context under the law. No standard is a legal document on its own and is therefore not legally binding; all trade agreements ruled by international standards are based on a set of voluntary agreements. To the extent that international law on the subjects treated in standards dealing with sampling aspects has been adopted, this law must be adhered to. International law implemented in national laws also takes precedence to non-legal documents in case of conflict.

Be this as it may, there are very many advantages in not being complacent with the fact that sampling issues are mentioned in the existing body of relevant standards and norm-giving documents. Mentioning is not enough, only the principles guaranteeing representativity matter. A directed effort has been in place for some five years, involving a systematic critique of selected standards, specifically with respect to the full set of sampling errors outlined in TOS. Two examples of this work are presented below, which suffice to show how one should approach any part of a standard etc. that purports to recommend proper sampling procedures and equipment etc.

Analysis of sampling standards for solid biofuels

Assessment of all sampling procedures from CEN standards for sampling solid biomass (CEN/TS 14778 part 1 and part 2)²,³ has shown that most of the recommended procedures do not lead to a fully satisfactory result, a representative sample. Correct delineation and extraction of many standardised methods as well as depicted, and thus recommended, tools and equipment are not ensured. While for grab and shovelling methods, correct delineation and extraction is hardly ever possible, other recommended sampling methods lack sufficient specification regarding application conditions, which invariably increases the potential for incorrect sampling error effects. Table 1 gives an overview of the evaluation results with respect to potential incorrect sampling errors (ISE) caused by the methods stated, recommended or allowed in the standard for primary sampling CEN/TS 14778.^2,3 ISE comprise the three so-called bias-generating errors: Increment Delimitation/Delineation Error (IDE), Increment Extraction Error (IEE) and Increment Preparation Error (IPE), all concerning sampling equipment and sampling procedures. The full assessment of these sampling standards can be found in Wagner & Esbensen.⁴

Table 1. Assessment of incorrect sampling errors of CEN/TS 14778.

Insufficient specifications and the existence of incorrect sampling errors must under all circumstances be eliminated in sampling standards as the result will unavoidably be an inconstant sampling bias, always and for ever out of control; it is not possible to make any bias correction regarding the sampling bias, DS 3077¹ and Esbensen & Wagner.⁵ Incorrect sampling methods, room for personal interpretation and the vertical standardisation approach of CEN specifying different procedures for each material group makes sampling a complicated issue with a highly uncertain and varying validity. Any procedure and standard that has not eradicated all such potential sampling bias elements, as illustrated above, does not comply with TOS’ stringent and demands for sampling correctness. The result is always a biased sampling procedure—which is always unacceptable. The full assessment of CEN/TS 14778 has been published, but so far no reaction or response has been forthcoming.⁴ 

Analysis of grain sampling guide

The “Home Grown Cereals Authority” (HGCA) is a division of the “Agriculture and Horticulture Development Board” based in the UK, which is mainly responsible for research and knowledge transfer in the cereal and oilseed sector. In 2013 the HGCA published a guide on grain sampling to define key requirements for effective grain sampling at various process locations from harvest to storage until departure and arrival of the grain.⁶ Besides physical extraction of a grain “sample”, focus is also on monitoring moisture, temperature, pests and moulds, especially mycotoxins. The described sampling practices therefore must have an obligation to contribute to ensure procedures that reliably are able to assess harvested grain quality, to protect this quality level throughout the storage phase as well as to determine the quality level after storage (before transportation to buyer) and upon arrival at the buyer. For various commodities the latter two aspects (differences in quality level at departure vs quality level at arrival) have in the past caused major law cases, often due to inappropriate or inadequate sampling procedures. Besides these kinds of discrepancies which cause serious economic disputes, extraction of representative grain samples is also crucial with regard to impurity detection (e.g. GMO quantification, toxins), as regulated by international standards (e.g. ISO 24276:2006⁷). Table 2 gives an overview of the evaluation results for the HGCA⁶ with respect to potential TOS-incorrect sampling errors. The full assessment can be found in TOS forum.⁸

Table 2. Assessment of incorrect sampling errors of HGCA sampling guide.

