The Chemical Ledger · #1

The Chemical Century

How Tunisia's phosphate coast was built, measured, and not saved — 1885 to 2026

2026-08-14 · 18 min read
How to read this investigation Evidence badges · Entity links · Corrections

This is the narrative edition of The Chemical Century, the DeepTunisia investigation into Tunisia's phosphate coast. Each factual claim carries an evidence badge showing its grade. Each person and organization name links to the knowledge graph. Each key claim links to the evidence ledger.

Evidence badges

Every figure in this article carries one of these grades:

  • documented — located in a primary record: the 2025 E&S audit (including its relay of the 2010 campaign), ministry annexes, JORT decrees, the official production series, peer-reviewed papers, or the group's own history pages. Documented negatives sit here too. "Searched and not found" describes the search, not the world; it can always be broken by a record the search missed.
  • reported — credible secondary source relaying a figure or attribution that could not be verified against a primary document.
  • unsubstantiated — circulating with no verifiable base. Kept, never deleted: an investigation holding only what it can prove could not show why the popular account is wrong.
  • disputed — shown as an exclamation mark on a claim chip. Two or more sources disagree, unresolved; each side's position sits in the ledger entry.
  • air — a layer tag, not a grade. An air-layer item is still graded documented, reported or unsubstantiated like everything else.
  • derived — arithmetic done for this investigation lives in the ledger notes, graded by the quality of its inputs.

Conventions in this article

Entity mentions appear inline with the name of a person or organization — for example, the Groupe Chimique Tunisien. Each entity links to the DeepTunisia knowledge graph.

Claim references appear in square brackets — for example, — linking factual claims to the evidence ledger in `evidence.yaml`. Where a claim ID appears, the statement it tags is sourced and graded in the evidence table.

Interpretation panels are set off from the prose. They carry the editors' own readings of the documented record: each names who makes the claim, states the reasoning, and gives the observable that would prove it wrong. They are not facts, and they do not render like facts.

Corrections from earlier cuts

  1. "8.2 Mt (2010)" is a rounding of 8.131 Mt — 2000 (8.3 Mt) was the record.
  2. 14–15k t/day is a 2025 measurement, not a 1972 rate.
  3. The group's founding is 1992, not 1972.
  4. "95% of atmospheric pollution from GCT" is not a measurement — it is a 2003 court-expertise claim (Medhioub), relayed by the 2017 EU study and the 2025 audit ("d'après une étude"), never re-measured.
66 claims
133 sources
8 disputed
1 unresolved

The hook

In October 2025, the city of Gabès stopped. On the eleventh, protesters stormed the gates of the — the state-owned chemical group that has run this complex since 1992. On the twenty-first, the called a general strike , and the governorate stood still. On the twenty-second, more than twenty schoolchildren were hospitalised at the Collège Chatt Essalem after a gas leak from a GCT unit — the school named for the beach the factory faces. The gas itself: the record does not say .

And the measurements that had been true for decades finally became public documents. Into the sea: every day, fourteen to fifteen thousand tonnes of gypsum . Into the air: sulphur dioxide, nitrogen oxides, fluorides, ammonia, dust — measured once, in March 2010, at up to twenty-six times the legal limit for sulphur, seventy times for fine particles , and never measured again . From the state's own company, which the state owns at 99.998 percent . But this story does not start in 2025. It starts before the factory. It starts with a rock, a railway, and a Gulf that fed a country.

The discovery and the railway

In April 1885, a French amateur geologist named Philippe Thomas found thick layers of calcium phosphate on the north slope of Jebel Thelja, above the town of Metlaoui. In 1897, a company was created to dig it — the CPGCFG , backed by the Mirabaud and Hottinger banking houses of Paris, granted an exclusive sixty-year concession plus thirty thousand hectares of state land in exchange for building a railway to the coast. The line reached Sfax in 1899: 242 kilometres . In 1916, a branch reached Gabès — 83 kilometres to the town that would one day host the factory .

