Learning Objectives
By the end of this knowledge development session, student divers will be able to:
- Describe the sequence collect, analyze, understand, predict, duplicate, and identify which dive in this course corresponds to each stage.
- Explain why the same station is occupied three times rather than twice, and what the third occupation makes possible.
- Distinguish u, the uncertainty of one value; C, the uncertainty-based comparison threshold; R, the empirical threshold from Dive 2 against Dive 4; and max(R, C₄₅), the threshold a Dive 5 change claim must clear, and explain why one pair of occupations does not estimate the repeatability of the whole procedure, and state which applies to which comparison.
- Explain why a prediction must be written down and dated before the measurement that tests it.
- Name two mechanisms that change the water over a reef within a single day, and state what each would do to the temperature profile.
- State the two questions that must be answered about any observed difference before it can be called a change.
- Explain why a prediction that fails is a useful result.
- Explain why re-occupying a station is not the same as repeating a dive.
Presentation Notes
a. Measurement is only the first step of a cycle. Collect the data. Analyze it. Understand what it means. Predict what will happen next. Then go and see whether you were right — and if you were, do it again somewhere else and find out whether the understanding travels. Most citizen science stops at the first step, and produces archives that nobody can draw a conclusion from. This course runs the whole cycle across two station days.
b. Dive 2 collects. Dive 3 asks whether the picture is complete. Dive 4 asks how well the whole procedure repeats. Dive 5 tests a prediction about change. Dive 6 asks whether the relationships hold in different water. The dives are not six repetitions of one procedure. Each answers a question the previous one raised, which is why the sequence must stay intact.
c. The step most programmes skip is Dive 4. It is tempting to occupy a station in the morning, occupy it again in the afternoon, find a difference larger than the instruments' own uncertainty, and call it a diurnal change. That reasoning has a hole in it: nobody has established how far two occupations of the same station differ when nothing is expected to have changed.
The bath measures how much the instruments disagree. It says nothing about how much the procedure varies — the deployment, the depth- holding, the kick cycles, the observers, and the water's own day-to-day wandering, all of it together. Dive 4 measures exactly that: same station, same hour, one day later. That difference — or C₂₄, whichever is larger — is the floor that every claim on Dive 5 must clear.
This is why the course occupies one station three times rather than twice. The third occupation is not redundancy. It is what turns the second into evidence.
d. The predictions are written and dated before Dive 4. This is the rule that makes the exercise honest, and it is not a formality. A prediction recorded afterward is not a prediction. Human memory is generous to itself: shown a result, we reliably recall having expected it. Writing the claim down, with the reasoning and the time, removes that possibility. Anyone reading the Station Report can see what was expected and what happened, in that order.
Two predictions are required, because Day 2 makes two comparisons. One says what the repeat occupation will show — usually that nothing will be resolvable. The other says what the afternoon occupation will show, and by what mechanism. The first is the one teams find strange, and the one worth insisting on: predicting that nothing will happen feels like a wasted form until the day it does happen.
e. A prediction must be capable of being wrong. "The water will change" is not a prediction. "The temperature at the shallowest logger will be higher this afternoon than this morning, because it is calm and sunny and the surface layer is heating, and the deepest logger will be unchanged" is a prediction. It names a direction, a location, a mechanism, and something that would falsify it. Before Dive 4, each team should be able to say what result would show them to be wrong.
f. Two mechanisms are usually available on a diving day. The first is diurnal heating: sunlight warms the surface layer through the morning, building or strengthening a thermal gradient, which mixing then works against. In calm sunny conditions the shallow water warms measurably between morning and afternoon while deeper water lags or does not move at all. The second is tidal exchange: a flooding tide brings offshore water over the reef and an ebb draws water off it, and the two can differ in temperature and clarity. Which mechanism dominates depends on the site, and choosing the right one is part of the prediction.
g. Remember the confound from Topic 4. Warm water draining off a reef flat on a falling tide arrives at the fore-reef as a warm surface layer, and looks exactly like local surface heating. A prediction that names diurnal heating and is then confirmed by a warm shallow logger may have been confirmed by the wrong mechanism. This is why tide state is recorded, and why a good prediction says what else could produce the same result.
h. Other things change on the same timescale. Wind rising through the afternoon can stir a thermal layer away as fast as the sun builds it. Swell increasing can lift sediment and change the optical measurements without touching temperature. Cloud can arrest surface heating entirely. A good prediction states which mechanism is expected to dominate and under what conditions it would be overwhelmed.
i. Re-occupation is not repetition. Dives 4 and 5 return to the same station, use the same logger depths, the same procedure, the same calibration. Everything is held constant except the time. That is what makes any difference attributable to the passage of time rather than to a change in method. A dive that used different depths, or a different disc orientation, or an uncalibrated array, would produce a difference that could not be interpreted at all.
j. Two questions must be answered about any difference, in this order. First: is it larger than the applicable threshold? And note that the threshold means two different things at two stages. Comparing Dive 4 against Dive 2, the floor is C₂₄, computed from the two days' uncertainties. Comparing Dive 5 against Dive 4, the threshold is max( R, C₄₅ ) — the larger of the empirical threshold Dive 4 produced and this comparison's own uncertainty-based threshold. Second, only if the difference passes the first: is it consistent with the predicted mechanism, in direction, in magnitude, and in where it appeared?
k. The second question is where most of the thinking happens. Suppose the shallow logger reads warmer in the afternoon and the deep logger does not. That is the diurnal heating signature, and the prediction is supported. Suppose instead that every logger warms by the same amount. That is not diurnal heating — heating is applied at the surface and should appear there first. Something else moved the water, and a uniform warming across the whole column suggests that different water arrived rather than that this water was heated. The pattern carries the information, not the size of the change.
l. A failed prediction is a real result. If the water did not do what you expected, one of three things is true: the mechanism you named was not the one operating, the mechanism was operating but something suppressed it, or your measurement was not good enough to see it. Working out which is more instructive than being right, and a Station Report that says so is more valuable than one which quietly rewrites the prediction to match the outcome. Do not do that. The written record exists specifically to make it impossible.
m. Not everything can be attributed, and saying so is correct. Some differences exceed the noise floor and still cannot be assigned a cause, because two mechanisms were available and the data cannot separate them. "Resolved but not attributed" is a legitimate entry in a Station Report. It is more honest than a confident guess, and it tells a future reader exactly what an additional measurement would have to do.
n. Dive 6 asks the last question in the cycle. Understanding that only works at one site on one day is not understanding. Occupying a different station — or the same one under distinctly different conditions — tests whether the relationships you observed travel. The optical product from Topic 7 is the natural thing to test: if it holds at two different sites, you have evidence that the relationship extends beyond the first station, at the resolution of this course. If it does not, you have found something about the difference between those two sites. Either outcome is a result.
o. This is what separates a measurement from a monitoring record. A single occupation describes one moment. Two occupations with a prediction in between describe a process, and a process can be tested, argued about and built on. The instruments in this course are inexpensive and modest. The method is not.