What settles a dose, and what stops one.
Which instruments settle which question, what gets written down before a vessel moves, what halts a pass, and who else is holding the record while it happens.
The layers
Four sources of truth, and they answer different questions.
The common failure in this category is letting the most impressive layer answer a question it cannot reach.
Earth observation
Sets the search area and the priority order across a zone, days before a shore sample would notice anything. It cannot see under the surface, cannot name a species and cannot set a dose. HABscan sets out where these products bend and why.
Moored instruments
Continuity at a fixed point. A buoy string is the only thing that catches the overnight crash nobody was on the water for, and the only record that shows whether a treated volume held its gains a week later.
Profiles at the working site
Dissolved oxygen, temperature, salinity, pH and turbidity through depth, before the pass and after it, at the same depths. This is the layer that settles which depth to draw from and what the reactor runs. Nothing above it can substitute.
Samples and assay
Cell counts, toxin assay and the discharge parameters a probe cannot read, sent to an independent laboratory. Slow, and the only thing that answers what a bloom was carrying. A falling count is a weak proxy and we do not treat it as the verification step.
On the day
The standard is written down before the vessel moves.
The sequence matters less than the order of the paperwork. What gets fixed in advance, and what stays adjustable, is the whole difference between a controlled operation and an improvised one.
Fixed before departure
What the effluent has to meet at the outlet, agreed with whoever consents the work and set against a measured background for that site. It is written as a discharge limit, not as a description of what the plant aboard could produce on a good day.
Decided on arrival
Which depth to draw from, how many passes, transit order across the tasking. All of it from the profile taken on site, because a water column three days old is a guess with a timestamp on it.
Never adjusted at sea
The standard. If the water turns out to need more than it allows, the answer is a new plan agreed ashore with the oversight party, not a crew deciding to push it. An inspector reconstructs this from the log, and the log has to survive that.
The organic background is the part that catches people out. Oxidant is consumed by whatever is already dissolved before it reaches the material you were aiming at, and two bays with identical cell counts can need very different plans for that reason alone. It is measured, not assumed.
Stop conditions
Any one condition halts a pass, and so does any one voice.
The set is defined in writing before the first pass is ever run, which is the only time everybody involved is calm enough to write it properly. Nothing in it is ranked. A single item stops the work.
Among them: a discharge assay that comes back outside the standard; dissolved oxygen at the outlet or in the water around it moving the wrong way; a large animal at the intake bar-rack; an intake drifting into a mapped exclusion area; distressed or dead marine life near the vessel; and two instruments that ought to agree failing to, with nobody yet able to say which is lying.
Part of that does not wait for a person. The inline instruments hold interlocks and cut the oxidant themselves — the fastest stop on the boat. The treatment operator can also stop it, and so can the master, the duty analyst ashore and the oversight scientists, individually, without first persuading anybody else. Restarting takes agreement from all of them.
That asymmetry is not politeness. Stopping is cheap and reversible; continuing through an unexplained reading is neither.
Thresholds
The numbers a duty engineer actually needs.
Three values, each taken from the body that published it, and each carrying the qualifier the source attaches. These are the numbers that end up inside an operating procedure, so they are worth reading at full length before they get there.
The value an alarm must never be set from
A figure of 0.5 to 1.0 mg/L circulates widely as an approximate lethal dissolved-oxygen threshold for fish, and it traces back to the same Delaware River Basin Commission review quoted above. It is a nineteen-hour result for juvenile White Perch alone, sitting in an appendix table. The review's own conclusion is the 4 to 6.4 mg/L band.
An operator who set a night alarm from the appendix figure would be an order of magnitude low, and would learn about it from a fish kill rather than from a report. Alarm levels on a program are set against the review's conclusion, then against what the species and the season in that water actually demand.
Routing
Hull-hours are the scarce thing, and routing is how they get spent.
A zone rarely has one problem at one time. It has four, ranked badly.
Every morning the same question comes back: which water, in what order, for how long. Forecast risk across the zone, the state of anything treated in the last fortnight, tide and wind for the next thirty-six hours, gas remaining, and where the crew sits in its rotation.
