Jump to main content.


Ecological Risk Assessment

Frequently Asked Questions

Composite Sampling within Ecological Assessments

  1. What is composite sampling?
  2. Where, When, and Why is composite sampling used?
  3. How is composite sampling applied spatially? How is it applied temporally?
  4. What is an acceptable composite sample?
  5. What are adjusted benchmarks, how are they derived, and how are they applied?
  6. What are the benefits of composite sampling?
  7. What are the limitations of composite sampling?
  8. How can we define ecological exposure points (using composite samplings)?
  9. How does the receptor being protected affect the use of composite sampling?

1. What is composite sampling?

Composite sampling is a technique whereby multiple temporally or spatially discrete, media or tissue samples are combined, thoroughly homogenized, and treated as a single sample.

[ Back to Top ]

2. Where, When, and Why is composite sampling used?

Composite sampling can improve spatial or temporal coverage of an area without increasing sample number. As the resulting information on contaminant extremes and variability is substantially reduced compared to discrete sampling, the appropriateness of composite sampling is dependent upon the sampling objectives and the site characteristics. Project managers and risk assessors should consult with the BTAG, a statistician, and possibly other technical specialists when developing a sampling plan that includes composite sampling for the following purposes:

Composites are appropriate for inorganic contaminants and persistent, nonvolatile organic compounds such as PCBs (EPA 1991) in all media and biota under the following conditions (Carson 2001, Correll 2001):

The need for subsequent sampling (i.e., composite exceeding adjusted benchmark - see #5 below) is minimized if:

Specific circumstances and objectives may also warrant composite sampling:

[ Back to Top ]

3. How is composite sampling applied spatially? How is it applied temporally?

Compositing can be performed on spatially distinct component samples if they have equivalent exposure potential. Thus, horizontal compositing should be limited to a particular medium or habitat type over which exposure is expected to be uniform. For example, sediment samples should not be composited across an entire bay if the receptor spends most of its time in shallow areas along the periphery. Likewise, vertical compositing is limited to layers in which exposure potential is uniform. For example, sediment dwelling receptors spend disproportionate amounts of time in the surface sediment layers. Thus, composite sampling should combine discrete samples from within, but not between 0-6, 6-12, and 12-24 inch sediment layers.

Temporal composite sampling can be valuable in assessing average exposure in media if contaminant discharges are expected or known to be randomly distributed (i.e., surface water in a tidal river, air downwind from a discharge point). For example, when substantial time or flow related concentration changes are expected, compositing would likely not be acceptable.

The sampling program must be designed to appropriately account for both spatial and temporal variation in contaminant concentrations and receptor habitat use. Weighted averaging of the composite sample concentrations can then be used to account for known differential spatial or temporal habitat use in deriving an overall exposure point concentration.

[ Back to Top ]

4. What is an acceptable composite sample?

Composite sampling must be designed to ensure that it fulfills the specific sampling objective, site characteristics, and statistical assumptions. For example, the design to derive exposure point concentrations for an ecological receptor may be substantially different from that to determine the success of a soil removal action in the same area. The number of discrete samples forming the composite must consider the expected ability to detect exceedances of the benchmark (see Carson 2001), while minimizing the error rates (see Correll 2001). The composite must be limited to discrete samples from areas where contaminants are expected to be randomly distributed and variability is expected to be low. Areas with known or expected biased distribution of contaminants (e.g., waste trenches, discharge pipes) must be sampled separately with either discrete samples or composite sampling restricted to within the biased area. Habitat type and receptor home range would still dictate whether and how data from unbiased and biased area should be combined to estimate exposure concentrations.

[ Back to Top ]

5. What are adjusted benchmarks, how are they derived, and how are they applied?

With discrete sampling, data are compared to the benchmark appropriate for the particular objective (e.g., soil samples are compared to earthworm toxicity reference value). When discrete samples are composited, concentrations of chemicals within individual samples that exceed the benchmark may be masked within the composite. Therefore, the benchmark is adjusted to indicate if chemical concentrations within any of the individual samples are likely to exceed the benchmark. To perform effectively, the adjusted benchmark must balance both the false positive and false negative error rates to fulfill the specific objective. Statisticians have derived different approaches of varying degrees of statistical complexity for adjusting the benchmark (Boswell et al. 1996, Carson 2001, Correll 2001, US EPA 1991, 1992) and for selecting discrete samples for separate analysis when composites exceed adjusted benchmarks (Patil and Taillie 2001). Each approach has assumptions that markedly influence the adjustment parameters and should be considered in light of the objective. To ensure that the assumptions and methodology are compatible with the sampling objective, consultation with a statistician is advised.

