As described in Table 3.2, the design of the site investigation must be clearly linked to the preliminary conceptual model of the site, and procedures must be in place to allow regular and systematic review of the strategy. A key aspect of the site investigation plan is the design of an appropriate sampling strategy, e.g., decision about sampling locations, patterns and frequencies to meet the objectives of the site investigation.

Where remediation is the probable outcome of the site characterisation, it is essential that the survey design is suitable to allow waste volumes to be predicted. In particular, in the case of radioactively contaminated land, disposal costs (per unit volume) for exempt wastes are considerably lower than for low-level radioactive wastes (LLW). Over-estimation of LLW based on poor data results in high and unrealistic project budgets. Conversely, under-estimation of LLW has the opposite effect.

A summary of the key guidance issues is given below, while extensive guidance is given in the next sections.

There are two approaches to soil sampling:

- Targeted or judgmental sampling, which focuses on known or suspected sources of contamination, such as storage tanks, disposal pits and pipelines. The results from non-intrusive surveys (such as geophysical surveys, radiological surveys and drains surveys: see Section 3.6.3) are used to support the design of the targeted sampling.
- Non-targeted sampling, which aims to characterise the contamination status of an area or volume of ground.

In the case of targeted sampling of a known area of significant contamination, for example where significant levels of radioactivity have been detected by a non-intrusive radiological survey, it may not be necessary to characterise the area in detail during the early stages of the investigation. Instead, it may be more valuable to characterise the surrounding area in order to define the “envelope” of contamination and to provide information on the area of land that may require remediation. More detailed characterisation of the most contaminated areas will be required to define a remediation and waste management strategy, and it may be appropriate to undertake this as a supplementary investigation.

Two approaches to designing non-targeted sampling grids are presented in the existing guidance.

- In BS10175:2001 for exploratory investigations [BSI-2001]; Typical densities of sampling grids can vary from 50 m to 100 m centres and for main investigations from 20 m to 25 m centres. A greater density of sampling grid may be considered appropriate where heterogeneous contamination is indicated, for example, on a former gasworks site where in localized areas 10 m centres may be necessary. A high density sampling grid may also be necessary where a high level of confidence is required for the outcome of a risk assessment (for example for a housing development.
- In contrast MARSSIM presents a statistical approach in which the number of sampling points required to detect a certain size of “hot spot” with a certain level of confidence can be calculated
^{2}. Given this frequency of sampling, it is possible to state that, at the level of confidence specified:- A hot spot of specific size (if one exists) will not be missed.
- If contamination is not found, a hot spot of at least the specified size does not exist.

The size of the hot spot can be considered in a number of ways. It may be the expected size of the contaminated area or the maximum size of contamination that could be economically and safely remediated. Further, it could be the size of an area of contamination in an otherwise uncontaminated site or an area of greater contamination (for example, above some guideline “trigger” concentration) within a site that is generally contaminated.

The two approaches to designing the sampling grid take into account the same broad issues:

- The need for more frequent sampling to provide higher levels of confidence, and
- To characterise areas with smaller contaminant sources.

The statistical approach applied in EURSSEM is the more rigorous approach. However, the information needed to parameterise the model can only be obtained from a conceptual model of the site. The statistical approach proposed also addresses the identification of an appropriate sampling pattern (e.g., square grid, random, herringbone), stating that the “efficient sampling pattern should satisfy four conditions”.

- It should be stratified (i.e., the area to be sampled should be partitioned into regular sub-areas).
- Each sub-area should carry only one sampling point.
- It should be systematic.
- Sampling points should not be aligned.

A square grid pattern satisfies 1-3 above but, because sampling points are aligned, reduces the ability to detect elongated hot spots aligned parallel with the grid. A herringbone pattern is considered to be the optimum type of non-targeted grid pattern. In practice, on operational sites there will be restrictions on the possible positions of sampling points due to the presence of underground services, buildings, etc. This aspect is discussed further in Section 3.7.2. The consequence will be that the actual non-targeted sampling grid will probably not conform to the ideal pattern. A judgement then has to be made as to whether deviations from the ideal grid geometry are so great as to render the statistical measures of confidence invalid.

.

Summarizing: Determination of sampling frequencies:

- The survey grid should be designed to take into account:
- The proposed measurement technique.
- The size of the area to be surveyed.
- The anticipated size of anomalies that may be present.
- Desk study information on potential sources of contamination in the area.

- The majority of radiological surveys are carried out using a grid of some type. The scale of the grid should be selected to ensure that it is unlikely that features of interest will be missed, but should be compatible with the proposed survey instrumentation and with the scale of the overall survey area. The scale of the grid may vary over the site of interest, to allow for focused surveying in the areas of most interest. The statistical design of surveys is discussed below for the following situations:
- Contaminant present in background – determining numbers of data points for statistical tests;
- Contaminant not present in background – determining numbers of data points for statistical tests;
- Determining data points for small areas of elevated activity.

– by Rafael Garcia-Bermejo Fernandez about 6 years ago