Sampling error as a consequence of particle size and sample handling

Decision errors can occur because very small amounts of soil (sometimes as little as 0.25–0.5 g) are actually analyzed from the jar that is sent to the laboratory. Differential contaminant loading of small vs. large soil particles has already been discussed, and further examples are provided in Hyperlink 12. The effect on laboratory subsampling shows up as data variability in the sampling results.

Microscale heterogeneity exerts its effects as soon as soil is placed into a container. The settling of soil that occurs during container movement and sample shipment is governed by particle size and density. Settling stratifies a soil sample such that the larger particles usually end up at the top of the jar as smaller particles work their way to the bottom. If the subsampling procedure involves simply opening the jar and scooping from the top, very few small particles will end up in the analytical subsample, which may bias the concentrations low. On the other hand, the type of scoop used to take the subsample may discriminate against larger particle sizes if the surface is flat or very small so that larger particles can roll off. The very process of weighing out the analytical subsample can select for small particles if they are preferentially tapped onto the balance to slowly bring the subsample up to the desired weight. Laboratories seldom have standard operating procedures (SOPs) for obtaining a representative analytical subsample. Each laboratory, and each technician in the same laboratory, is likely to handle samples somewhat differently. As a result, the analytical subsample may not be representative of the bulk average in the sample container but may over- or underrepresent certain particle sizes from one subsample to the next (Gerlach et al. 2002).

Field duplicates, colocated samples, and laboratory subsampling duplicates data primarily provide information about heterogeneity at different spatial scales and not solely about analytical issues at the lab.

Unfortunately, typical sampling and analysis procedures make little or no effort to control for particle and microscopic effects. In fact, common mixing techniques, such as cone-and-quartering, can even exacerbate the problem (Gerlach et al. 2002).Therefore, it is not surprising that analyses of subsamples repeatedly taken from a single jar of soil can have widely varying results, as reflected in high relative percent difference (RPD) between field or laboratory duplicates. Most of this difference is not due to analytical issues, as is commonly assumed, but is primarily caused by heterogeneity between replicate subsamples.

In summary, though soil sampling seems like a simple process, it is actually quite complex and subject to many kinds of errors. For example, errors occur when the ratio of large to small particles in the subsample do not match the ratio present in the sample container. Taking a representative sample from a heterogeneous bulk particulate material like soil requires careful planning at each stage of sample collection and analysis. Planning to avoid errors requires an understanding of all types of heterogeneity and the spatial scales at which they occur.