Shortly after noon on Saturday, July 10, 1976, a reactor at the ICMESA chemical plant in Meda, north of Milan, overheated and vented a cloud carrying 2,3,7,8-tetrachlorodibenzo-p-dioxin, usually called TCDD, over the neighboring town of Seveso. No human deaths in the immediate aftermath were attributed to the release. Thousands of animals died, though, and children developed chloracne. The first residents were not evacuated until more than two weeks later. The disaster gave its name to the European directive on major-accident hazards. This guide covers the reaction, the slow disclosure, the zone map, and how the European model differs from the American one.

ICMESA at Meda: a trichlorophenol batch left in the reactor for the weekend

ICMESA, short for Industrie Chimiche Meda Societa Azionaria, ran a chemical plant in Meda, in the Brianza district of Lombardy roughly 20 kilometers north of Milan. The company belonged to Givaudan, a Swiss fragrance and chemical firm that was in turn part of the Hoffmann-La Roche group. One of the plant’s products was 2,4,5-trichlorophenol, usually shortened to TCP. TCP is an intermediate used to make the antibacterial agent hexachlorophene. The plant stood among houses, small farms and workshops, and the town of Seveso lay just to the south. I had no role in any of this. Everything here comes from the published record, chiefly the technical analyses of the accident, the epidemiological literature that followed the exposed population, and the European legislation the accident prompted.

The plant made TCP in batches. The process reacted 1,2,4,5-tetrachlorobenzene with sodium hydroxide in a solvent of ethylene glycol, and steam supplied the heat. TCDD forms as a by-product of that chemistry. It appears in trace amounts when the reaction runs at its intended temperature and in far larger amounts when the temperature climbs above that range. The industry knew this well before 1976, because earlier accidents at trichlorophenol plants, including one at BASF in Ludwigshafen in 1953, had already caused chloracne among workers. TCP was also the starting material for the herbicide 2,4,5-T, and TCDD is the same contaminant that made Agent Orange notorious.

Accounts of the accident agree on the basic sequence. A batch was in progress on the night of Friday, July 9, and it was stopped early on Saturday morning before the process was finished, because the plant was shutting down for the weekend. By the usual account the operators shut off the heating and then the agitator. They did not complete the step that would normally have distilled off much of the ethylene glycol. That left the reactor holding a hot, unfinished mixture with no stirring. Most histories say the shutdown was required by Italian rules on weekend working, and the technical literature repeats that explanation.

July 10, 1976: a runaway reaction and a relief line open to the air

The process safety literature explains the runaway through the steam used to heat the reactor, and Trevor Kletz’s widely used case studies give the same account. The steam came from the exhaust of a turbine elsewhere on the site. Because much of the plant was shutting down, the turbine was running at reduced load, and its exhaust steam was considerably hotter than normal. That steam heated the reactor wall above the liquid line to a temperature the process never reached in normal running. Once the agitator stopped, heat from the wall passed into the stagnant upper layer of the liquid. That was enough to start a slow exothermic reaction, which then accelerated on its own heat. This is the accepted technical explanation. Readers who want the primary record should go to the Italian parliamentary inquiry and the engineering reviews built on it rather than to secondary summaries.

At about 12:37 p.m., several hours after the batch had been stopped, pressure inside the reactor burst its rupture disc. The relief line discharged directly to the atmosphere above the building. It had no catch tank, knockout drum or scrubber, so the vapor and entrained liquid went straight into the air. The discharge is commonly described as lasting around twenty minutes. There was no fire and no structural explosion, and the plant showed little physical damage. That helps explain why the seriousness of the release was not obvious to people outside the fence that afternoon.

How much TCDD left the reactor is genuinely uncertain. Published estimates range from a few hundred grams to several kilograms, and some run higher still. I have not been able to tie the two ends of that range to specific sources firmly enough to name them here. If you need a figure, take it from a named scientific review and quote that review’s own caveats. The direction of travel is not disputed. A light wind carried the cloud south and southeast from the plant, and the heaviest deposition fell in a narrow band across Seveso. Lighter contamination reached Meda, Cesano Maderno, Desio and parts of neighboring municipalities.

The first two weeks: a slow disclosure and a slower evacuation

The release came at lunchtime on a hot summer Saturday, when many families were at home and children were outdoors. Residents near the plant later described the cloud and a sharp chemical smell. Within days rabbits, chickens and birds began to die in large numbers, and leaves on trees and garden plants yellowed and dropped. Children were brought to doctors with burn-like lesions on exposed skin, although those signs did not identify the agent and bore little resemblance to the familiar picture of a toxic gas release, in which people collapse in the street within minutes.