[ Back to Top ]

6. What are the benefits of composite sampling?

Composite sampling can improve spatial or temporal coverage of an area without increasing sample number.

Depending on the scale of sampling and objectives, composite sampling can provide more information about average contaminant concentration over space or time.

With an appropriately adjusted contaminant benchmark, composite sampling can increase the ability to detect hot spots by increasing the number of locations sampled.

Composite sampling can provide more representative estimates of mean concentrations than could be achieved by the same number of discrete samples.

Composite sampling may provide more accurate exposure point concentrations for certain receptors in certain habitats.

For the same size area, composite sampling can reduce sampling cost.

[ Back to Top ]

7. What are the limitations of composite sampling?

Preliminary discrete samples and/or site history details are necessary to demonstrate that the media meets the requirements and to derive adjusted benchmarks.

The resulting information on contaminant extremes and variability in the media or biota is reduced upon compositing.

Composites may mask hot spots if benchmarks are not adjusted to account for the number of component samples per composite and the known/expected variability. For example, to achieve 91% detection probability of hot spots of 3 ppb with a discrete sample criterion of 1 ppb, the adjusted benchmark for 4 discrete samples per composite is 1.05 ppb, while for 8 discrete samples it is 0.8 ppb. Using an adjusted benchmark of 1.05 ppb with 8 discrete samples per composite would lower the detection probability to 64%.

Some of the statistical approaches to deriving the appropriate number of components and adjusted benchmarks are complex and time consuming.

Valid methods for adjusting the benchmarks are limited to single contaminant evaluations. Therefore, addressing mixtures will require multiple adjustments and selection of the most appropriate surrogate contaminant.

Composite sampling can reduce the utility of the data for multiple purposes. Physical averaging of the media to form composites prevents the recombination that is possible with mathematical averaging of different sets of discrete data. For example, composite sampling designed to derive exposure point concentrations for a receptor may not provide sufficient detail for delineating the extent of unacceptable contamination within that receptor's habitat that is necessary to determine the volume of material to be remediated.

[ Back to Top ]

8. How can we define ecological exposure points (using composite samplings)?

Composite samples must be designed to be representative of an area. Once a sample is taken that meets the sampling design requirements it can be treated the same, with respect to exposure point concentration, as the mean of discrete samples from an area. It is possible that a series of composite samples over a small area or in a short time frame would yield a more defined exposure point concentration than fewer discrete samples over a larger area or greater time frame. The resulting ecological exposure point could be compared to a benchmark for chronic exposure. Composite samples provide less information on maximum exposure concentrations, and therefore, they are not appropriate for situations where maximum exposure is used (e.g., assessing potential acute exposures, initial screening of COPCs).

[ Back to Top ]

9. How does the receptor being protected affect the use of composite sampling?

Provided that the sampling plan has been based upon the sampling site characteristics and the potentially exposed receptors, analysis of composite samples can result in better estimates of receptor exposure. Consideration must be given to the relationship between compositing and the receptor's spatial or temporal use of the habitat. If the receptor has equal exposure over an area, then broad composites that cover the area in a representative way might be an appropriate and low cost approach to estimate exposure. Compositing that is suitable for one receptor may be totally inappropriate for another. If receptors with widely different home ranges are being assessed in the same area, then the compositing scheme must be designed to accommodate the receptor with the smallest home range. For example, compositing of sediment samples throughout a marsh would be appropriate to estimate exposure for mobile benthic fish, while compositing for frog tadpoles should be limited to sediment in shallow areas along the margins of the marsh.

[ Back to Top ]

Bibliography

Region 3 | Mid-Atlantic Cleanup | Mid-Atlantic Risk Assessment |EPA Home | EPA Risk Assessment Homepage


Local Navigation


Jump to main content.