The acceptable (meaningful) variability of analytical methods

Leaching tests and percolation tests intend to mimic the natural or man-made conditions where waste could be stored or used. The tests are therefore naturally different from one country to another, since the natural conditions (rain, evaporation, drainage, frost, storage conditions, etc.) and the conditions they mimic are different.

The variability for a given leaching or percolation laboratory test should nevertheless be under control. This is seldom the case in practice, as shown below.

Leaching tests are, from a chemical point of view, partial extractions and the result depends on strict compliance with the test conditions (particle size, leaching, duration, temperature, agitation, solid/liquid separation – decantation, centrifugation, frontal filtration, tangential filtration, filter pore size, tolerance of a filter cake, volume of leaching to be filtered, etc.) to obtain good repeatability and reproducibility (Hennebert et al. 2022).

The variability of standardized leaching test is high compared to the variability of total extraction methods (same reference), and frequently exceeds what is significant for a normal distribution (95% of the data of repetitions included in the range (mean – 2 standard deviations ; mean + 2 standard deviations). The variability is conveniently expressed as the relative standard deviation (RSD – the standard deviation divided by the mean) also called coefficient of variation (CV). This ratio is unitless or can be expressed in percentage. It allows immediate comparison of variability of methods. As there are no negative concentrations, the lower limit of the range should not be negative, and the standard deviation/mean, the CV, should be < 0.50 = < 50%, and for the 99% confidence interval should be < 0.33 = < 33%.

 

 

Figure 1 : The distribution of repeated measurements of a standardised analytical method is bell-shaped. As the minimal observed value cannot be lower than zero, the variability is limited CVr or CVR < 0.50 = 50% (95^th centile confidence interval of the mean) or < 0.33 = 33% (99^th centile of the confidence interval of the mean)

 

The variability of the EN 12457 series ((batch leaching test, 2 or 10 litres of deionised water/kg material grinded to 4 or 10 mm, shaking for 24 h, decanting, centrifugating if filtration is slow and filtrating with 0.45 µm membrane) is too high. From the validation data, the inter-laboratory variability CVR exceeds in some cases the value of CVr or CVR < 0.33 = 33% or < 0.50 = 50%. In these case, the intra- or inter-laboratory repetitions will give values between the limit of quantification and … something, with an average value statistically not different of zero…

 

Figure 2: Variability of intra-laboratory analysis, intra-laboratory leaching and analysis CVr and inter-laboratory leaching and analysis CVR (source: EN 12457 series)

 

A validated standardized analytical method must have a limited variability when analyzing test portions homogeneous at their scale, to ensure consistency of results and comparability of data.

The variability of leaching and percolation tests is developed further in this page.

A leaching test is a succession of three operations: (i) sample preparation, (ii) leaching of the test portion, and (iii) quantification of elements, substances or parameters in the leachate.

The leaching operation, is from a chemical point of view, a partial extraction. The quantity of extracted substances depends on the test conditions, that must be strictly described and followed, to be reproductible (intra-laboratory) and reproducible (inter-laboratory).

 

Variability of leaching test as a function of leachate concentration? No

The variability of analyses, repeatability (CVr) and reproducibility (CVR) of the validation data detailed in the annexes of the batch leaching test EN 12457-2 are presented in Figure 5 as a function of parameter concentration (data > limit of quantification – LOQ).

 

Figure 3 a, b and c: CVr of analysis of leachate, CVr and CVR of (leaching and analysis of leachate) (validation data of EN-12457-2)

 

The 24-hour batch leaching tests are mainly used for landfill acceptance or allocation of inert waste in the EU. The average intra-laboratory CVr is 0.17 (n=38, different waste, element concentration > LOQ) and the average inter-laboratory CVR is 0.36 (idem).

There is no statistical relationship between the variability and the concentration.

 

The huge variability of accelerated percolation test

The situation is worse for the accelerated (8 hours rather than 24 hours) and with low test portion mass related to grain size less of the accelerated percolation test EN 16637-3. Material is submitted to upward flow of deionized water in a column.

The intra-laboratory CVr and inter-laboratory CVR are presented as a function of concentration, of L/S ratio, and by sample (Figure 6). The statistical analysis of variance indicates no influence of these three factors neither of parameters on CVr and CVR. The validation trials are not conclusive: the very high variability for all the elements (mean CVR = 0.57) and organic substances (mean CVR = 0.73), all waste and all percolation fractions indicate the simultaneous presence of low values and high values during repetitions, so that the results are not normally distributed.

 

 

Figure 4 a, b, c and d: CVr and CVR of the accelerated percolation test of construction materials are high and not function of the concentration of the elements (upper figures), neither of the liquid-to-solid fraction (mid figures), nor of the sample (lower figure – three samples, for each sample left column CVr, right column CVR) (validation data of prEN 16637-3)

There is no statistical relationship between the variability and the samples, the elements, the concentrations, and the percolate fraction (details in Supplementary Information of the cited paper).

 

Conclusion

As the variability of the quantification step (after extraction) is limited in waste and soil analyses to about 1% (intra-laboratory CVr = 0.01 = 1%), the analytical variability stems from three main sources:

• non-homogeneous test portions;
• variable extraction rate, due to presence of options in the method or insufficient time for equilibrium; and
• ill-defined solid/liquid separation (leaching or percolation tests).

Experiments indicates that this last point is critical since there are colloids and nanoparticles in the leachates, representing from 0 to 100% of the element fraction in the leachate. Numerous authors have documented that point.

In the Supplementary Information of Hennebert et al. (2022) (above-mentioned reference) are presented:

• A very limited literature (7 papers) review on presence of particles in leachates. Much other papers could have been cited
• Experimental date with counts of particles < 0.45 µm in leachates (100 million per milliliter), and the fact that, counter-intuitively, the centrifugation before the 450 nm-filtration delivers leachates more concentrated in particles (1 sample) and statistically more concentrated in elements (+13%, 27 samples, 287 paired data) (Hennebert et al. 2022, this page). Without centrifugation, a filter cake can build up on the membrane and act as additional filter for the particles. That point is discussed in detail in the EU leaching test ((annex E of EN 12457 series).

 

Experts in standardization bodies should stop to deny the variability of the tests, and accept strict modus operandi to reduce that variability, avoiding to jeopardise these methods of assessment of the mobile fraction of waste, useful for risk assessment and circular economy.

 

More data in the cited paper and its supplementary information:

Hennebert P, Stoffel A-F, Hubner M, Fortmann D, Merdy P, Beggio G. 2022. The inherent variability of some environmental analytical methods hampers the circular economy of materials. Detritus. 17 p. https://doi.org/10.31025/2611-4135/2022.16225

 

Also useful:

Beggio G, Hennebert P. 2022. A novel method to calculate the size of representative waste samples based on particles size. Detritus, 18, 3–11. https://doi.org/10.31025/2611-4135/2022.15187

Hennebert P, Beggio G. 2021. Sampling and sub-sampling of granular waste: size of a representative sample in terms of number of particles. Detritus 17 – 2021, 1-12. https://doi.org/10.31025/2611-4135/2021.15139

 

Contact: pierre.hennebert@gmail.com

https://www.researchgate.net/profile/Pierre-Hennebert

Consultant in waste management and circular economy. Former convenor of the “leaching” and the “sampling” working group of the European Committee for Standardization CEN TC 444 “Analytical methods for the environmental characterization of soils, waste, bio-waste and sludge”.