And the sea that coastline faced? The richest fishing ground in the southern Mediterranean — surveyed from 1925 by Le Danois , mapped by Calypso in 1965, and in 1971 by two Tunisian scientists of the Salammbô institute, Ktari, Chakroun & Azouz , who published the last great survey of the Gulf before the factory. The Gulf was not a waste pipe. It was a nursery — mapped, measured, prized. The timeline records both things, in that order.

The first factory and the state

Meanwhile, Tunisia decided to stop exporting rock and start processing it. In 1947 the state created the SIAPE — SIAPE — at Sfax, with a French industrialist, François Breynert , as its first president. In 1952 its TSP factory started — triple superphosphate, 330,000 tonnes a year .

And after independence, the state took the whole chain, company by company: 1962, the M'dhilla company became fully Tunisian; 1967, the old concession railway was absorbed into the state railway, the ; 1969, the pieces merged; and in January 1976 everything was consolidated under one name — the , the CPG — "an enterprise in which the state holds the majority of the shares", in the group's own history; entirely owned by the Tunisian government, says the NRGI report.

But the biggest single decision was still ahead: where to put the chemical plants. The answer was Gabès — a town on the Gulf, at the end of a railway branch that had been waiting fifty-six years for its cargo.

The complexes go in — two streams from day one

Then the factories went in, one after another — each with its own company, each with its own president, and each with the same two disposal habits.

In 1972, the phosphoric acid plant opened at Gabès, in the Ghannouch industrial zone, under the name ICM — the ICM — 470,000 tonnes of merchant acid a year. Who built it? The French engineering group SPIE Batignolles — and the company's own environmental director, Noureddine Trabelsi , told investigators in 2015 that SPIE Batignolles "programmed" the sea discharges from the start.

And the air: the same plants that made the acid also made the plume — sulphur dioxide from the sulphuric units, nitrogen oxides from the nitric acid, fluorides from the phosphoric acid, ammonia and dust from the DAP granulators and the ammonium-nitrate prilling towers . Two streams, both from day one: the grey stream into the Gulf, the pale plume over the town.

In 1974, a second acid line. In 1975, a new company, the SAEPA — SAEPA — whose president, Ali Boukhris , was simultaneously president of the ICM, of SIAPE, and of SAEPA through the whole first decade — one manager over the three companies that built both pollutions. In 1979, the DAP plant — 650,000 tonnes a year. In 1983, the ammonium nitrate unit, 330,000 tonnes — a unit whose prilling tower is, even today, the single most expensive air fix in the corrective plan. In 1985, a second DAP line under EG , and a TSP plant at M'dhilla run by the ICG . In 1988, the phosphoric acid plant at Skhira . Four sites by 1989.

And every one of them — every single one — discharged its gypsum into the sea and its gases into the air . Neither stream is a malfunction that arrived later. Both were there from the first day of production.

They knew — the documentation clock

Now the part of the century that changes the meaning of everything before it.

The sea damage was measured early: the state's own institute at Salammbô INSTM began its surveillance in 1976 , four years after the plant opened — Darmoul, Hadj Ali Salem and Vitiello measured lethal concentrations at the outfall by 1978 and 230 hectares of dead seabed by 1980 ; in 1988 Darmoul recommended storing the gypsum on land ; in 1989 the state adopted its own water standard, NT 106.002 , seventeen years after the discharges began.

The sea was measured. Now the air.

In March 2010 — the only time it has ever been done — the company commissioned an engineering consortium, COMETE-SNC-LAVALIN-IHE, to measure what the stacks put out. The results, relayed by the 2025 audit: sulphur dioxide at up to 7,800 milligrams per cubic metre against a limit of 300 — twenty-six times . Fine particles at 352 against a limit of five — seventy times. Fluorides at four times the limit.

The same year, 2010, the state adopted its own source-emission standard, Décret 2010-2519 . And then — nothing. No campaign since March 2010. No continuous monitoring . The audit calls it a major non-conformity: "no measurement campaign more recent than March 2010 was identified." Fifteen years, one measurement.

And the famous number about the air — "ninety-five percent of the region's atmospheric pollution comes from the plants" — is not a measurement at all . It comes from a 2003 expertise commissioned by the court at Gabès, a man named Medhioub , relayed by the 2017 EU study, repeated by the 2025 audit with the hedge "according to a study" — and never re-measured.

