Phytoplankton at the Football Desk: Auditing a Domain-Classification Failure

The Stage-2 template opened with nine rows. Seven came back empty.

The first row asks how sophisticated a team's tactical structure is. The second demands a wage bill, an amortisation schedule, a net debt line. The fifth wants a tick against FFP/PSR compliance. Underneath those rows sat a file whose opening sentence was about the colour of the sea — and it was stating plainly that the water was not acquiring a new pigment, that its phytoplankton were thinning out. On top of the file, one label: Domain Label: football.

A single number in the whole document could be made to fit a financial row — 83 per cent. That is NOAA's forecast probability that this El Niño would be a "particularly intense" episode. It is a likelihood, not a currency. An automated pipeline finds exactly here its point of failure: the number is retrieved, the conditional clause around it is dropped.

Start with what the source actually contained, because an audit that skips that step is just an accusation.

El Niño is easier to file as weather than to understand as an ocean system. Surface water in the central and eastern equatorial Pacific warms above normal, and the sequence begins with trade winds slackening. Slack winds slow the process off the South American coast called upwelling — cold, nutrient-rich water rising from depth. Fewer nutrients mean fewer phytoplankton near the surface. Less chlorophyll means a different reflectance: the green-brown tilt recedes, the blue rises.

There is no mystery in that chain. It is standard ENSO oceanography and it holds together internally. What is new is the instrument. NASA's PACE — Plankton, Aerosol, Cloud, ocean Ecosystem — launched in February 2026, and the article calls this the first El Niño the platform has observed. New instrument, old mechanism. Kelvin waves, the equatorially trapped warm pulses travelling east, are the early signal specialists watch, and they were reported again.

The article also blocked its own easiest misreading: a colour change does not mean the water is literally taking on new pigmentation. That is a genuine quality marker in science communication, and I am recording it — because the same discipline is missing one layer down.

Now the audit.

Seven empty rows are a finding, not an omission. Tactical analysis returned null because the material contains no formation, no pressing trigger, no PPDA, no xG. Club finance returned null because there is no club, no owner, no wage bill. League landscape returned null because there is no table, no tier, no promotion-relegation structure. Governance returned null because FIFA, UEFA and every national association are absent. Dressing room returned null because no coach, captain or player is named. Sporting risk returned null because no contract, competition or transfer is exposed.

The three dimensions that did return something returned it from outside football. Narrative analysis placed the piece in an emergence phase — the novelty premium that attaches to first-of-instrument observation and pulls more media weight than confirmatory value. The results-cycle analogue could only confirm that the process evidence is internally consistent. The transmission map that emerged was ecological: physical forcing into nutrient supply into primary production into colour signal, food chain and CO₂ drawdown. All three conclusions are valid. All three belong to another desk.

The fastest signal was the field that came back as an instruction. Stage-1's "Entities Involved" column arrived unpopulated, carrying a note: identify from the information points above. That is not a routine glitch; it is an admission. The extractor found nothing and returned its emptiness in the template's own grammar. A pipeline that reads that field as "to be filled later" can never fail — it has no failure state to reach.

The temptation test. I could have written twelve hundred clean words. Upwelling as the lower block breaking under sustained pressure. Trade winds as the direction of build-up. Kelvin waves as the transition travelling from midfield into the final third. It would have read well and been entirely false. Analogy and invention differ on one point: analogy needs a shared mechanism, not a shared shape.

My own method rests on that distinction. Before the 2026 Under-17 World Cup I built a twelve-zone grid to map England's 4-2-3-1 press, then wrote on Rhian Brewster's movement — eight goals in the tournament — and Phil Foden's half-space occupation, with a clip timestamp against every zone, or the file did not go. After Belgium beat Japan 3-2 in Russia in 2026 I sat in Ahmedabad until four in the morning re-watching Roberto Martínez's shift from 3-4-3 to 3-4-2-1. It wasn't magic; it was a causal chain with three broken links, and every link carried a timestamp. The grid is not a cage; it is a receipt, and you have to show it with every claim.

On this ocean piece the receipt does not apply, because the claims are not football claims.

Two-tier provenance. Eight of twenty-three information points are sourced to "None" or "Provided information." The institutional claims are solid — NASA as PACE operator, NOAA as the source of the 83 per cent figure. The explanatory scaffolding carries no attribution and reads like textbook-standard compilation rather than original reporting. On a science desk that is a small matter. On any desk it is the discipline I have been asking for across three decades.

The 83 per cent figure needs its own warning. It is a probability for a "particularly intense" episode, and "particularly intense" is a definitional judgement. Quoted alone, the judgement disappears and the number hardens into certainty.

One correction, and this piece gets only one. The chronology collides. Observations are dated 2026, PACE launched in February 2026, and the article calls this the first El Niño PACE has observed. All three can coexist; none is verified. Until the original publication metadata is checked, treat it as open — this may be a forward-dated piece, or a transcription artefact.

The part least likely to be heard: the classifier is not the culprit.

The template is. Nine rows make an answering machine, and an answering machine returns what it is asked. A system with no slot for "this does not belong here" will eventually answer everything — sometimes weakly, sometimes with total confidence, always with an answer. In my experience the football desk's worst failures were never wrong analyses. They were moments when the desk had no format in which to decline.

One good-faith explanation belongs in the record, because not every error is a conspiracy. A question-format headline — why is X changing? — can trip a template-default label, and a human router on deadline makes the same call. The error is natural, not rare. The danger is that it stays invisible.

And the sharpest thing in this whole episode is that the material did carry a real football thread — not tactics, heat. Ocean and atmosphere data connect to climate and thermal load; thermal load connects to tournament scheduling, kick-off windows, wet-bulb globe temperature protocols and pre-season dates. Any match played on a May afternoon in Ahmedabad can speak to that. The pipeline had a row for formations and a row for wages, and no row for climate. That is where the information was actually lost: nobody asked the question that should have produced a different article.

Every domain failure is a causal chain wearing a disguise.

The way forward is technical and small: a domain-assertion gate that closes when zero domain entities are extracted. Fail-closed, not fail-open.

A checkable target has to sit alongside it, or this becomes one more good intention. Across the next hundred sampled items, count how often the entity field returns an instruction instead of a list. Today's number can be anything; the target is zero, and the number should be published.

The larger question outlasts the template: how many of those nine rows leave space to write "this does not arrive here"? Until that space exists, the colour of the sea will keep returning to the football desk — and next time it will not come back so honestly empty.