By the usual account, the company told local authorities the following day that chemicals had escaped from the reaction, but it did not identify TCDD in that first notice. The mayors of Seveso and Meda posted warnings telling residents not to eat local fruit and vegetables and to avoid touching plants in the affected area. Samples went to Givaudan’s laboratory in Switzerland, which confirmed TCDD. Histories place that confirmation in the middle of the following week, and the regional and national authorities grasped the scale of the problem only after that. The exact dates differ somewhat between accounts. How much of the lost time was the company’s fault and how much belonged to the public authorities was argued over at the time and still is.

The first evacuation from the most contaminated area began on July 26, by the date most histories give. That was sixteen days after the release. The evacuation was extended in the following days as sampling widened, which means residents of the most contaminated zone had lived inside it for more than two weeks. Part of the delay came from the slow flow of information from the operator. Part came from the fact that nobody in Lombardy in 1976 had a plan or a legal threshold ready for persistent soil contamination by a substance that caused no immediate human casualties. The European directive that followed addressed both problems. It required operators to report major accidents promptly, and it required authorities to have off-site plans in place before anything happened.

Zones A, B and R: a map drawn from soil samples

The authorities divided the affected area into zones based on TCDD concentrations measured in soil samples. Zone A, nearest the plant, had the highest concentrations, and Zone B around it had lower ones. Beyond those two lay a wider buffer called Zone R, from the Italian zona di rispetto. Measured contamination there was lower still, and restrictions were imposed as a precaution. The epidemiological literature usually puts Zone A at mean soil levels above 50 micrograms of TCDD per square meter, with Zone B below that. The lower cut-offs for Zone B and Zone R are reported less consistently, so take them from a named paper. In practice the boundaries followed streets and other recognizable features so that residents and police could tell where one zone ended, and they were adjusted as more sampling results came in.

Population figures for the zones usually come from the cohort studies run by researchers at the University of Milan. Those studies give roughly 700 residents for Zone A, a little under 5,000 for Zone B, and somewhat over 30,000 for Zone R. Different papers report slightly different totals depending on how they counted people who moved, were born or died during follow-up. Use the figure from a named paper with its date rather than a rounded number from a news story.

The measures differed by zone. Zone A was evacuated and fenced off. Residents of its most contaminated core never went home, because their houses were demolished and the contaminated topsoil removed. Residents of Zone B stayed, but farming, keeping animals and eating locally grown food were prohibited, and histories describe additional advice aimed at young children and pregnant women. Zone R carried restrictions on local produce and animal husbandry. Thousands of people in Zones B and R spent months and years under rules about what they could grow and eat, and that is where the long tail of the incident lay.

The Seveso map shows that a chemical incident involves two different kinds of zone. Responders draw initial isolation and protective action distances in minutes from a guidebook table or a plume model, on the assumption that the hazard is airborne and passes. A contamination zone for a persistent substance is drawn over days and weeks from laboratory results. It governs where people may live and what they may eat, and it can stay in force for years. A jurisdiction therefore needs to know in advance who will collect and analyze the samples, and who has the legal authority to declare a residential area off limits based on the results.

The health record: what it supports and what it leaves open

A claim still circulates that the Seveso cloud killed residents outright, and it is false. No human deaths in the immediate aftermath were attributed to the release. The documented acute harm fell on animals and on people’s skin. The figures most often cited are about 3,300 animals found dead, mostly rabbits and poultry, and roughly 80,000 more slaughtered to keep TCDD out of the food chain. These numbers appear consistently in reviews of the accident, though I would still cite them with the source in hand. Chloracne is a persistent acne-like skin condition and a recognized marker of dioxin exposure. It was diagnosed in a number of people commonly reported as 193, most of them children, and the literature describes most cases as resolving over time.

The longer-term record comes mainly from epidemiological follow-up. Pier Alberto Bertazzi and colleagues at the University of Milan followed mortality and cancer incidence in the zone populations for decades. They have reported elevated rates of some cancers, including certain lymphatic and blood cancers, in the more exposed zones, with the qualification that the numbers in Zone A are small. The Seveso Women’s Health Study, led by Brenda Eskenazi at the University of California, Berkeley, has examined reproductive and other outcomes in women who were young at the time. Paolo Mocarelli and colleagues reported a shift in the sex ratio of children fathered by men exposed at a young age. Much of this work relied on blood serum samples collected in 1976 and stored. Those samples later allowed TCDD levels to be measured in individuals rather than inferred from where they lived.