The output is a tasking, and the discipline is refusing to write it from yesterday's complaint. A patch that has already closed a beach is the loudest item on the list and often the worst use of a hull. The forming patch in the next bay is quieter, cheaper to work, and the one that decides whether next week has an incident in it — which is what the adjacent Deep Strike in the base subscription is for.
Data services
What a program produces, besides treated water.
A running program generates evidence whether anyone plans for it or not. Designing that record in advance is cheaper than assembling it afterward from memory, and it is the only version that will hold up in the year something goes badly.
The pass record
Position and time, the depth drawn from and returned to, the throughput sustained, the duty the reactor ran, the readings taken on the influent and the effluent, and the profile either side. Two of those columns hold the same position, because treated water goes back into the water it came out of. One row per pass, written as it happens. This is the artifact a regulator will ask for, so it is built for that reader rather than for a newsletter.
The water-quality series
Continuous telemetry from moored instruments and vessel profiles, unbroken through the year, at fixed stations as well as at treatment sites. Untreated control locations included, because a change without a comparison is an anecdote.
The zone risk surface
Forecast conditions across the contracted area for the coming days, assembled from earth observation, in-situ data and meteorology. It drives tasking. It is also handed over, because a client who can see next week is a client who can plan around it.
The program account
What was treated, where, when, and what the water did — set against the parameters agreed at the outset, including the ones that were missed. It reports on the quiet months too, since those are the ones the argument for working them has to be proved in. The independent laboratory results sit inside it under their own dates, because a confirmatory assay is only worth anything if somebody can see when it was taken.
The second copy sits outside this company
Independent oversight takes the form of access rather than assurance. Local scientific institutions engaged for the life of the subscription hold the same live feeds the duty crew steers by — not a curated quarterly digest — and are free to moor instruments we have never touched. They start before the first pass, stay for as long as the program does, and keep their copy afterward.
What that changes
Nothing, on a good day. On a bad one it changes everything: the pass that missed, the patch that came back overnight, the assay nobody expected. Those readings land on two desks at the same moment.
If a program's targets and the health of the water ever pull against each other, a ministry hears it from the institution rather than from us. The documents behind a program are listed in one place.
Where the decision stays with a person
Three things the control layer leaves to somebody who signs.
Instruments narrow a question. They do not close it.
- It does not set its own limits. Interlocks can cut the oxidant without asking anybody, because stopping fast is what a machine is for. Deciding what the effluent must meet, and signing it, stays with a named person.
- It does not forecast a bloom into existence with any useful precision more than a few days out. Risk surfaces narrow a search and rank a tasking. They do not tell you what August holds in March.
- It does not replace your own monitoring obligations. Whatever a regulator requires of you, you still owe. Our feeds are additional to that record, and available to whoever you are obliged to report to.
Asked by data teams and duty engineers
Four about the instruments
What dissolved-oxygen number should an alarm be set at?
Higher than the number that defines hypoxia. NOAA's 2 mg/L is a classification threshold, not a safe operating level, and treating it as one is how a stocked pen gets lost overnight.
A Delaware River Basin Commission literature review puts lethal-level requirements for sensitive estuarine species at 4 to 6.4 mg/L, with roughly 5 adequate for many species and 6 to 7 protective. Set against that band, then against what your own species and season demand.
Who can stop a pass?
Any of four people, acting alone: the treatment operator, the master, the duty analyst ashore, or an oversight scientist. None of them has to convince the others first.
Faster than all of them are the interlocks, which cut the oxidant on their own when a reading crosses the line. Restarting, though, requires all four. Making it easy to stop and difficult to resume is the cheapest safety control available to a program like this.
Does the oversight body get the raw feeds or a report?
The feeds. Live, as the crew sees them, plus anything they choose to bring and moor themselves.
A summary written by the operator and a telemetry stream are different artifacts, and only one of them can be independently checked. Handing over the second one is the point of the arrangement.
Can a satellite decide what the reactor runs?
No. It sets a search area and a priority order, and stops there.
Bounding a search and ranking a tasking is genuinely valuable and worth paying for. What the treatment stage actually runs comes from the profile taken on site and from the instruments watching the intake and the outlet, on the day.
Where the detail lives
Bring your monitoring people
Send the record you already hold.
Send the water body, the season and whatever record you already hold. The site read comes back with the stations we would want, the thresholds we would set, and the gaps in the existing data that would have to be filled first.