The sea was measured, and measured again. The air was measured once. The claim about the air was never measured at all.

The merge and the peak

In 1992, the state finished the foundation it had begun in 1947: SIAPE and SAEPA merged into a single group — the — after the ICM, Engrais de Gabès and the Industries Chimiques de Gafsa had been absorbed into SIAPE. One name. One owner. And in 1994, the state put the mining company and the chemical group under a single chief executive — Rafaâ Dekhil , from 1994 to 2001.

From then on, the top of the chain was concentrated to a degree the numbers make concrete: 99.998 percent of the group in one hand , and from 1994 one chief executive over both the mining company and the chemical group . And what did that concentration do with the machine?

It ran it to its record: the official series shows 7.8 million tonnes in 1998, 8.03 in 1999, and a peak in the year 2000 at 8.3 million tonnes . That is the true record — not the famous 2010 figure, which is 8.131, rounded to "8.2" by everyone who repeated it .

In 2008, output dipped to 7.6 — the year the mining basin of Gafsa revolted over a hiring contest . In 2010, it recovered. And in 2011 — the year of the revolution — it collapsed: 2.28 million tonnes . The machine lost two-thirds of its output in a single year. The waste did not collapse with it.

And by 2014, the seagrass was ninety percent gone — from roughly 1,300 square kilometres in 1970 to less than 150 — the calculation of El Zrelli & Rabaoui . The output curve fell. The waste curve did not. And the plume — the plume had no curve at all: it had one measurement.

The half-broken machine, the pricing, and the air

After the collapse, the machine never recovered. A decade at three and a half million tonnes — half of the old output. The Gabès complex ran at forty percent or less, by its own director-general 's account; some units at twenty, said the engineer Houcine Rhili .

And the air kept doing what the air does.

The hospital data exists — a study by Kwas and colleagues at Gabès University Hospital, a pneumonia cohort from 2022, found that outdoor sulphur dioxide was significantly associated with patients needing non-invasive ventilation and higher severity scores, and fine particles with lung abscess and pleural effusion . The ambient measurements exist — Ahbil, Sellami & Azri measured the suburban air in 2017 to 2019 and found it unhealthy for sensitive groups more than one day in five . And the lived record exists — the 2025 audit's own surveys of residents and workers: unanimous that the situation is "very grave"; the workers naming sulphur oxides and fluorinated gases; asphyxiation, respiratory and eye and skin irritation, a persistent sharp sulphur smell . The audit itself notes these are perceptions, not yet measured — its own recommendation.

And the state's responses to the health question have a pattern: in 2013 and 2014, a project for an offshore cancer hospital at Gabès collapsed in mutual recriminations . In March 2021, a fire at the unit killed five workers, and the president announced a cancer-treatment centre . In October 2025, the health minister announced a specialised cancer-disease pole for Gabès within two years . Three responses in twelve years. And no incidence statistics: the only circulating figure — "ten percent lung cancer" — traces to a single academic article whose own source is unread . The state treats the burden. The burden is not measured.

And in the ledger of the damage: the fisheries at sixty million euros a year , the ecosystem loss a hundred and five million , the machine at forty percent — and the air, measured once, in 2010, never again.

The crisis year

The present, then — and it is a story of its own, in two streams.

On the fifth of March 2025, a restricted cabinet council — chaired by Prime Minister Kamel Maddouri, presented by the Minister of Industry Fatma Thabet Chiboub — multiplied the production target by five, to fourteen million tonnes, and in the same meeting reclassified phosphogypsum as a "value-added product" . Same factory, same squares, new word.

In July, the environmental audit — commissioned by the company because the African Development Bank required it — published the daily numbers: fourteen to fifteen thousand tonnes of gypsum into the Gulf , and the record of the air: the 2010 campaign's exceedances , the absence of any campaign since, the absence of continuous monitoring .