None of these authors presents Seveso as either a mass-casualty event or a harmless one, and an emergency manager should not either. Their conclusions are limited by small numbers and by the long latency of the outcomes they study, which makes it hard to separate dioxin from everything else that happens to a population over decades. Some associations have been reported and others have not been confirmed, and the question remains open in the scientific literature. Fear of birth defects was intense in 1976. Italian authorities permitted therapeutic abortions for exposed women at a time when abortion was otherwise illegal in Italy, and that decision was bitterly contested. A reader should approach it through the historical literature rather than a summary like this one.

Cleanup, 41 missing drums and the court cases

Decontaminating Zone A took years. The contaminated topsoil, the rubble of demolished buildings and other waste went into two large engineered containment basins built within the zone. The land above them was later landscaped as a public park, the Bosco delle Querce. The ICMESA reactor and the residue inside it were a separate problem, because they held the most concentrated material of all.

That residue was packed into 41 drums, which left Italy in 1982 under disposal arrangements that kept their destination from the public. The drums then dropped out of sight, and their disappearance became a political scandal in several countries. They were found in May 1983 in a disused slaughterhouse in northern France and were eventually incinerated in Switzerland. The episode is commonly cited as one of the events behind the European Community’s 1984 rules on supervising cross-border shipments of hazardous waste. It is also cited as one of the events behind the wider international effort that produced the Basel Convention on transboundary movements of hazardous wastes in 1989.

The legal aftermath ran for years and was overshadowed by violence. In 1980, members of the armed left-wing group Prima Linea shot and killed Paolo Paoletti, ICMESA’s technical director. Criminal proceedings against company managers produced convictions at first instance that were reduced or overturned on appeal. Givaudan and Hoffmann-La Roche paid compensation to the Italian state, the Lombardy region and affected residents. I am not giving sentence lengths or compensation totals because the accounts I have seen differ in detail. Anyone who needs them should go to the Italian court record and the companies’ own published statements on the settlements.

The 1982 Seveso Directive and the two that replaced it

A tidy version of this history says the European directive was written because of Seveso alone, and that version leaves out too much to be accurate. On June 1, 1974, an explosion at the Nypro plant at Flixborough in England killed 28 people. It had already led the British government to set up an Advisory Committee on Major Hazards, and several member states were working on the problem of large chemical installations before July 1976. Seveso gave that work urgency at the level of the European Community, and it supplied the name. The result was Council Directive 82/501/EEC of June 24, 1982, on the major-accident hazards of certain industrial activities, which member states then had to transpose into national law.

The 1982 directive placed its main obligations on operators. An operator of a covered installation had to take the measures needed to prevent major accidents and limit their consequences. It also had to be able to show the competent authority that it had identified the hazards and taken those measures. Installations holding listed substances above threshold quantities had heavier duties. They had to send the authority a notification describing the substances, the installation and possible major-accident situations, and they had to plan for emergencies on site. Authorities had to make sure off-site emergency plans existed. People likely to be affected by a major accident had to be told about the safety measures and how to behave if one occurred. Operators also had to report major accidents, and member states passed that information to the European Commission. The resulting database continues today as the Major Accident Reporting System, run by the Commission’s Joint Research Centre.

The directive was amended twice in the late 1980s. The first change followed Bhopal. The second followed the 1986 fire at the Sandoz warehouse at Schweizerhalle near Basel, which sent contaminated firefighting water into the Rhine. In 1996 Directive 96/82/EC, known as Seveso II, replaced it. Seveso II based coverage on the presence of dangerous substances rather than lists of activities. It added requirements for safety management systems, land-use planning around establishments, attention to domino effects between neighboring sites, and inspection by the authorities. Seveso II was amended in 2003 after the fireworks explosion at Enschede in the Netherlands in 2000 and the ammonium nitrate explosion at Toulouse in 2001. The current instrument is Directive 2012/18/EU, called Seveso III, which has applied since June 1, 2015. It aligns its substance categories with the European classification and labeling regulation and strengthens the public’s right to information and to take part in decisions. Anyone citing a specific article should read the consolidated text on EUR-Lex, the EU’s official legal database, rather than a summary.

