6.2.2.7 Analytical splitting and subsampling techniques

These sample splitting and subsampling options can provide representative aliquots.

After the entire sample has been dried, sieved, or otherwise prepared, a variety of techniques may be employed to complete the incremental subsampling process for target analyte or moisture determination. Of the multiple processes that exist, several are omitted from this section based on their low performance rankings in regard to grouping and segregation and agreement to calculated sampling error as seen in Table 8 of Guidance for Obtaining Representative Laboratory Analytical Subsamples from Particulate Laboratory Samples (Gerlach and Nocerino 2003). The following techniques are reviewed in this section according to the preferential order in the aforementioned Table 8: sectorial sample splitters, paper cone sectorial splitters, simple incremental sampling (1-D or 2-D Japanese slabcake), and riffle splitting. Alternative shoveling, fractional shoveling, and cone and quarter techniques are available but generally not recommended unless sample characteristics prevent the first four techniques from functioning properly or DQOs can be met even with these less rigorous techniques (Gerlach et al. 2002).

With each of these subsampling techniques, consideration should be paid to the potential for contamination. Decontamination processes must be developed and checked using a matrix such as blank Ottawa sand at an established frequency between samples. The composition of the subsampling equipment should also be considered as a potential contamination source. For example, plastic parts containing phthalates should be avoided if SVOC phthalates are contaminants.

An important element to consider when using a subsampling process is that the final subsample mass must be used completely in the analytical sample preparation step. For this reason, the final target mass for each of the following approaches and the mass needed for analytical sample preparation must be considered when choosing the process.

As with all aspects of field sampling, coordination should take place between the laboratory and the end data user to determine which method would be most appropriate. Each of these processes may have different biases. In general, projects requiring a greater amount of reproducibility should be processed with the smallest particle sizes and the largest final target masses acceptable (ASTM 2003).

Of particular concern are methods that use small masses such as the 1 g amount typically used in metals digestions. Increasing the initial mass to a minimum of 10 g at a <2 mm sample particle size improves reproducibility. See the discussion of FE in Hyperlinks 7 and 18 for further details. There are generally only two options to reduce the FE: increase the sample size or reduce the particle size. For a typical soil and analyte concentrations of 1 ppm, to reduce the FE to ≤15%, either the sample mass must be increased to 32 g (2 mm particle size) or the particle size must be reduced to less than 325 mesh (0.044 mm) for a 1 g sample.

The following techniques for splitting and subsampling may or may not be appropriate, depending on project-specific DQOs.

Sectorial sample splitting is the preferred process that results in the least sample heterogeneity of the methods discussed. It requires investment in a rotating sample splitter and dust-abatement measures. The device consists of a rotating head with several chutes sitting on top of a motor. The chutes are spaced equally apart from each other and are of the same dimensions. A hopper is mounted above the rotating head with a vibrating tray that delivers the soil sample to the splitter at a variable rate, depending on the intensity of the vibrations. The rotation speed should also be adjustable. The sample falls from the hopper into the chutes as they spin. Collection devices such as sample bottles are mounted on the bottom of each chute to receive equal portions of sample. In general, slower feed rates from the hopper and faster rotational speed make for better subsamples.

The entire sample must be poured into the hopper initially and the resulting subsamples are equal in mass to the initial sample mass/the number of subsamples. If the desired target mass is not achieved on the first split, recombinations of individual splits may be required to achieve a larger final target mass or resplitting of one of the previous spilt samples (serial splitting) if a smaller mass is needed. Small amounts of fine particles may adhere to the device and should be pushed through the device by tapping or by a small burst of compressed air.

Limitations to this technique include equipment cost and availability, trained staff availability for correct operation, equipment cleaning issues, and equipment maintenance.

Paper cone sectorial splitting achieves a result similar to that of the rotating sectorial splitter and does not require the purchase of expensive equipment, but is far more labor-intensive and more sensitive to operator technique. A square piece of paper is folded in such a way as to have several equally spaced ridges and valleys in a downward conical shape. A funnel is held in one hand and a container holding the sample in the other. The entire sample is poured from the container into the beaker while rotating the funnel around the top of the cone. Individual containers are placed at the base of each valley to receive the sample as it falls.

