A Call, or a Probability
148 chloride samples screened across a decommissioning programme working to two different limits. Away from the limit the system returns a call. Close to it, it returns the odds instead — and only the calls are scored.
EXECUTIVE SUMMARY
148 samples were screened for chloride on this programme and 93 of them were also sent to an accredited laboratory — one of the largest checked sets AISCT® Sal™ has on record. A screening result close to the limit is not a decision the system pretends to make: any sample reading within 20% of the applicable criterion is returned as a probability of exceedance for the crew to resolve on site, and carries no pass or fail to grade. That leaves 75 samples with a call against them. On those, the screen caught 13 of the 15 exceedances the laboratory found, cleared 59 of 61 correctly, and raised 1 false alarm.
AT A GLANCE
148 screened • 93 laboratory-checked • 75 carrying a scoreable call
1.THE CHALLENGE
A conductivity meter measures how well soil conducts electricity, which is not the same question as how much salt is in it. Background minerals, moisture and soil texture all move that reading. On a decommissioning programme with two different limits — one for shallow soil and a higher one below three metres — and a schedule set by laboratory turnaround, a field number that cannot be compared directly to the limit is a number nobody can act on.
2.THE APPROACH
Sal measures the chloride itself and reports it in milligrams per kilogram, the same units as the limit, so each result is compared to the applicable criterion the moment it is taken. Away from the limit that comparison is a decision. Within 20% of it the system reports the odds instead, because the spread between two grabs of the same soil is wide enough that a single reading in that band cannot settle the question. The crew screened densely and sent the laboratory the samples carrying the decisions worth confirming.
3. THE DATA
23 of the 148 screened samples landed inside the band and came back as a probability rather than a call — the crew’s cue to re-test, duplicate or step out while still on site. Of the 93 samples the laboratory checked, 18 sit in that band and 75 carry a call that can be graded.
WHERE THIS RESULT APPLIES
- Large programmes where laboratory turnaround sets the schedule
- Salt delineation across shallow and deep criteria on one site
- Deciding which samples are worth a laboratory analysis
- Programmes that have to show how their screen was verified
4.THE VERIFICATION
Because the answer arrives in the units of the limit, the crew could sort soil while it was still in front of them and send the laboratory the samples that carried the real decisions. The samples reading close to the limit did not stall the work either: they came back with odds attached and a clear instruction to resolve them on the spot, which is the band where a laboratory check earns its cost.
5.THE DECISION
Because the answer arrives in the units of the limit, the crew could sort soil while it was still in front of them and send the laboratory the samples that carried the real decisions. The samples reading close to the limit did not stall the work either: they came back with odds attached and a clear instruction to resolve them on the spot, which is the band where a laboratory check earns its cost.
6.THE VALUE
Two things came out of this programme. The first is coverage: 148 samples' worth of delineation for 93 analyses, with most of what the screen flagged actually checked rather than assumed. The second is the scoring discipline. The system grades itself only on the samples it was willing to call, reports the rest as odds, and still publishes a rating held below what the score alone would give it. A screen that will not grade itself on the questions it declined to answer is one you can put your name to.
BASIS & LIMITS — WHAT THIS CASE DOES AND DOES NOT ESTABLISH
- Scoring scope: a sample whose screening result falls within 20% of the applicable criterion is reported as a probability of exceedance and carries no pass or fail to grade. 23 of the 148 screened samples and 18 of the 93 laboratory-checked samples sit in that band. The rule is applied to the screening
- value alone, without reference to the laboratory result.
- The rating is held at Moderate because Catch is 0.87, below the 0.90 floor. The score of 16 of 18 would otherwise band Effective.
- Duplicates: 15 field pairs, median split difference 12%. The screen-to-laboratory median difference was 22%, giving a closeness of 1.89 against the heterogeneity floor — Moderate, not Effective.
- TRIUM personnel were on site for initial training only. Sample locations and laboratory submissions were selected by others, and it cannot be confirmed that the system was used for the whole of the screening programme.
- The extraction method and chemistry are proprietary. Field screening supports, and does not replace, accredited laboratory analysis.
.svg)