This assessment shows that most of its recommended sampling procedures and equipment (for both primary sampling and sub-sampling) do not lead to a representative sample. The guide’s sampling procedures have a high error potential for incorrect sample delineation and extraction, which unavoidably will lead to a significantly detrimental, or even fatal sampling bias.¹ Most of the guide’s recommended sampling equipment, when rated with TOS criteria, reveal major incorrect sampling errors (ISE), vastly jeopardising grain control validity.

It is noteworthy that the body responsible for the HGCA guide undertook a careful response to the above critique, which was published in TOS forum (see box section).⁹

It is in the interest of the science of sampling to bring this kind of discussion to the attention of everybody interested in representative sampling. While the present authors of the critique of the HGCA⁶ do not agree with most of the “reasons for lowering the standard w.r.t. representativity” in the rebuttal (see above), both science and industry will benefit from the clearly stated argumentation vs the original critique. It is, as always, up to the reader to form his/her own conclusions based on the evidence presented pro et con.

Sampling for GMO risk assessment

Currently an EFSA-funded project is a.o. engaged in a similar critique of all standards and norm-giving documents governing sampling for GMO risk assessment. The project reports will, after approval by EFSA, be available on the appropriate homepages within the EFSA portal.

Examples of too glib recommendations

For want of space, we end this column by showing a few examples “from undisclosed standards” of a few “recommended” sampling procedures/equipment, which would not under any circumstances find acceptance under the systematics of the Theory of Sampling, TOS (Figures 1–5).

Figure 1. From the geological realm. Sub-sampling a drill core (vertical) with an aim of providing analytical samples for diverse analytical modalities [SEM, porosity and permeability measurements (poro-perm), thin section (petrological microscopy), XRF, XRD]. Although a serious attempt at creating a standardised sub-sampling scheme (left: original recommendation), there is clearly room for improvements as seen in the centre and right-most evolutions towards the smallest coeval sample volume, i.e. the maximum possible joint sample support. There are plenty of arguments from geologists to the tune: “the rocks in the drill core are pretty well homogenous, so some 15–30 cm vertical separation does not matter much… ” or endless variations on this theme, which, however, completely misses the point: There is no need deliberately to create significant IDE and IEE errors in the sampling process (which is tantamount deliberately to create the fatal sampling bias). Also, homogenous materials do not exist in the real world, especially not in the poly-phase heterogeneous worlds of rocks… While there may very well be rocks of particular low vertical heterogeneity that need drilling (e.g. limestone/chalk oil reservoir rocks), setting of a standardised sub-sampling scheme for all reservoir rocks based on this scenario can only lead to significant sampling bias. Allowing for this is not the role of a standard.

Figure 2. A particularly ill-conceived recommendation of a “grain stream sampler”. When this example was used as a basis for an exam question in a PhD course on “Representative Sampling, TOS” a student wrote: “The mind boggles!”.

Figure 3. A recommended sampler with a much better chance of being approved by TOS—although the conditions under which this equipment is to operate are totally missing, making it open to several fundamental uncertainties.

Figure 4. A potpourri of verbatim quotes from discussions in sampling committees and fora. The mind boggles at the lack of respect for representativity, while logistics, practicality and economics would appear to be the only drivers. The effect of letting such proxies dictate sampling procedures, operations and equipment alone was discussed thoroughly, and dismissed, by Esbensen et al.^10,11 and Minkkinen et al.¹²

Figure 5. Using a cylindrical coring tool for cheese sampling does not allow a representative sample of the highly irregularly distributed components of a mature blue cheese. Only the two right-hand approaches will pass muster for TOS.