One-dimensional Japanese slabcake is produced by pouring the sample into a line using at least 20 passes back and forth to distribute the sample particles over the line. A square scoop is cut across the line to remove a subsample aliquot. Combine as many of these aliquots as needed to accomplish the mass reduction (Gerlach and Nocerino 2003).

Two-dimensional Japanese slabcake or incremental sampling is a method that emulates the field incremental subsampling process in the controlled laboratory setting. The entire sample is spread evenly onto a 2-D surface at a depth that can be easily penetrated by a square scoop. A scoop is then taken by removing an increment that equally represents the entire vertical column of the slabcake and the material is placed in a receiving container. This process is repeated at least 30 times at systematic random locations around the entire sample. This technique may introduce more bias than the previous three techniques, as it is impossible to extract an ideal increment (a cylinder or rectangular solid) from a noncohesive soil, even when using a square scoop with vertical sides (the bottom of the slab is underrepresented in the increment).

The laboratory default should be to use 30 increments to build the analytical aliquot. If project-specific planning has determined that other increment numbers are needed to meet DQOs, use them. Replicate subsamples are recommended to determine whether the subsampling meets the DQOs.

A process should be established to document that the increments are collected from random or systematic random locations over the entire exposed surface to ensure adequate representation of the sample. Increments for replicate samples should be collected from independent locations, or alternatively, the entire sample may be stirred, respread, and replicate increments collected in the same manner as the primary sample. Repeat the process for as many replicate samples as applicable.

A good example setup is a 20 × 30 inch aluminum baker’s tray lined appropriately. The tray can easily take a 2 kg sample spread across it at a depth of no more than 1–2.5 cm. A scoopula is used to push the sample around and spread it to an even depth and ideally as thin as practical. As the sample is spread, the fine particles tend to migrate downward toward the tray while the larger, less-dense ones rest on top. A scoop is used that minimizes the discrimination of taking more of the large particles on the top. A square-walled, blunt-end scoop with a minimum 16 mm width tends to perform the best because it facilitates equal collection from both the top and bottom of the slab. The sides reduce the tendency of particles to fall off the scoop during increment collection. Before taking increments, the target mass should be considered. Each scoop (increment) will ideally represent 1/30th of the desired target mass. For a 30 g subsample, each increment should weigh about 1 g. Before starting the scooping process, a few trial scoops should be taken and weighed, to calibrate the amount needed for each scoop. This technique works best when used after disaggregation or milling in conjunction with particle size selection via sieving to reduce the range of particle sizes (see Figure 6-3).



Example of 2-D Japanese slabcake incremental subsampling on dried and sieved soil.

Figure 6-3. Example of 2-D Japanese slabcake incremental subsampling on dried and sieved soil.



The 2-D Japanese slabcake subsampling process may be applied to moist “sticky” samples as well. The best results are achieved with moist sieved soils (see Figure 6-4), but this process can also be applied to as-received samples. Spread the moist soil into an even-depth Japanese slabcake as described above. Use a square-walled, blunt-end scoop with a minimum 16 mm width for 2 mm particle size to collect 30 or more increments to produce the final analytical subsample. Coring tools may also be used for subsampling if the moist sample is sufficiently cohesive. See the tool width discussion in Section 5.2.

Example of 2-D Japanese slabcake incremental subsampling on moist sieved, as-received soil.

Figure 6-4. Example of 2-D Japanese slabcake incremental subsampling on moist sieved, "as-received" soil.



Riffle splitting generally divides the sample into two equal portions by directing the sample portions into opposite pans with alternating chutes. It can be used sequentially to further subdivide a sample into smaller aliquots (Gerlach and Nocerino 2003).

Alternate shoveling divides the sample into two subsamples by placing alternate subsample scoops of the original sample into two separate sample containers (Gerlach and Nocerino 2003).

Fractional shoveling is similar to alternate shoveling except the sample is divided into three or more subsamples (Gerlach and Nocerino 2003).

Cone and quartering splits the sample into two subsamples by pouring the sample into one large cone, flattening the top, and dividing into four sections. Opposite sections of the sample are then combined to form the two subsamples (Gerlach and Nocerino 2003).