Knowing where it is tells you little about where it goes.
Detection and prediction share a vocabulary, which is how monitoring budgets buy one and get billed for both.
The distinction
Two different products, one word.
Detection answers a question about the present: is there something in this water now, where, and how much. It is validated by going and looking.
Prediction answers a question about a state that has not happened. It is validated statistically, over many events, against what a naive guess would have achieved, and it needs different procurement language.
The confusion is expensive. A program buys a system on a compelling hindcast, then learns in its second season that a hindcast is a curve fitted to known events and the operational product has never been scored.
What a forecast owes you
Four things, all of which should be in writing.
Lead time, stated as a decision window
A number of hours or days, tied to a decision somebody actually takes — reduce abstraction, cancel a harvest, staff a night, retask an in-water program already working the zone. After the decision point a forecast is worth nothing, however accurate.
Skill against a named baseline
The baseline is usually persistence: tomorrow looks like today, and it is embarrassingly hard to beat over short horizons. A model that does not report skill relative to persistence and climatology has been demonstrated and never evaluated.
A false-alarm rate you can live with
Every threshold trades missed events against false alarms; no setting avoids both. A plant that can throttle cheaply wants a sensitive trigger; a resort that must not close twice in a summer does not.
Verification, published, after the fact
Each season, scored against what happened, in a document somebody outside the supplier can read — ideally one that held the raw observations all along. Systems that publish their misses improve. Systems that publish only their hits are marketing.
Where the physics ends
Transport is arithmetic. Growth is still biology.
That split explains most of what bloom forecasts get right and wrong.
Moving an existing patch of water is a fluid dynamics problem, and fluid dynamics is mature. Given a decent circulation model, wind forcing and a known starting distribution, where a surface bloom drifts over the next two or three days is tractable.
Predicting whether a population initiates, blooms, produces toxin or collapses is a biology problem with nutrient supply, light, temperature, grazing, strain variability and competition in it. Those processes are parameterized from limited data, and the parameters are not stable between regions.
So forecasts of drift and arrival tend to be useful, and forecasts of initiation and toxicity tend to be weak. Ask a supplier which one they are selling.
Models are only as good as their initial conditions, which come from sparse, cloud-interrupted remote sensing observations. And most coastal regions lack the long, consistent, quality-controlled record a statistical model needs to learn from.
Building the chain
What sits behind an operational bloom forecast.
Five components. A program missing any one can still produce a map, which is why so many programs produce maps.
- A physical model of the water body Circulation, stratification, exchange and residence time, forced by a meteorological model. In an enclosed basin or engineered coastline, local bathymetry and structures outweigh any global product.
- An observation stream to initialize from Satellite fields for extent, moored instruments for rate of change, samples for identity. Most engineering effort goes into assimilation, which demonstrations rarely show.
- A biological component, bounded Growth and decay terms with parameters fitted somewhere. Ask where. Parameters carried from a temperate estuary into a warm, hypersaline gulf are a known source of confident nonsense.
- A decision layer Thresholds, alert levels and a defined recipient, agreed with the people who act. A forecast with no named recipient is a research output, and should not be billed as an operational service.
- A verification loop Scores computed every season and fed back into the model and thresholds. Without it neither the system nor its users improve.
The same discipline, applied to treatment
A mechanism explains. A measurement is what counts.
A mechanism tells you why an effect should occur, and nothing about its size in your water last August. A forecast built on mechanism without measurement inherits every assumption underneath it.
Nanobubble treatment sits across three tiers of certainty. Some of it is measured and barely disputed: oxygen nanobubbles in water carry zeta potentials of roughly minus thirty-four to minus forty-five millivolts, which keeps them from coalescing.
Oxygen transfer is measured and still moving: a laboratory study reports nanobubble aeration 1.5 times as efficient as coarse bubbles, and other comparisons report more from different setups. So it enters a forecast with a wide error bar, and anything sized off it uses 1.5.
Radical generation is contested. A 2020 study from Moleaer and Arizona State University reported that injected nanobubbles produce reactive oxygen species including hydroxyl radicals. A controlled 2023 study by Chae and colleagues in ACS ES&T Engineering found that generation minimal, if it occurred at all, under the ambient conditions tested.
Build on the first, bound the second, hold the third open. A forecast owes the same handling to a well-constrained variable, a loose one and a parameter nobody has pinned down.
The same discipline is what turns a forecast into something a third party can act on financially. Put the comparator beside the skill figure. Put the false-alarm rate beside the hit rate. Say how many seasons and how many events the score was computed over, and who computed it. A product specified that way can carry a trigger written before the season and adjusted on a published cycle; one specified loosely can carry a map and an opinion.
Analytics inherit the treatment case
A program that routes an intervention with analytics inherits responsibility for saying what it did. If the treatment case rests on an unresolved mechanism, so does the verification, and the chain becomes unfalsifiable.
If you are evaluating a system
Six questions, in order.
Which decision is this forecast for, and who takes it. What is the lead time at that decision point. What skill does the product show against persistence over the last two seasons on our own water. What is the false-alarm rate at the threshold you propose. Where were the biological parameters fitted. And who publishes the verification.
They should arrive as answers. A demonstration is no substitute for the skill score.
A seventh sits underneath: what will act on the answer? Skill is capped by response. Seventy-two hours of warning is seventy-two hours of work where an in-water program already runs the zone, and a better-documented loss where nothing does. The two purchases are usually made by different directorates in different years, and the second decides what the first was worth.
In-water treatment is therefore sold as a continuous subscription rather than a response summoned once a forecast verifies. Stressors accumulate the whole time, so the program works the whole time; the forecast changes what it does on a given day, never whether it is out there. Alarivean's control stack is built to turn a risk surface into a vessel's next task; HABguard covers the site when that task arrives; HABslayer covers the treatment and its permit.
Sources
- Science of the Total Environment — mass transfer of nanobubble aeration and its effect on biofilm growth. The abstract is paywalled and the method unread, so only the literature range is used.
- Springer — nanobubble stability and zeta potential.
- Chae, Kim, Kim and Fortner — reactive oxygen species generation from nanobubbles, ACS ES&T Engineering, 2023. The 2020 Moleaer and Arizona State University result reaching the opposite conclusion was issued as a press release and has not appeared in the peer-reviewed record.
Aquatic Prosperity as a Service
Score us against your own record.
Send two seasons of observations and the events you remember. Alarivean will say what that data could have anticipated, what it could not, and what an additional layer would have bought you.