In October, the city erupted — and the air was the trigger. On the fourteenth, children at the Collège Chatt Essalem were asphyxiated after inhaling toxic gases from a GCT unit; some lost consciousness . On the twenty-second, more than twenty were hospitalised after a gas leak . The gas is never named in any report — the record ends at "a unit of the GCT" .

On the twentieth, the minister Slah Zouari told parliament that six unfinished environmental projects, about two hundred million dinars, were "directly linked to toxic gas emissions" — and gave the progress: nitrogen oxides, ninety-eight percent; sulphur dioxide, seventy-five; ammonia, eighty-four . The same projects the company's own 2009–2020 environmental plan promised, still unfinished in 2025 . Unfinished through three governments since 2009 , they support two readings at once: deprioritised by choice, or stuck in the machinery of the state itself. The record does not separate them. Of the two, incapacity is not the milder finding.

On the twenty-sixth of February 2026, the tribunal rejected the motion to halt the units: the harm, it said, was not established . The target was walked back three times . The minister was dismissed . took both groups . The guarantees grew by three hundred million dollars and euros . The word changed. The target shrank. The money grew. The discharge continued — into the sea, and into the air.

The wall

And the rules that are not written — now in both layers.

The company's own annual reports stop at 2014 . The ministry's 2023 row is blank . The task force's plan — Ali Ben Hammoud's team — never published . The transfer price, never published .

And the air: the emissions were measured once, in 2010 — and the audit's own finding is that there has been no campaign since and no continuous monitoring : the air is the least-measured thing in this whole story. No air-quality series for Gabès turned up in the search . No cancer-incidence statistics either . The gas that hospitalised twenty children on the twenty-second of October: unnamed .

One document exists, and it exists because the bank required it .

And the editor of the magazine that publishes the company rankings — Hédi Mechri — wrote, in his editorial for the 2026 edition: "The black-out resurfaces. Move along — there is nothing to see" .

Across a century, that concentration produced factories, output, waste, standards, documents, a task force. Into the air it produced a plume, measured once. What it failed to produce was the record. The timeline keeps walking. The wall stays up.

The answer

Can the moral question be answered? Three times yes.

Legally: the state's own audit says the discharges violate the state's own standards — the water standard of 1989 , the air standard of 2010 .

Economically: El Zrelli and colleagues put the 2014 damage at one hundred fifteen percent of what the factories add . One team's estimate; its components and method are noted in the ledger.

Definitionally: the state reclassified the waste instead of treating it .

How much does it make? Two to three billion dinars a year, on the record that exists — and the most famous figure is the least reliable. How much does the state make? Less than it spends, on the record. How much do the people pay? More than the factories add. Is it okay? By its own documents, no.

And the air: the state's company measured its own stacks once, in 2010 — sulphur at twenty-six times the limit, particles at seventy — and never measured again . The "95%" the whole country repeats was never measured at all . The gas that hospitalised twenty children was never named . The cancer the state keeps promising to treat has no count in the searched record .

The sea is measured. The air is named. The distance between those two facts is the finding this whole ledger has been building toward.

The timeline reaches 2026, and it does not stop. The next point on it is being decided now.

