Statistical quantification of sub-sampling representativeness and uncertainty for waste-derived solid recovered fuel (SRF): Comparison with theory of sampling (ToS).

The level of uncertainty during quantification of hazardous elements/properties of waste-derived products is affected by sub-sampling. Understanding sources of variability in sub-sampling can lead to more accurate risk quantification and effective compliance statistics. Here, we investigate a sub-sampling scheme for the characterisation of solid recovered fuel (SRF) - an example of an inherently heterogeneous mixture containing hazardous properties. We used statistically designed experiments (DoE) (nested balanced ANOVA) to quantify uncertainty arising from material properties, sub-sampling plan and analysis. This was compared with the theoretically estimated uncertainty via theory of sampling (ToS). The sub-sampling scheme derives representative analytical results for relatively uniformly dispersed properties (moisture, ash, and calorific content: RSD ≤ 6.1 %). Much higher uncertainty was recorded for the less uniformly dispersed chlorine (Cl) (RSD: 18.2 %), but not considerably affecting SRF classification. The ToS formula overestimates the uncertainty from sub-sampling stages without shredding, possibly due to considering uncertainty being proportional to the cube of particle size (FE ∝ d), which may not always apply e.g. for flat waste fragments. The relative contribution of sub-sampling stages to the overall uncertainty differs by property, contrary to what ToS stipulates. Therefore, the ToS approach needs adaptation for quantitative application in sub-sampling of waste-derived materials.