The reader is invited to try to determine which sampling error(s) are compromised in each specific example. It is not relevant to refer to the specific standards from which the examples originate; they are shown here in complete anonymity with the sole purpose of illustrating that sampling is not a game in which anything goes… More seriously, they are examples of what can happen when committees are guided by a regimen of consensus where truly anything goes, as long as it is unanimously voted and agreed on… Pierre Gy often used to deliver a wry comment on this state of affairs in his lectures and courses: “With this approach a committee could vote that Newton’s second law no longer applies”. The few examples are a vivid illustration to this dictum—very many “recommended” sampling procedures and equipment are nothing but a showcase of not having invested the necessary effort to investigate the basics of TOS principles. But, there is always room for improvement.

Summary

There is no need for unnecessary confrontations, but there is a need for absolute clarity with respect to the responsibility carried by international (and national) standardisation authorities. There is no excuse for recommending non-compliant sampling procedures and equipment; the result can only be inferior sampling and inferior, indeed compromised, decision making. A chain is only a strong as its weakest link. TOS-compliance is the missing link in very many standards etc. There is only one remedy—get involved, get TOS literate!

There are plenty of relevant courses, lectures, consulting companies, experts on the subject matter of representative sampling, all contributing and doing a remarkable job in the last 15 years (for some up to 40 years), but none will receive specific identification here. All the reader needs is a willingness to start looking for the singular operative characteristic: representativeness—as in representative sampling and the Theory of Sampling (TOS).

Que faire?

Start here: DS 3077!¹

References

1. DS 3077, “DS 3077. Representative sampling—Horizontal Standard”. Danish Standards (2013). www.ds.dk
2. CEN/TS 14778-1:2005 Solid Biofuels. Sampling. Part 1: Methods for Sampling. British Standards Institution, London, UK (2006).
3. CEN/TS 14778-2:2005 Solid Biofuels. Sampling. Methods for Sampling Particulate Material Transported in Lorries. British Standards Institution, London, UK (2006).
4. C. Wagner and K.H. Esbensen, “A critical review of sampling standards for solid biofuels – Missing contributions from the Theory of Sampling (TOS)”, Renew. Sust. Energ. Rev. 16, 504–517 (2012). doi: http://dx.doi.org/10.1016/j.rser.2011.08.016
5. K.H. Esbensen and C. Wagner, “Theory of Sampling (TOS) versus measurement uncertainty (MU)—a call for integration”, Trends Anal. Chem. (TrAC), 57, 93–106 (2014). doi: http://dx.doi.org/10.1016/j.trac.2014.02.007
6. HGCA Grain Sampling Guide. HGCA Publications, Warwickshire (2013). http://www.hgca.com/media/248889/grain_sampling_guide_2013.pdf (accessed February 2014).
7. ISO 24276:2006 Foodstuffs—Methods of Analysis for the Detection of Genetically Modified Organisms and Derived Products—General Requirements and Definition. International Organization for Standardization (ISO), Geneva, Switzerland (2006).
8. C. Wagner and K. Esbensen, “A critical assessment of the HGCA grain sampling guide”, TOS forum 2, 16–21 (2014). doi: http://dx.doi.org/10.1255/tosf.18
9. D. Bhandari and K. Wildey, “Letter in response to ‘A critical assessment of the HGCA grain sampling guide’ published TOS forum Issue 2”, TOS forum, 4, 4–4 (2015). doi: http://dx.doi.org/10.1255/tosf.36
10. K.H. Esbensen, C. Paoletti and P. Minkkinen, “Representative sampling of large kernel lots – I. Theory of Sampling and variographic analysis”, Trends Anal. Chem. (TrAC) 32, 154–165 (2012). doi: http://dx.doi.org/10.1016/j.trac.2011.09.008
11. K.H. Esbensen, C. Paoletti and P. Minkkinen, “Representative sampling of large kernel lots – III. General Considerations on sampling heterogeneous foods”, Trends Anal. Chem. (TrAC) 32, 179–184 (2012). doi: http://dx.doi.org/10.1016/j.trac.2011.12.002
12. P. Minkkinen, K.H. Esbensen and C. Paoletti, “Representative sampling of large kernel lots – II. Application to soybean sampling for GMO control”, Trends Anal. Chem. (TrAC) 32, 166–178 (2012). doi: http://dx.doi.org/10.1016/j.trac.2011.12.001