Evidence ledger

ClaimGradeTextSources
C001documented1885 — Philippe Thomas discovers Metlaoui phosphate (April, Jebel Thelja)S41
C002documented1897 — CPGCFG founded (convention 1896; décret 22-05-1897); Mirabaud-Hottinger / Mokta-el-Hadid / NervoS41
C003documented1899 — Sfax–Gafsa–Métlaoui rail, 242.2 kmS41
C004documented1916 — Graïba–Gabès rail, 83.1 kmS41
C005documented1925 Le Danois; 1965 Calypso (De Gaillande 1970); 1971 Ktari-Chakroun & Azouz (INSTM Bull. 2(1):5-27)S26, S37
C006documentedOutput series 1900–2011 (0.2 → 8.3 → 8.131 → 2.28)S14
C007documented'8.2 Mt (2010)' = rounding of 8.131; 2000 (8.3) is the recordS14, S51
C008documentedSIAPE 1947 (vs 1948)S41
C009documentedFrançois Breynert — first PDG SIAPE (1947–1953)S41
C010documentedTSP 1952, 330,000 t/yrS41
C011documentedNationalisations 1956 → 1962 (CIPHOS) → 1967 (SNCFT) → 1969 (merger)S41
C012documentedCPG formed 01-1976; instrument not foundS41
C013documentedComplexes with names + capacities: ICM MGA 470k (1972) · SAEPA I DAP 650k (1979) · SAEPA II ammonitrate 330k (1983) · EG DAP (1985) · ICG M'dhilla TSP 465k (1985) · Skhira PA 375k (1988)S41, S8
C014reportedSPIE Batignolles built the first plant — per GCT env. director Noureddine Trabelsi (2015)S81
C015documentedAli Boukhris — PDG ICM (70–81) + SIAPE (70–76, 79–81) + SAEPA (75–81), parallelS41
C016documentedGérard Valier — first PDG EG (73–77); Mohamed Mghaieth — first PDG ICG (81–85)S41
C017documentedDischarge to sea since 1972S8, S25
C018documented1976 — six-year surveillance (INSTOP/INSTM); siltation first describedS37, S35
C019documented~1977–78 — lethal PG concentrations in situ (Darmoul & Vitiello)S37
C020documented1980 — 230 ha benthos destroyed (Darmoul, Hadj Ali Salem & Vitiello)S37
C021documented1988 — red tide (Hamza & Ben Maiz 1990); 1988 — land storage recommended (Darmoul 1988)S37
C022documented1989 — NT 106.002 adopted (arrêté 20-07-1989, JORT n°59, effect 01-10-1989)S37
C023reported1990 — no living plants (Zaouali 1993, via El Kateb 2018)S35
C024documented1992 — GCT formed (SIAPE+SAEPA); absorption 1988/89 disputeS41
C025documented1994 — single PDG CPG+GCT: Rafaâ Dekhil (1994–2001)S41
C026documented2014 — ~90% Posidonia loss; ~1,300→<150 km² (El Zrelli et al. 2023)S26
C027documented2008 Gafsa revolt (context beat; output 7.6 Mt)S41
C028reportedActivity ≤40% (Hédi Youssef, 03-2026); ~20% units (Houcine Rhili, 02-2026)S20, S21
C029documentedM'dhilla 2 created 2010; idle since 2020S41, S22
C030documented14–15k t/day (2025) = 5.1–5.5 Mt/yrS8
C031reportedCumulative ~200 Mt (Samir Kazbar, via TAP) vs >500 Mt wet (El Zrelli 2023)S27, S26
C032reported2003 — Medhioub expertise (tribunal de Gabès); orchards halved, date palms −80%; origin of '95%'S8, S25
C033reported2017 — EU study (PARTICIP GmbH & ETI Consulting) ≈76m TND/yr (2015), 47/33/7 — floorS32, S33
C034documented2017-06-29 — relocation decision under PM Youssef Chahed, never implemented (2030)S8, S37
C035documentedFisheries ≈€60M/yr; >€750M cumulative (El Zrelli 2020); 65% of landingsS34, S35
C036documented€105M (2014) ≈ 115% of factories' added value (El Zrelli 2023)S26
C037documentedCMR 05-03-2025 (PM Kamel Maddouri; Min. Industrie Fatma Thabet Chiboub): ×5 + reclassification + VAT + new units + M'dhilla 2 supportS51, S52, S85
C038documentedAudit 24-07-2025 (I2E/IHE; AfDB P-TN-BB0-010), 160 pp, bank-requiredS8, S47
C039reportedProtests/storming/strike/Tunis march Oct 2025; >20 schoolchildren (Collège Chatt Essalem)S54, S55, S77, S80
