Question: how well does this procedure repeat?
The same station, the same logger depths, the same procedure, the same calibration, at the same hour of the day — separated from Dive 2 only by twenty-four hours. If the water has not changed and the method is sound, this occupation should reproduce Dive 2.
For each parameter, Dive 4 establishes an empirical comparison threshold R: the larger of the observed Dive 2-4 difference and that comparison's uncertainty-based threshold C₂₄. These are the comparison thresholds — one per parameter — for the whole procedure. Not for the instruments alone; that was measured in the bath — but of everything: the instruments, the deployment, the depth-holding, the kick cycles, the observers, and the water's own day-to-day variation, all together. Every claim made on Dive 5 must exceed the threshold for its own parameter to mean anything.
Follow the Station Occupation Procedure in full, without modification.
Dive 4 is conducted at approximately the same time of day as Dive 2, and may not be conducted until the written predictions have been recorded and dated.
Specific to this dive
Before entry, read the repeatability prediction aloud, and read the Dive 2 specification aloud alongside it. Every member of the team should know what is being reproduced and what was predicted.
Confirm nothing has changed but the day. Station, logger depths, disc bearings, background sighted against, working depth, dwell time. Read them from the Dive 2 record.
An imperfect Dive 2 is repeated as it was. An invalid one is not. This distinction is not optional and the instructor is responsible for making it.
Imperfect but valid means the procedure was followed and the result is usable, but not as tidy as intended: a logger clipped at 10.2 m instead of the planned 10.0, a disc bearing a few degrees off, a dwell that ran long. Reproduce these exactly. Correcting them between occupations changes the method and destroys the comparison, which is worth more than the tidiness.
Invalid means the data cannot be used or the procedure should not have been conducted that way at all — a safety standard not met, a course standard breached, a logger that was never at depth, a calibration that was skipped, a measurement made outside the agreed separation. These are never repeated. Correct the procedure, occupy the station properly, and treat that occupation as the new baseline. A comparison against an invalid baseline is worthless, and preserving comparability is never a reason to repeat something unsafe or non-compliant.
Where a fault is repeated, record what it was and why it was preserved.
Compare the calibration sets. The Day 1 post-dive offsets and the Day 2 pre-dive offsets are taken twenty-four hours apart on the same instruments. A large discrepancy means an instrument moved, took a knock, or is failing, and must be resolved before the comparison is attempted — otherwise an instrument change will be read as a change in the water.
Additional Analysis
After the standard analysis steps:
- Plot the Dive 4 profiles against the Dive 2 profiles on the same axes, temperature and light.
- For every parameter, compute the difference between the two occupations and its uncertainty-based threshold C₂₄ = 2√(u₂² + u₄²), using each occupation's own uncertainty for that parameter.
- Classify each difference as resolved only where its magnitude exceeds C₂₄.
- Record R for each parameter: the larger of the observed difference and C₂₄. For temperature it is R(T), in degrees; for the attenuation coefficient R(Kd), in reciprocal metres; for extinction distance R(y), in metres. The floor matters — where nothing resolved, R is C₂₄ rather than the observed difference, so a quiet day cannot produce a threshold smaller than the instruments could have seen. They are not interchangeable and there is no single number, because the quantities do not share units — 0.08 °C and 1.2 m cannot be compared, still less ranked.
- State whether the repeatability prediction was supported.
What this threshold is, and what it is not
Be precise with student divers about what has just been measured, because the temptation is to claim more than the data supports.
It is one difference between one pair of occupations. Two numbers do not estimate a spread. A statistician would want several repeat occupations before speaking of repeatability, and this course has time for one.
And it contains more than the procedure. The difference between yesterday and today includes the instruments, the deployment, the depth-holding, the kick cycles, the observers — and whatever the water itself did overnight. Those cannot be separated with the data in hand.
Which is exactly why it is useful. Because it contains the water's own day-to-day variation as well as the method's, it is larger than the method's variation alone. Using it as the bar for Dive 5 therefore errs in the safe direction: it makes a claim about diurnal change harder to make, not easier. A difference that clears this threshold has cleared something conservative.
Call it the comparison threshold rather than the repeatability, and say in the Station Report what it was derived from. A reader can then judge it for themselves.