Sampling strategy in a phase 2 site investigation

A phase 2 sampling strategy is only as good as the conceptual site model it tests. The investigation exists to confirm or remove the pollutant linkages the preliminary risk assessment identified, so every exploratory hole should be traceable to a linkage, and every determinand on the laboratory schedule should be traceable to a source. An investigation that covers a site on a uniform grid with a standard suite, regardless of what the desk study found, produces a thick report that answers a question nobody asked.

Design it the other way round. Start from the linkages, decide what evidence would confirm or refute each, then work out where to dig and what to test.

Designing from the conceptual site model

Take each linkage in turn and write down what would settle it. A suspected fuel tank needs targeted holes at the tank location and immediately downgradient. Made ground of unknown origin across a former yard needs coverage sufficient to characterise it rather than to find a single hotspot. A controlled waters linkage needs groundwater monitoring, which needs wells, which needs a second visit.

Then set the objectives explicitly in the method statement: characterise the made ground, determine whether hydrocarbons are present beneath the former tank, establish ground gas regime, obtain geotechnical parameters for foundation design. Writing them down disciplines the scope and gives the reviewer something to check the results against.

Where the contaminated land assessment will also support foundation design, agree the combined scope before mobilising. A Geotechnical Assessment sharing the same holes costs a fraction of a separate visit.

Targeted and systematic sampling

Targeted sampling goes where the model says contamination is likely: former tank positions, filled ground, the old workshop floor, drainage runs, areas of staining or unusual vegetation. It is efficient at confirming a known linkage and poor at characterising an area.

Systematic sampling covers the site on a pattern where the distribution is unknown. A staggered or herringbone arrangement gives a better chance of intersecting a hotspot of a given size than a square grid with the same number of points, because a square grid leaves regular gaps. The spacing is a judgement about the size of hotspot that would matter for the end use, and the report should say what that judgement was.

Most investigations use both, and the report should distinguish them. Averaging a targeted hotspot result into a site-wide mean is a classic error, because targeted points are not a random sample.

Depths, horizons and made ground

Sample by horizon rather than by depth interval alone. Take a sample at the surface for direct contact and produce uptake where gardens are proposed, at each change in material, at the base of made ground where it meets natural strata, and within the natural strata where migration is plausible.

Record the made ground properly: thickness, composition, presence of ash, clinker, brick, timber, plasterboard and any fibrous material, and whether it is putrescible, since that governs the gas regime. Log odours and staining at the time rather than reconstructing them from photographs afterwards.

Trial pits give you the best view of made ground and the worst access in constrained gardens. Window sampling gives you continuous recovery in narrow spaces. Cable percussion or rotary methods follow where depth or groundwater require it. Choose by what the model needs rather than by what the driller has on the lorry.

Ground gas and vapour monitoring

Ground gas is the item that determines the programme, because it cannot be compressed. Monitoring wells are installed, allowed to stabilise, and then visited repeatedly, with each visit recording methane, carbon dioxide, oxygen, flow rate, borehole pressure and atmospheric pressure.

Published guidance sets expectations for the number of visits, commonly at least six spread over a period of months for a lower risk site and considerably more where the risk is higher, and it expects the monitoring to include occasions of low or falling atmospheric pressure, when emissions are highest. A set of visits taken in a single stable high pressure week is not a defensible dataset, however many of them there are.

The output is a gas screening value combining concentration with flow rate, which places the site in a characteristic situation and drives the protection measures designed under the relevant British Standard. Our note on ground gas assessment goes into how those measures are specified and verified.

Vapour monitoring follows a similar logic where volatile organic compounds are suspected, and the sampling method has to suit the compounds involved.

Laboratory schedules and detection limits

Two rules govern the schedule. Test for what the model identifies, plus a screening suite broad enough to catch what the model may have missed. And check that the laboratory's reporting limits sit below the assessment criteria you intend to use, because a result reported as less than a value higher than the criterion tells you nothing.

Use a laboratory with the appropriate accreditation for each determinand, and be aware that accreditation is granted by matrix and method rather than across the board. Asbestos in soil deserves specific attention: screening identifies presence, quantification is a separate and more expensive test, and the assessment approach differs from that used for other contaminants.

Where material may leave the site, waste acceptance criteria testing is a different exercise again, done for disposal classification rather than risk assessment. Budget it separately and do not expect the risk assessment suite to answer it.

Assessment criteria and their derivation

Generic assessment criteria are derived for defined land uses and standard exposure assumptions, so the first check is whether the published criterion matches the proposed use. Residential with gardens, residential without gardens, allotments and commercial all carry different values, and using a commercial criterion on a scheme with private gardens is the error that gets reports rejected.

Where no generic criterion exists, or where the site conditions differ materially from the assumptions behind one, site-specific assessment criteria can be derived using an appropriate exposure model. Say which model, which parameters were changed and why. A derivation that quietly adjusts soil organic matter content to produce a pass is the thing a reviewer looks for first.

For controlled waters, the criteria come from water quality standards rather than human health models, and the assessment has to consider the pathway to the receptor rather than concentrations in isolation.

Reporting so the remediation strategy follows

End the report by returning to the model. Restate each linkage, say whether the investigation confirmed or removed it, and revise the conceptual site model accordingly with an updated figure. Where a linkage is confirmed, state the extent as far as the data allows and be explicit about the uncertainty.

That structure makes the remediation strategy almost write itself, because each confirmed linkage points to a measure: a clean cover system for direct contact, a gas membrane for ingress, removal or treatment for a hotspot, a barrier or monitoring for controlled waters. A report that lists exceedances without reconnecting them to linkages leaves the next consultant to redo the interpretation.

Include the raw data, the logs, the calibration records and the chain of custody documentation. A reviewer checking someone else's work reads those before the conclusions.

Browse the 66 report types to see how the ground investigation evidence sits alongside the rest of a submission. PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval. For specialist technical reports, qualified professional input may be required.

This is general guidance, not legal advice; check the position with your local planning authority or a qualified professional before relying on it.