What a Field Screen Is Actually For

Ask ten environmental professionals what a field screen is for and most will give a version of the same answer: it tells you roughly how contaminated the soil is, so you know where to send samples. That answer is not wrong. But it quietly imports an assumption that causes more problems in practice than any instrument ever has — the assumption that a field screen is a rough laboratory analysis, and that its job is to approximate a number the laboratory will produce later.

It isn't, and it shouldn't try to be. A field screen exists to support a decision that has to be made while the crew is still on site. Once that is the standard, almost everything about how field data is collected, judged and used has to change.

The question a field screen is actually answering

Stand at the edge of an excavation with the machine idling. The decision in front of you is not "what is the exact benzene concentration in this soil?" It is "do I keep digging?" At a borehole it is "do I go deeper, or step out?" Back at the truck it is "which twelve of these ninety samples go to the laboratory, and what question does each one answer?"

Every one of those is a binary or near-binary decision, and every one has to be made now. A concentration estimate is only useful insofar as it resolves the decision. This is why the useful outputs of a modern field screen are not a single number but three related things:

• Prediction — is this sample most probably above or below the applicable guideline, and by roughly how much?
• Precision — how variable is the soil at this location? Split duplicates and proximity triplicates quantify heterogeneity in real time. High variability says get more data before calling it; low variability says move on.
• Probability — given that measured variability, how likely is it that a repeat reading or a laboratory result would disagree with this call?

Together those three turn a reading into a risk-informed decision. A single number cannot do that, because a single number carries no information about how much to trust it.

A conventional field reading offers false precision — one number implying a certainty it cannot deliver. The goal is calibrated confidence instead: a probable outcome with its uncertainty stated.

Why "it didn't match the lab" is the wrong test

The most common way field screening gets judged — and dismissed — is by comparing a field number to a laboratory number for the same sample and treating any gap as instrument error. That test is misleading, for a reason that has been documented for decades.

U.S. EPA guidance on matrix effects and sample heterogeneity establishes that heterogeneity in the soil matrix has roughly nineteen times more influence on result variability than the choice of analytical method. When a field result and a laboratory result disagree, the most probable explanation is not that either instrument is wrong. It is that the two measurements were made on different subsamples of a genuinely variable material.

The laboratory itself demonstrates this. Laboratories routinely report QA/QC tolerances of ±30–40% on duplicate analysis, and relative percent differences above 50% — sometimes above 100% — are ordinary in heterogeneous soils. Two splits of the same jar, analysed by the same accredited laboratory, can disagree substantially. Holding a field instrument to a standard of concentration equivalency that the laboratory does not meet against itself is not rigour. It is a category error.

The right performance metric is outcome alignment: does the field result predict the same exceedance outcome — above or below guideline — that the laboratory does? That is the question the decision actually rests on, and it is the question a field screen can be held accountable for.

The two errors are not equal

Once outcome alignment is the metric, screening performance decomposes into four possible outcomes, and only one of them is genuinely unacceptable.

Read the two error boxes against each other, because they are not equivalent. A false positive costs one confirmatory laboratory sample and perhaps a small volume of clean soil handled conservatively — it costs money, not safety, and the project recovers. A false negative leaves impact in the ground: a failed confirmation, a return mobilization, and a liability that stays live long after the crew has demobilized. It is not recoverable in the field, because by the time it surfaces the field programme is over.

This asymmetry is why the acceptance ranges used in Triad practice are asymmetric too. A field screen is generally considered decision-ready when it misses fewer than 5% of true exceedances and over-calls fewer than 20% of clean samples. Those two thresholds — FN under 5%, FP under 20% — are the bar. AISCT® programmes are run against a tighter internal target of under 10% on both.

Read that asymmetry carefully, because it also tells you how to interpret a screen that errs. A conservative instrument that over-calls in difficult material is behaving the way you want it to. On a recent coarse-sand chloride programme, AISCT® Sal produced three false positives across 31 laboratory-confirmed samples — roughly 10%, inside the Triad range — and zero false negatives. Every disagreement was the recoverable kind. That is a good result, not a mediocre one.

Precision is what makes a number comparable

There is one more fundamental worth being explicit about, because it is where conventional practice quietly fails. Accuracy — how close a reading is to the true value — is not the property that makes field data useful at scale. Precision is: producing every measurement under the same conditions, so readings are comparable to each other.

A screening programme that is precise but biased still tells you where the site changes, where the boundary is, and which sample is the worst case. A programme that is imprecise tells you nothing at all, no matter how accurate any individual reading happens to be, because you cannot tell a real change in the soil from a change in how the sample was handled. Standardize sample mass, temperature, equilibration time and technique, and the population of readings becomes something you can do statistics on. Leave those uncontrolled and you have a collection of anecdotes.

Screening and the laboratory are not competitors

The last misconception worth retiring is that field screening and laboratory analysis are rivals, with the laboratory as the arbiter of whether screening works. They do different jobs, at different points in the workflow, and each is best at what the other cannot do.

Field screening delivers density and spatial intelligence — hundreds of data points per mobilization, mapped across the site, in minutes rather than weeks. The laboratory delivers analytical precision and regulatory-grade documentation for the specific samples that matter. The screen's real power is that it selects those samples with quantified confidence.

In conventional practice, laboratory samples are chosen from a sparse data set on the basis of appearance, odour and a PID reading. The practitioner cannot know whether the sample submitted is the worst case, the boundary, or something in between. With high-density screening completed before a single jar is shipped, the selection changes character entirely: this location is the highest at 3.2× guideline with low variability; these three sit on the boundary; these fifteen are confidently clean. Twelve samples then answer every question a regulator will ask, and the laboratory bill falls accordingly.

And the relationship runs both ways. Every laboratory result that comes back is compared to the field reading for the same sample, quantifying the site-specific bias and feeding the correlation model. The next programme starts smarter than the last one did. A PID reading has no memory; a modelled field system does.

The One Thing to Remember

A field screen is not a cheap laboratory analysis, and judging it as one guarantees you will under-use it. It is a decision instrument, judged on outcome alignment rather than concentration equivalency, valuable in proportion to its precision and its density, and at its most powerful when it chooses which samples the laboratory should see.