C040reportedRéféré 26-02-2026 (Me Mehdi Talmoudi; 300 certificates, 30 filed): 'préjudice non établi'S78, S44
C041reportedTarget walk-backs: 13.6 (Youssef, 03-2026) → ~11.4 (Plan, 06-2026) → 9.4 by 2035 (CNRD)S20, S62, S19
C042documented27-04-2026 — Chiboub dismissed (no reason published); interim Slah ZouariS61
C043documented01-06-2026 — décret n°79: Omar Bouzouada PDG of CPG and GCTS30, S83
C044documentedGuarantees 2026 ≈ $120M ITFC + €110M+€7M EBRD + $110M AfDB (lois 49/50/51/2026)S69, S70, S71, S47
C045documentedReports stop 2014; 2023 blank; task-force (Ben Hammoud team) unpublished; transfer price/health/LF2026/envelope not publishedS86, S10, S75, S88, S94, S79
C046documentedEditor quote — Hédi Mechri, edito #937: 'move along — there is nothing to see'S64
C047documentedRevenue series 2019–2022: 1,275 / 936 (−608 net) / 1,979 / 3,179 m TND; 3,400m (2024) '+247%' single outletS9, S10, S1
C048documentedPlume composition: SO₂, NOx, HF, NH₃, dust from acid/DAP/ammonitrate/prilling unitsS8
C049documented2010 campaign (COMETE-SNC-LAVALIN-IHE, March 2010): SO₂ 7,805/2,008/1,235 vs 300 · NO₂ 924 vs 500 · HF 20/15/13 vs 5 · PM₁₀ 352 vs 5 · dust 123–248 vs 50 · CO 96/248 vs 10 — start-up framing is the audit's ownS8
C050documentedNo campaign since March 2010; no continuous monitoring — audit major non-conformity §6.3.4S8
C051documentedDécret 2010-2519 (28-09-2010) — source-emission limitsS8
C052documentedDécret 2018-447 (18-05-2018) — ambient-air limitsS8
C053reported'95%' chain: Medhioub 2003 (court) → EU study 2017/18 → audit 2025 'd'après une étude' — never re-measuredS8, S25
C054documentedKwas et al. 2024 (Cureus 16(8):e66578, open access): Gabès hospital pneumonia cohort Jan–Oct 2022 — SO₂ ↔ NIV/CURB65/PSI; PM₁₀ ↔ lung abscess/pleural effusionS39
C055documentedAhbil, Sellami & Azri (Environmental Forensics 27(1):90-108): suburban Gabès 2017–19 — >53% moderate (PM₁₀), >20% unhealthy sensitive, ~3% very unhealthy-to-hazardousS40
C056reportedSellami, Gautam & Azri 2026 — Ghannouch (GCT fence-line) air + health studyS39
C057documentedAudit resident surveys: 66 questionnaires (§3.4.4); §6.8 — unanimous 'very grave', asphyxiation, respiratory/eye/skin irritation, sulphur smell; workers name SOx + fluorinated gasesS8
C058reported14-10 / 22-10-2025 school asphyxiations (Collège Chatt Essalem); 18-11 wave — gas unnamedS77
C059reported20-10-2025 — Min. Slah Zouari: six unfinished projects ≈200 MDT 'directly linked to toxic gas emissions'; NOx 98%, SO₂ 75%, NH₃ 84%S36
C060documentedGCT 2009–2020 plan: same 10 Gabès projects, declared 1,323 MDT, unfinished; 'surveillance continue' claim contradicted by its own auditS8
C061reportedCancer-pole chain: 2013–14 'cancéropôle offshore' collapse · 13-03-2021 fire (5 dead) + Saïed centre · 21-10-2025 Ferjani pole (2 yrs)S77
C062unsubstantiated'10% lung cancer' — Rousselin 2018, source inside unread paper — circulatingS77
C063documentedGabès cancer-incidence statistics: none found; air-quality series: none foundS8
C064documentedAir corrective plan ≈159.3 MDT (DAP NH₃ 24 · DENOX 6.2 · prilling 93 · SOx 4 · odour 20 · N₂O 6.1 · CO₂ 5 · stations 1)S8
C065documentedAhbil 2024 STOTEN (dry-deposited particles — air→ground channel); Bouajila 2022 (phthalates in Gabès soils); Mokadem 2012 (Gafsa-basin analogue)S40
C066documentedGCT state share 99.998% of capital (MoF LF2021 Annexe 9; capital 476.084m TND)S67