Everything you make here is aimed at a dive that happens after the course ends.
When you go back to your station in six months, a year, or five years, you repeat Dive 3 exactly: relocate from the documentation, re-occupy, re-capture, re-score. That time the interval is long enough that a difference might be real.
That visit, not this course, is where change is detected. Dive 3 is your rehearsal of it, run while the answer is already known — the reef has not changed over a few days, so every difference you measure is your own method. A diver whose first return visit is a real one cannot tell which part of the difference is theirs.
Comparable means the same place, the same standard, the same method. The station relocated to within the offset your documentation supports. Frames that met the capture standard. Charts down in the same rotation. The same class list and point count. Scoring done the same way. Where any of that differs, you state the difference and narrow the comparison — a return visit is an audit of your own earlier record, and the rules are the ones you just learned for somebody else's.
A difference smaller than the spread your own method produces with nothing changed is not change. Beating it clears a floor — it does not settle the question.
You measured that spread during the course. It has three parts: the sampling error from your point count, your disagreement with another scorer, and your capture repeatability from Dive 3. The last is usually the largest, and it is the one you can improve.
You measured all of it in one day, with one set of charts and one calibration. A real return visit adds what that day did not contain, and your claim widens to hold it. What you leave this course with is a floor, not a detection limit good for every visit that follows.
A null result is a claim too. No change was detected is a statement about the reef only if it arrives with the smallest change you could have seen. Without that figure it is a statement about you.
And change is not health. A second visit can support a claim that something changed and by how much. It cannot support a verdict on the reef's condition, because a verdict needs a reference your record does not contain — a comparable site, a regional standard, a longer history. A reef can decline and still be in better shape than its neighbors; it can improve and still be in poor shape. What you hand over is what changed and by how much, offered to whoever holds the wider evidence. That is worth considerably more than a verdict your data cannot carry.