Duty to demonstrate and duty to disclose: Seveso and the US approach

I covered the origins of the American right-to-know system in the Bhopal article, so this section compares the two designs rather than retelling that history. Since 1982 the European model has started from the operator’s duty to demonstrate. An operator above the thresholds must analyze its own major-accident hazards, write them down, and satisfy a competent authority designated by the member state that it has done so adequately. Public information and off-site planning are built around that core. The Emergency Planning and Community Right-to-Know Act of 1986 starts from disclosure instead. Facilities report the hazardous chemicals they hold and the releases they have, and state emergency response commissions and local emergency planning committees use that information to write plans that the public can also see.

In the United States the duty to analyze hazards sits in separate programs. OSHA’s Process Safety Management standard, 29 CFR 1910.119, requires covered employers to perform process hazard analyses framed around worker safety. EPA’s Risk Management Program, established under section 112(r) of the Clean Air Act as amended in 1990, requires covered facilities to analyze off-site consequences, including a worst-case release scenario, and to submit risk management plans to EPA. The Seveso directives put all of this in one instrument, with a single line of accountability from the operator to the competent authority. The American system spreads the same work across several agencies and statutes. Neither arrangement is automatically better, and both depend on regulators with enough staff to read what operators submit.

The American answer to the lost days at Seveso is the release notification requirement in EPCRA section 304 and in CERCLA. A facility must report a release above a reportable quantity immediately: to the LEPC and the state under EPCRA, and to the National Response Center under CERCLA. That requirement is only as useful as the facility’s ability to say what actually escaped. For a working emergency manager, the information about a hazardous facility exists, but it is split up. Part of it is in a Tier II inventory filed with the state and the LEPC. Part is in an RMP filed with EPA, if the facility is covered. The rest is in process safety documentation kept on site for OSHA, and nobody hands you a single document like a Seveso safety report. Seveso also exposes a limit shared by both systems. An inventory lists what a facility stores, and the TCDD at Meda was never stored, because it formed in an upset. That hazard shows up only in the facility’s own process hazard analysis, which is why talking with the plant’s process engineers matters more than the filing.

What to do at your agency

• Have your LEPC chair or emergency manager pull the Tier II reports for the five facilities in your jurisdiction with the largest chemical inventories, check which of them also file an EPA risk management plan, and record both facts on one page the EOC can find at 2 a.m.
• Have your hazmat team leader ask the facility emergency coordinator at one RMP-covered plant this month for the decomposition and side-reaction products identified in its process hazard analysis, and add any that are not on the Tier II report to that facility’s preplan.
• Have the emergency manager write one paragraph into the existing hazardous materials annex naming the agency that collects environmental samples after a release, the laboratory that analyzes them, and the expected turnaround time, confirmed by a phone call to each.
• Have the communications center supervisor confirm that the dispatch center has a written procedure for receiving an EPCRA section 304 release notification, including who is paged and where the call is logged, and test it this month with one facility’s emergency coordinator.
• Have the public information officer add one paragraph to the existing public information annex describing what will be said in the first hour when the identity of a released substance is unconfirmed, including a plain statement that it is unconfirmed.
• Have the training officer add a Seveso segment to the next scheduled hazmat refresher that covers the runaway, the open relief line, the sixteen-day evacuation delay and the zone map, and that states plainly that no immediate deaths were attributed to the release.

Takeaways

• On July 10, 1976, a runaway reaction in a trichlorophenol reactor at the ICMESA plant in Meda burst a rupture disc and vented a cloud containing TCDD over Seveso and neighboring towns.
• The accepted technical explanation is that hotter than normal steam heated the reactor wall after the batch was stopped for the weekend, and the relief line discharged straight to the atmosphere with no containment.
• Estimates of how much TCDD escaped range from hundreds of grams to several kilograms, and no single figure is settled.
• The company did not identify TCDD in its first notice, and the first evacuation from the most contaminated zone began about sixteen days after the release.
• Authorities mapped Zones A, B and R from soil sampling, evacuated Zone A, and imposed food and farming restrictions on many thousands of residents in the outer zones.
• No immediate human deaths were attributed to the release, but thousands of animals died, about 80,000 were slaughtered, and chloracne was documented in a number commonly reported as 193 people, while long-term health findings remain qualified.
• Directive 82/501/EEC, followed by Seveso II in 1996 and Seveso III in 2012, requires operators to demonstrate safety to a competent authority, while the American system splits disclosure, worker process safety and off-site consequence analysis among EPCRA, OSHA and EPA.