Virridy Home | Lume — Water Quality Sensing Water for Carbon

In-Situ AquaTroll sondes against the Lume sensors

One barrel, 1–2 September 2026, eighteen settled windows through a five-analyte challenge. Three AquaTroll sondes and 36 Lume units read the same water over the same windows. Four channels have an instrument on both sides: temperature, turbidity, FDOM and chlorophyll-a.

Neither instrument is calibrated, so nothing here is read in absolute terms. The Lume fluorescence readings are temperature corrected, normalized to 20 °C on each sensor's own quench fitted inside the settled windows; the sonde readings are not, because this run cannot resolve a coefficient at their reporting rate. Everything is then expressed against its own clean-water window, so both sides are dimensionless and share one axis. The fits are on magnitude: for the fluorescence channels the log–log slope, which an unknown gain on either instrument cannot touch, since multiplying an axis by a constant moves the intercept and leaves the slope alone; for temperature and turbidity, changes from the clean-water window, so an offset cancels. A gain error still rides on those two slopes and is not corrected for. Intercepts and absolute agreement are carried in the tables and are not interpreted.

SummaryDo they agree, do they respond, do they discriminate

Every figure is computed over the ladder each channel is answerable for, on readings referred to 20 °C on the Lume side and to each instrument's own clean-water window on both. Slopes are in logs for the two fluorescence channels, where an unknown gain moves the intercept and not the slope.

Do they respond

Do they discriminate

SensitivityHow small a dose each instrument sees

The same test on both sides. For each channel, the lowest rung of its own ladder whose change clears three times the scatter of repeat readings in that ladder's rung-0 window, which is the water the ladder started from. Nothing about calibration enters, because both are being asked only whether they can tell the water changed.

Between buildsCorrelation, specificity and sensitivity

The comparison against the sondes says how the two instruments relate. This says how the Lume builds relate to each other and to the analytes: how closely units of one build agree, how much each build separates its own analyte from the others, and how small a dose it takes to see.

Response of each build to each ladder

Median fold change from each ladder's rung 0 to its top, over the units of that build, on readings normalized to 20 °C. The cell for a build's own analyte is marked. A value below one is a fall.

Agreement, specificity and detection

All of itEvery step, every channel

The whole run in one place: all eighteen settled windows in time order, all four channels that have an instrument on both sides. Every series is divided by its own clean-water window, so counts, RFU, NTU and degrees share one dimensionless axis and the two instruments can be laid over each other. Shaded blocks are the five ladders. Each channel is then taken up on its own below, over the steps that actually test it.

Sondes in solid lines on the right-hand axis, Lume units dashed on the left. Both are fold change from clean water on log scales, so a flat line is a channel that did not move, and the two axes let each side use its own range instead of one crushing the other. The Lume fluorescence readings are temperature corrected before normalizing; the turbidity panel is the Lume time-of-flight return and the temperature panel its board sensor, neither of which is corrected.

The barWhat agreement is available: sonde against sonde

Three instruments of the same model in the same water set the ceiling on any comparison against them. Before asking how well a Lume tracks a sonde, this is how well the sondes track each other.

Each row is the window in which that channel was driven hardest, so the ratio is between real readings and not between three numbers near zero. The absolute gap is the largest difference between any two sondes at any window in the run.

ScreenWhich instruments are tracking the water

Every instrument checked on the ladder its channel is answerable for, and set aside only where the reading contradicts the water rather than merely fitting loosely.

Fold is from clean water to the top of that channel's own ladder: quinine for FDOM, the turbidity ladder for the time-of-flight return, rhodamine for chlorophyll. The z score is against the median and scaled median absolute deviation of the other instruments on the same channel, in logs. A unit is only set aside for a reason stated in words; a large z on its own is not one.

Channel 1Temperature

The control. Both instruments measure it directly and neither depends on a fluorophore, so agreement here says the two are in the same water and the windows line up.

Through the eighteen windows, sondes on the right-hand axis and Lume board sensors on the left, both in degrees.
Change from the clean-water window on both axes, so any offset on either instrument cancels. Both carry the same unit here, so the axes run over the same range at 1:1 and the pale line is identity.

Channel 2Turbidity

The Lume time-of-flight return against the sonde turbidity reading. The turbidity ladder is the only part of the run in which anything turbid was added, so it is the only part that can test this channel.

Through the eighteen windows, every series against its own clean-water reading, so both instruments share one dimensionless axis.
Change from the clean-water window on both axes. Points are colored by which ladder the window belongs to.

Channel 3FDOM

The Lume FDOM units against the sonde FDOM channel. Quinine sulfate is the analyte this channel exists to answer to, so the quinine ladder is the part of the run that tests it.

Through the eighteen windows, every series against its own clean-water reading, so both instruments share one dimensionless axis.
Log–log, so the slope survives the sonde's unknown gain. Both axes span the same number of decades on a square plot area, so a slope of one would sit at 45°.
Fold change from the clean-water window, temperature corrected on the Lume side. The left column is the ladder this channel exists for; the right is a ladder it should have ignored.
The pooled row is over all eighteen windows and is the weakest of these numbers, because it mixes the quinine response with the rhodamine cross-talk.

Channel 4TLF against the sonde FDOM channel

The Lume TLF units against the sonde's FDOM channel. The sonde has no tryptophan channel, so this pairs two UV-excited channels rather than two measurements of the same thing, and the tryptophan ladder is the part of the run that drives both.

Through the tryptophan ladder, Lume readings normalized to 20 °C and every series against its own clean-water reading, sondes on the right-hand axis.
Log–log on a square plot area with equal decades on both axes, so a slope of one would sit at 45°.
Fold change from the clean-water window. The left column is the ladder this channel exists for; the right is one it should have ignored.

Channel 5Chlorophyll-a: 500151 against 1033548

500151 is used as the example for this channel, paired against 1033548 throughout. The rest of the batch is covered on the raw chlorophyll page.

Through the eighteen windows, Lume readings normalized to 20 °C and every series against its own clean-water reading, so both instruments share one dimensionless axis. Gaps are windows in which 500151 produced no lit sweep.
Log–log, both axes spanning the same number of decades on a square plot area and anchored at the clean-water reading, so the ladder climbs from the bottom left corner and a slope of one would sit at 45°. Points are colored by ladder, so the two instruments can be checked for agreeing on the same analyte as well as on the same water.
Fold change from the clean-water window, temperature corrected on the Lume side.

LimitsWhat this comparison cannot say