Experimental Methodology

Scope

This document sets out how experiments on this aquarium are run. It is written as a procedure so that a reader can judge the results elsewhere on this site, and so that a future survey can be run the same way as a past one.

How variables are controlled

1. 01Change one variable per test. Where two changes are unavoidable, the test is recorded as confounded and its conclusion weakened accordingly.
2. 02Hold the automated inputs constant: photoperiod, CO₂ target, dosing schedule and flow profile are all under automation and are not adjusted during a test.
3. 03Record the state of the aquarium continuously so that an unrelated change during the test window is visible afterwards rather than invisible.
4. 04Run the measurement at the same point in the photoperiod each time.

What runs the tank

How instruments are calibrated

• /The pH probe is calibrated against standard buffer solutions. Because CO₂ is controlled from a pH drop rather than an absolute value, day-to-day stability is more important here than absolute accuracy.
• /The conductivity probe is calibrated with a K0.1 solution set, tracked as an inventory consumable so the solution's age is known.
• /The quantum meter carries its factory calibration. No independent reference has been used, and no accuracy claim is made beyond the manufacturer's.
• /Where an instrument's calibration state is unknown, that is recorded rather than assumed.

Milwaukee pH probe · Atlas EZO conductivity

How PAR measurements are taken

1. 01Fixture at 100% output, mid-photoperiod, room lighting minimized, water clear.
2. 02Sensor submerged and held level at the numbered position.
3. 03Reading allowed to settle before recording.
4. 04Value logged against the position number, with depth and fixture height recorded alongside it.
5. 05The full set of positions completed in one session, in the same order each time.

PAR Mapping Across a 90P

How photographs are standardized

Comparison photographs are taken at matched camera settings: the same exposure, white balance and position. An uncontrolled photograph can be made to support almost any claim about color.

This is the stated procedure for comparison work. It has not yet been exercised on a published comparison, because no second fixture has been tested. The hourly camera archive is captured under fixed conditions but is not a controlled photographic comparison and is not used as one.

How comparisons are performed

1. 01The same aquarium, in the same state, for both items being compared.
2. 02The same instrument and the same numbered positions.
3. 03Both items measured at a matched mounting height, and again at the height each is intended to run at, where those differ.
4. 04Measurements repeated rather than taken once, so run-to-run variation is visible.
5. 05The procedure published alongside the result, in enough detail to be repeated.

Experiment 004 · Solar RGB II

How conclusions are written

• /The question is stated before the result, and the hypothesis is recorded before measuring so that it can be shown to have been wrong.
• /Observations are separated from interpretation. What was measured is stated first; what it might mean is stated second and labelled as such.
• /A result that contradicts the hypothesis is reported as the result, not reframed.
• /Where the data does not support a conclusion, the conclusion is that the test was inconclusive.
• /Language is kept to what was observed. Claims of proof are not made from a single aquarium.

How limitations are identified

Every write-up carries a limitations section, and it is written before the conclusion rather than after. The standing questions asked of each test are: what was not controlled, what was not recorded, how many times was this measured, and what would a skeptical reader challenge first.

Limitations that emerge only after publication are added to the document rather than left out.

How future experiments are planned

Planned work is published before it is run, with its question and hypothesis stated, so that the record shows what was intended rather than only what succeeded. Planned entries carry no invented figures and are clearly marked as not yet run.

Project roadmap